LEARNING BY BREWING: HOMEBREWING LEGALIZATION AND THE BREWING INDUSTRY.
Amendment XXI, ratified in 1933, repealed Prohibition and made the commercial production of beer and other alcoholic beverages legal again in the United States, although it left it to the states to allow and regulate brewing, vinting, and distilling within their borders. Importantly, the amendment solely omitted homebrewing, the brewing of beer at home for personal consumption, from the list of legal activities. The reason for this omission has long been debated, with explanations ranging from a clerical error to, the more likely, purposeful omission for tax revenue reasons. As such, brewing beer at home for personal consumption was illegal and regulated by the Bureau of Alcohol, Tobacco, and Firearms.
From 1933 to 1978, 13 states affirmed the right to homebrew in spite of the federal ruling (Appendix). In 1978, President Carter signed H.R. 1337 which legalized home-brewing, although federal law deferred to state statutes. At that time, only an additional nine states opted in to legalize homebrewing. The remaining 28 states gradually legalized homebrewing over the next 35 years, with Alabama and Mississippi being the last in 2013.
With respect to homebrewers, the impacts of such regulation extend beyond just the recreational and consumptive benefits of homebrewing alcohol. Restricting homebrewing also limits the growth of a "beer culture" or a "beer community" where people come together to enjoy homebrewing and share their respective wisdom. This in turn, limits the development of knowledge and skills important to the growth of any productive industry. In this way, the act of home-brewing appears to epitomize the definition of learning-by-doing. To cite Arrow in his seminal work on the implication of learning-by-doing (Arrow 1962, 155): "Learning is the product of experience. Learning can only take place through the attempt to solve a problem and therefore only takes place during activity." We postulate that states that legally restricted homebrewing may have hindered the development of future brew-masters and therefore the expansion of their own brewing industry. (1)
That human capital development could lead to productivity and economic growth is not a new concept in economics. Lucas (1988) and Romer (1990) advocate that the accumulation of knowledge is a key determinant of economic growth. Lucas even suggests that learning-by-doing is of equal importance to formal education in explaining human capital development. Importantly, the literature emphasizes internal knowledge spillovers that raise productivity internal to the firm and external spillovers that flow to other firms in the industry.
Across a variety of industries, we find empirical support for both of these internal and external spillovers. Irwin and Klenow (1994) find significant internal learning rates in the semiconductor industry which are larger than knowledge spillovers. Thornton and Thompson (2001) find similar findings in World War II shipbuilding. Foster and Rosenzweig (1995) show that as Indian farmer experience increases, barriers to adopting new technology decline. Further, farmers with more experience and with experienced neighbors are significantly more profitable.
Moretti (2004) explores the productivity of manufacturing plants across cities. He finds that plants located in cities where the percentage of college graduates grew faster also experienced larger increases in productivity than similar plants in cities with less growth in college graduates. Levitt, List, and Syverson (2013) estimate the impact of learning-by-doing in automobile factories. They find a strong effect of learning-by-doing on both quantity and quality of production. Further, they suggest that such outcomes become institutionalized under a managerial organizational structure that promotes learning from mistakes.
The purpose of this paper is to explore whether state legislation legalizing homebrewing had an impact on the growth of state brewing industries across the United States from 1970 to 2012. As such, we differ from the previously mentioned learning-by-doing research as we focus on state outcomes rather than firm-level effects. Consequently, our work is more closely related to Alston et al. (2011) and Andersen and Song
(2013) who find a long-run relationship between investments in agricultural research and development at the state level and multifactor productivity in agriculture. In addition, Elzinga, Tremblay, and Tremblay (2015) specifically examine other factors contributing to the recent expansion in the craft beer industry. Our work attempts to provide greater insight into how learning via homebrewing impacted the brewing industry.
Our research hinges on the assumption that when states change homebrewing laws, more people begin to practice homebrewing. Furthermore, we expect that such learning-by-doing facilitates the expansion of the brewing industry. There are several reasons why we think these assumptions are not overly heroic.
With respect to learning-by-doing (i.e., internal spillovers), homebrewing allows product development and experimentation with low investment costs. If there are no legal restrictions to homebrewing, would-be brewers are only restricted by their time, creativity, and access to inputs. Failure in homebrewing only equates to suffering through a bad batch of beer, as opposed to lost profits. To this point, there are no legal requirements for becoming a brewmaster and any person can be successful in the brewing industry if they have learned to make a beer of sufficient quality. Further, legalization would facilitate access to higher quality and a wider range of ingredients (Hanson, McCullough, and Berning 2016). This is akin to an increase in technological capital and would increase the quality and diversity of beers brewed at home.
In terms of external spillovers, homebrewing and homebrewing clubs facilitate a community of idea sharing among like-minded individuals. Evidence of such comradery can be seen in the over 2,000 homebrewing clubs registered with the American Homebrewers Association (AHA; some of which operated illegally before state legislation changed), as well as numerous state craft brewer guilds, craft beer festivals, and in the numerous craft beer collaboration projects where two or more craft breweries develop and market a collaborative new beer.
Homebrewing was built upon a growing desire to create alternatives to mass-produced beer in the United States. For example, one influence on the growth of brewing in the United States came from abroad as American soldiers returned home from World War II and various Cold War deployments. Some of these servicemen came together to homebrew to recreate foreign styles of beer which differed greatly from the mass-produced American Light Lager (Corzine 2010). Removing legal restrictions on home-brewing allowed people to explore more creative beer recipes. Such diversity would also create more market opportunities.
Anecdotally, many observers credit the current success of the craft beer industry to early home-brewers who "lead the way" (Elzinga, Tremblay, and Tremblay 2015). The AHA, which formed prior to changing the homebrew laws in 1978, estimates that 90% of all professional brewers began as homebrewers (Talking Points 2015). In fact, Charlie Papazian, one of the founders of the AHA, taught two notable brewers in the early days of homebrewing: one of the founders of Boulder Brewing and one of the cofounders of New Belgium Brewing Company. Ken Grossman at Sierra Nevada and Jim Koch at the Boston Beer Company, two prominent craft breweries in the United States, also proudly proclaim their homebrewing roots. Grossman was also the proprietor of a homebrew supply shop before starting a brewery.
The heterogeneous rollout of homebrewing legislation provides us an opportunity to empirically examine this research question using a relatively long and wide panel dataset. We gather information identifying when each state legalized homebrewing and pair that with several measures of states' brewing industries. Since homebrew legalization could be endogenous to the number of breweries, we estimate a geographically dependent instrumental variables model with fixed effects.
Our results indicate that by removing home-brewing restrictions, states benefited via the expansion of their brewing industry. Specifically, we find significantly more breweries per million people and a larger growth rate of breweries per million. When we separate out breweries by employee size, we find the greatest change in "small" breweries, followed by "medium" breweries, but no change in "large" breweries. Finally, we find significant change in craft beer production but not in total beer production. These last two findings highlight growth in the craft industry, which is where we would expect homebrewers to have the largest impact.
This research provides insight into understanding how policies that, intentionally or unintentionally, restrict innovation and learning-by-doing might restrict long-term industry growth as well. This is particularly relevant for industries that can benefit from gains in efficiency and new product development, like the brewing industry.
II. EMPIRICAL SPECIFICATION
We start by specifying a linear model:
(1) [mathematical expression not reproducible]
[Y.sub.it] is used to evaluate the beer industry in each state using three different specifications. First, we calculate the number of breweries per capita in each state i from time t = 1970 to 2012. This measures the growth of the entire beer industry. Next, we separate breweries by employee size as reported by the U.S. Census. To our knowledge, there is no clear way to define small or medium breweries. We arbitrarily designate breweries with fewer than 20 employees as small, with 20 to 500 employees as medium and breweries with over 500 employees as large. We discuss different specifications of brewery size in the results section. Finally, we compare total state barrel production using data from the Beer Institute's Brewers Almanac (2013) with craft brewery production which is taken from Elzinga, Tremblay, and Tremblay (2015).
We estimate two potential effects of repealing homebrewing restrictions: an intercept shift which we specify as Legal, and a slope change given by Time. Since there is a likely lag effect of legalization, we specify Legal = 1 for the year after state legalization and 0 otherwise. (2) The variable Time is the number of years since the state legalized homebrewing. We speculate that as time goes by, the number of people homebrewing would increase which would lead to an accumulation of human capital via learning-by-doing and thus a greater supply of capable "brewmasters." As the pool of skilled labor increased, we expect there would have been an increase in the rate of growth of breweries in states that legalized home-brewing. As previously discussed, the legalization of homebrewing would also improve access to ingredients and knowledge, further enhancing growth in the industry. (3)
Market is a set of market regulations that we control for that could also influence the growth rate of the brewing industry. These variables vary over time and across states and are primarily obtained from the Alcohol Policy Information System (APIS) (National Institute of Alcohol Abuse and Alcoholism 2018). Brewpubs are a type of restaurant brewery that sells 25% or more of its beer on site (Brewers Association 2016) and could compete with the brewing industry. To control for the effect of changes in brewpub laws, we include a dummy variable to indicate when state laws changed to allow for this specific type of restaurant. Certain states require the registration of kegs sold which could create a transaction cost for customers and limit the type of production breweries engage in. We identify such keg laws using a dummy variable. Many states have had Sunday sales bans which have restricted or limited sales of alcohol on Sundays. These laws are identified using a dummy variable as well.
Some states have restrictions on the level of alcohol by volume (ABV) being produced by breweries. Importantly, this is different than restrictions on ABV being sold in stores to consumers as it only affects production. While limited ABV at both the retail and production side both have the potential to affect a brewery's decision to locate, the production restriction is more likely to limit growth of breweries. We include the max producer ABV, which we also obtain from APIS.
States also have various laws restricting their vertical relationships. Typically, states require three-tier distribution of beer from brewers to wholesalers to retailers. Certain states allow for self-distribution (generally brewers under a certain level of production), which we identify with a dummy variable. In addition, some states make it more difficult to exit relationships with wholesalers, often referred to as state franchise laws. We identify states with such state franchise laws in place to protect wholesalers using a dummy variable. The data we use to identify states laws on vertical relationships come from Burgdorf (2016).
We also control for several economic factors. First, we include the log of state gross domestic product (GDP) to control for changes in the state economy. Second, we consider potential bias from cross-state spillover. As homebrewing laws change in adjacent states, the flow of human capital could cross state borders. Furthermore, it is plausible that a state surrounded by states with a culture that supports a large number of breweries per capita is likely to have a similar culture and an increased number of breweries itself. Simply put, the expansion of the brewing industry in the United States is geographically dependent. Failing to account for this spillover could bias our estimates of the legalization of homebrewing.
Previous literature has addressed the problem of estimating interstate autocorrelation, or state spillover effects, in a fixed effects panel data setting through the use of spatial weighting matrices (Debarsy and Ertur 2010; Driscoll and Kraay 1998; Yu, de Jong, and Lee 2012). To capture spatial spillover, we account for all adjacent states' breweries per capita. A state spillover variable is created by calculating for state i, the total breweries per capita of neighboring states combined:
(2) [mathematical expression not reproducible]
where the weighting matrix [w.sub.i,j] takes the value 1 when state i physically borders state j, and 0 otherwise. Finally, [[beta].sub.i] and [delta] are parameters to be estimated, [[theta].sub.i] are state fixed effects, [[lambda].sub.t] are year effects, and [[epsilon].subit] is the residual error term.
A common approach for estimating the impact of a rollout of some policy change over time and space is a difference-in-differences specification. A concern with our analysis is the potential endogeneity of the legalization of homebrewing. Even though state fixed effects control for time-invariant unobservables, we cannot rule out the possibility that unobservable time-invariant state-level characteristics are correlated with both legalization of homebrewing and state breweries per capita. This could arise for several reasons. For one, homebrewing is most likely occurring because of a critical mass of a beer drinking population. As a result, states may have legalized homebrewing to meet their residents' call for changes. At the same time, breweries could be opening to meet expected market demand. These unobservable market trends would lead to bias in our model.
As the number of breweries in a state grows, their collective lobbying efforts may also grow as well. A group of breweries may want to encourage a beer culture and therefore help to lobby for changes in alcohol laws such as homebrewing restrictions. Consequently, there could be reverse causality between the number of breweries and the change in homebrewing laws. Alternatively, homebrewing could act as a substitute for craft beer production. As such, the growth of the industry could lead to greater pressures to restrict the legalization of homebrewing. In either case, these unobserved factors would bias our estimates of legalization.
We employ instrumental variable estimation to control for potential endogeneity of legalization in our specification. The challenge is to find a significantly strong instrumental variable that is also excludable from our primary equation. As an example, Stanig (2015) uses an instrumental variables approach to control for potential endogeneity when examining how the regulation of free press affects the coverage of corruption in Mexico. Specifically, he uses the severity of penalties for other criminal offenses as an instrument. He argues that criminal offenses are a measure of how punitive the laws are in general and therefore correlated with laws restricting free press. At the same time, such laws on criminal offenses are plausibly excludible from his primary equation.
Similarly, we attempt to find an instrument that measures a state's willingness to pass legislation in favor of individual rights. Our assumption is that states that are more likely to favor individual rights in previous legislation will also be more likely to allow for homebrewing when given the green light from the federal government.
For our instrument, we use the number of years since each state repealed their antimiscegenation laws. In the United States, antimiscegenation laws prohibited the marriage of members of different races. In general, this meant the restrictions of whites marrying blacks, Native Americans, and Asians, although there was variation in the laws from state to state. There were 9 states that never had such laws in place and 11 states that repealed the law some time before 1887. For such states, we count the time since becoming a state. From 1948 to 1967, 13 states repealed their laws before the Supreme Court invalidated all state antimiscegenation laws in 1967. This resulted in 17 states being forced to repeal their laws.
Like homebrewing laws, antimiscegenation laws were motivated by intolerance of individual freedoms. And both laws were legislated at the state level. States that were quicker to change their antimiscegenation laws, or never had them in place, were more socially progressive and therefore we argue that they would be more likely to change their homebrewing laws as well. In our empirical specification, we expect the time since a state repealed antimiscegenation laws to be positively correlated with the decision to repeal homebrewing laws. We measure the statistical strength of our instruments and discuss this in the results section.
Antimiscegenation laws have nothing to do with commercial brewing and are therefore excludable from our primary equation. A more challenging concern is whether our instrument is uncorrelated with the error term. One argument could be that certain social institutions that lead to the propagation of antimiscegenation laws also limit the growth of the brewing industry; for example, intolerance of social freedoms motivated by the strength of religious or political institutions. If this is correct, antimiscegenation laws would not be independent of the error term in our empirical model. We consider this not to be the case, however. First, our error term has state fixed effects that might be correlated with our instrument differenced out. In addition, we control for states with Sunday sales bans which were a part of religiously motivated blue laws, effectively controlling for states with a stronger view on temperance.
In addition, antimiscegenation laws were a product of racism and the exercising of power by whites over other minorities. However, there does not appear to be a strong link between any specific religious, regional, or political affiliation and antimiscegenation laws in the United States. In fact, the first antimiscegenation laws were passed in Maryland in the late 1600s. Browning (1951) argued that there was no clear categorical explanation for antimiscegenation legislation as the prohibitions enacted by the laws varied from state to state.
There could be other social phenomena not observed by the researcher that are correlated with antimiscegenation that could also affect commercial brewing. Again, this would mean antimiscegenation would not be independent of the error term. However, social phenomena that affected commercial brewing had their greatest impact via the temperance movement and the Eighteenth Amendment. In our dataset, prohibition was already repealed via the Twenty-first Amendment, so such social phenomena were not likely to impact commercial brewing directly. Finally, there could be some other social phenomena that we are not aware of which links antimiscegenation to the number of commercial breweries. We expect that state fixed effects will help to account for such unobservable effects.
Another potential concern is whether using antimiscegenation repeal as an instrumental variable also satisfies the monotonicity condition required to estimate the local average treatment effect (LATE) of the policy change (Angrist and Pischke 2009). In our data, 13 states changed their homebrewing laws prior to the federal mandate in 1978. Three of these states (Indiana, Nebraska, and Oregon) repealed their antimiscegenation laws after repealing their homebrewing laws. This would effectively make these states "defiers" under the monotonicity condition.
We expect that states that allowed homebrewing prior to the federal mandate would be unaffected by the federal change, however. As such, for these 13 states, we specify Legal = 0. This allows us to estimate the effect of the policy change on those states that waited to meet the federal guidelines, that is, the "compliers." This also then redefines the three states that were "defiers" as "never takers" in the parlance of treatment effects estimation. Essentially, they were never impacted by the change in homebrewing policy and are excluded from the estimate of the LATE effect on states that decided to repeal their homebrewing laws.
To employ our instrumental variable approach, we estimate the first stage of our model:
(3) [mathematical expression not reproducible]
where for state i, [Anti.sub.it] is the number of years since the repeal of antimiscegenation laws in year t, [X.sub.kit] are the other independent variables specified in Equation (1), y are parameters to be estimated, and [[micro].sub.it] is the residual series. Our final specification of Equation (1) therefore uses the first-stage estimate [Legal.sub.it]
Given the nature of the growth of breweries across the United States, there may exist temporal autocorrelation, a unit root, within each state that could bias our parameter estimates. We investigate this possible confounding factor by testing for temporal autocorrelation within a panel data setting. We test for temporal autocorrelation using the Levin-Lin-Chu test for balanced panels (Levin, Lin, and Chu 2002).
Results of the Levin-Lin-Chu test suggest that there are states that do not contain a unit root, they are stationary or are stationary about a trend, and that there are states that do indeed contain a unit root. To control for the potential presence of a unit root in our estimation, we take the first difference of the dependent variable in Equation (1). Importantly, this changes the interpretation of the estimated coefficients to the impact on the change in the dependent variable, or the impact of legalization on the growth of breweries. For this reason, we estimate two sets of models and compare results: one with breweries per capita as the dependent variable and one with the annual change in breweries per capita as the dependent variable.
We gather data from several sources to create a state-level panel. The number of brewing establishments by state is collected from the U.S. Census Bureau. From 1970 to 1997, the U.S. Census Bureau classified all industries using Standard Industrial Classification (SIC) codes. Since 1998, they have used the North American Industry Classification System (NAICS). We compile the number of breweries using SIC 2082 and NAICS 312120 (definitions are provided in Table 1) from 1970 to 2012. (4) According to the Census Bureau, these two codes are in concordance. The Census also identifies the number of employees for each brewery. We use these designations to categorize our breweries by the previous defined categories: small, medium, and large.
As can be seen in Figure 1, there has been a steady decline nationally over the past four decades in breweries we classify as large. At the same time, we observe substantial growth in small breweries and relatively steady growth in medium-sized breweries.
The Brewers Association, a nonprofit trade association of brewers, also counts the number of brewers in the United States (Brewers Association 2016). Their count follows the same pattern as Census data, but is larger. Importantly, the Census data only counts company headquarters and avoids counting multiple locations owned by the same company. In addition, the Census data may also neglect to count nanobreweries (a brewery producing a few barrels a year) because of their limited size and number of employees. Finally, brewpubs are not included in Census definition of Breweries. While there are differences in the data, the Census provides a consistent and frequently used count of the number of establishments for each state.
To compare the number of breweries across states, we divide the number of brewing establishments by the estimated Census population in millions for each state. We plot the number of breweries per capita for each state over time (see Appendix S1, Supporting information). In addition, for a visual reference, we have provided an Appendix S1 that maps the change in the count of breweries and the legalization of homebrewing across decades. Certain states exhibit significant growth in breweries per capita such as Alaska, Montana, and Vermont. Many other states have near flat growth in breweries per capita over the 42-year period. Some states exhibit signs of a downturn roughly corresponding to the U.S. economy as well, for example,
Colorado, Idaho, Oregon, Vermont, and Washington. Other states appear to be more resilient or unaffected.
We collected information identifying the year that homebrewing was legalized through a careful review of each state's statutes (Appendix), which we indicated in our state graphs using a vertical line and in the online maps (Appendix S1). Casual observation suggests that pre- and post-legalization trends were heterogeneous across states. For instance, after legalization, breweries per capita in Alaska grow significantly, whereas other states (e.g., Colorado, Idaho, and Montana) appear to have a lag between legalization and industry growth. Pretrends for some states also reveal a decline in breweries, such as Idaho and New Hampshire. Finally, some states exhibit a decline after legalization or modest change. Importantly, this visual analysis confirms that difference-in-difference estimation may not be appropriate with this analysis.
There also appears to be some interesting intraregional variation within the data. Specifically, we consider state groups defined by their Bureau of Economic Analysis (BEA) regions (Table 2). In none of the BEA regions do states legalize homebrewing within the same few years. This indicates that there was not meaningful coordination of legalization between states. There are significant differences in the volatility of breweries per capita as well. For example, New England, Vermont, and Maine exhibit more ups-and-downs, whereas Connecticut remains fairly stable over time. And while there appears to be great variation in the Southeast region, the scale is such that increases/decreases are in relatively small terms.
We next aggregate the state brewery data into their BEA economic regions and compare the difference between the number of breweries and the number of breweries per capita (Figure 2). This provides an interesting perspective on how breweries have evolved across BEA regions. The Far West has the most number of breweries overall, which is not surprising since it includes California, home of one of the oldest operating craft breweries in the United States, Anchor Brewing Company. The Far West is third, however, behind the Rocky Mountains and New England when adjusted by population, regions generally known for their love of craft beer and early homebrew legalization advocates (Hanson, McCullough, and Berning 2016). While the total number of breweries in the Great Lakes, Plains, and Southeast appear to be growing, they do not keep pace with the other regions.
Aggregated by region, a dip in the number of breweries that corresponds with the economic downturn in 2007 is easier to observe. The effect is most evident in the Far West and Rocky Mountains, but has limited effect in other regions such as the Great Lakes, Plains, and Southeast. Since the economic recovery, however, almost every region appears to exhibit near exponential growth in total breweries and per capita breweries. This is representative of the extensive growth of the craft beer industry in the United States over the past few years, even as larger brewers have experienced decreasing sales in recent years (Redding 2013). To this point, the Brewers Association (2015) recently reported that an additional 1,000 breweries were added from 2014 to 2015.
Our production data come from two different sources. Elzinga, Tremblay, and Tremblay (2015) provide craft production in 10,000 bbls (5) by state from 1979 to 2012 (see the Readme file to their online addendum for calculation specifics). Total brewery production in barrels by state from 1970 to 2012 comes from the Beer Institute's Brewers Almanac (2013). The Beer Institute data reflect shipments or sales of malt beverages at the wholesale level within a given state. Figure 3 illustrates the growth in craft beer production during a slow decrease in total beer production.
A. Fixed Effects Models
We first estimate Equation (1) for the number of breweries per capita without the first-stage instrumental variable (Table 3). Column 1 does not include state fixed effects, clustered standard errors, or spatial spillovers. Columns 2 and 3 include state and time fixed effects and errors clustered by state and column 3 includes the spatial autocorrelation spillover effect. Column 4 estimates the first difference of per capita breweries as the dependent variable. The clustered errors are calculated using 1,000 bootstrap iterations.
In the simplest model (1), we find that legalizing homebrewing had a positive effect on the number of breweries per capita. Nationally, the effect corresponds to an average increase of roughly 0.372 breweries per 1 million people in each state. It is relevant to note that this is not a measure of a 1-year jump in breweries immediately following the homebrew legalization. This number represents the average treatment effect (ATE) on breweries over the course of our sample period. In these terms, the impact estimated in this model seems relatively modest yet consistent with anecdotal evidence.
Most of the market variables are significant. Legalizing brewpubs is negative, which could indicate market competition. Restrictions on Sunday sales, producer ABV, and franchise laws all appear to restrict the growth of breweries per capita. Interestingly, keg registration which we expected to be restrictive is correlated with a greater number of breweries per capita. Self-distribution, which allows brewers to by-pass wholesalers, is also associated with greater breweries per capita.
After accounting for state fixed effects and clustered standard errors in column 2, we find the effect of legal, time since legalization, and the market variables are not statistically significant. The spillover of other states changing their laws is significant in column 3. This reveals the important geographic effect of neighboring states on the commercial development of breweries. Specifically, as states start to increase their commercial brewing their neighbors are more likely to follow suit. Based on Figure 1, there are likely to be limits to this spatial effect or more well-defined spatial boundaries, but we just capture the nearest neighbor effect. Finally, we first difference the dependent variable to control for a potential unit root (column 4) and find that none of the variables of interest are significant. As the GDP grows, we do observe an increase in the growth rate of number of breweries.
B. Instrumental Variable Models
We next estimate our instrumental variable specification including state and year fixed effects as well as cluster the standard errors by state. The first-stage estimate (Table 4, column 1) suggests that the time since changing the antimiscegenation laws is positively correlated with the legalization of homebrewing as we expected. Specifically, the longer the time since a state has eliminated antimiscegenation laws, the more likely they are to change their homebrewing laws. The Angrist-Pischke (AP) multivariate F-test is also significant, indicating the endogenous variable is identified.
In the second stage (Table 4, column 2), we find that after controlling for potential endogeneity, the legalization of homebrewing has a positive effect on the average number of breweries per capita. The estimate suggests roughly 7.1 breweries per 1 million people. This is the LATE over our entire sample period, which measures the change in breweries per capita for states that choose to overturn homebrewing legislation. This effect persists even after controlling for state spillover effects and other market factors which are not significant. The time since legalization does not have a statistically significant effect in this model.
We then estimate our model using the first difference as the dependent variable (Table 4, columns 3 and 4). The AP F-test suggests the instrumental variable performs well for this specification as well. The estimates show that legalization and time since legalization both have a significant effect on the change in the number of breweries per capita. This indicates that growth in this industry is increasing over time and at an increasing rate. Again, this is consistent with our examination of the data which reveals an exponential growth pattern. It also appears that states enforcing Sunday sales bans have significantly less growth than those that do not. This could be due to potential breweries perceiving such restrictions as representing a less beer-friendly culture and therefore choosing to open elsewhere or not at all.
It is important to note that our LATE estimates are relevant as we are interested in those groups of states that went along with the federal law to allow homebrewing. Additionally, in our data time frame, all but two of the states legalize homebrewing.
C. Other Measures
We next consider how changes in home-brewing laws affect breweries of different sizes. A priori, we expect people that start off as homebrewers would enter the industry as a small or medium-sized brewery, that is, with few employees per our measure of small, medium, and large. For one, the capital costs are prohibitive to enter as a large brewer (over 500 employees). Further, the type of brewing by small and medium brewers tends to be more consistent with the creative efforts of homebrewers, that is, a more diverse product line with many seasonal and one-off variations (Berning and McCullough 2017).
We start by estimating the same first-stage model. Specifically, the spillover variable is a measure of the spatial effect of all breweries, not just those of a specific size. Then, we estimate the second-stage model using the number of small, medium, or larger breweries per capita as the dependent variable. As can be seen in Table 5 column 2, the change in homebrewing laws had a significant effect on the number of small breweries. There were roughly 5.6 more breweries per 1 million people. Furthermore, the number of breweries is growing over time, consistent with Figure 1. Looking at medium-sized breweries (column 3), we find that the effect of legalization is smaller and not growing significantly over time. Still, the effect is significant which tends to confirm our assumptions that homebrewing legalization would impact these sized breweries the most.
There is no significant change in the number of larger breweries per million people following changes in homebrewing laws (column 4). Given the approach of these larger breweries, mass production, and economies of scale, this is not surprising. In addition, GDP only had a significant impact on the number of small breweries. This could indicate that smaller breweries are less able to weather economic downturns, or that the growth in smaller breweries is countercyclical due to other market factors. This would need to be explored further. In all, it appears that the overall growth we observe in breweries due to the change in homebrewing laws more directly impacts the smaller-scale breweries.
We also estimated models with different definitions of small, medium, and large (results available upon request). Importantly, the same finding holds that smaller breweries experienced the most growth and medium-sized breweries experienced less growth. Depending on the definition, large breweries might have experienced some growth, but this was relatively small.
We next compare barrel production for all breweries and just craft breweries as defined by Elzinga, Tremblay, and Tremblay (2015). In Table 6, column 2, we see that total production did not change significantly following legalization of homebrewing. Interestingly, we do find that there is a very small negative change in the production levels over time following changes in legalization. This is most likely reflecting the overall decrease of total beer consumption, see Figure 3. At the same time, craft production increases significantly following the legalization of homebrewing. The estimated impact is roughly 85,000 barrels per million people. Further, craft barrel production is increasing over time following homebrewing laws, suggesting a growing industry, which is consistent with industry trends. Altogether, this suggests a change in the composition of industry production in states that legalized homebrewing. However, craft production is still only a small portion of total market share.
An interesting marketing factor is Sunday sales bans which appear to favor large-scale production and work against craft production. This could have something to do with the distribution channels used by large brewers versus craft brewers. Specifically, larger brewers have wider distribution, whereas craft brewers have traditionally been more limited by their distribution channels. Therefore, any reduction in sales opportunities could drastically impact craft brewers compared with large brewers.
Similarly, states with franchise laws appear to favor large brewers. This could be that large brewers have more bargaining power with distributors and can benefit from their distribution capabilities. Finally, the legalization of brewpubs appears to hurt craft beer production; again, this is consistent with the idea that craft is competing with brewpubs.
The results of our analysis indicate that the legalization of homebrewing has had a significant effect on the average number of breweries per capita and on the growth rate in the number of breweries as well. We also find that smaller breweries grew more than medium-sized breweries following legalization, and large breweries did not grow at all. Finally, production of craft beer appeared to flourish following legalization whereas total beer production did not change.
While there are other factors that have contributed to the increase in breweries throughout the United States, these results suggest that home-brewing laws did in fact play a role in the growth of breweries at the national level. In fact, we find that few of the market factors play a significant role in our models. We also find no evidence that the number of breweries is spatially autocorrelated. This indicates that the effects of changing regulation within a state are somewhat isolated to that one state.
This analysis is the first of its kind and helps to validate prohomebrewing arguments made by organizations such as the AHA as an economic driver outside of the hobby itself. The effect of legalizing homebrewing has broader implications than just a few additional breweries per year. Breweries have large impacts on their regional economies by creating jobs, supporting input supply industries, as well as other local service industries. The economic impact has been measured at $55.7 billion dollars and over 424,000 jobs nationwide in 2014 (Brewers Association 2017).
Russian River Brewing Company is just one example where their annual release of the coveted beer Pliny the Younger brought in an estimated $2.4 million in 2013 to the county (Swindell 2015). Importantly, their brewmaster, Vinnie Cilurzo, learned the trade as a homebrewer in California shortly after the state legalized homebrewing. "Experimentation and a ton of reading, taught Cilurzo virtually all he needed to know about brewing" (McMorrow 2007). There are numerous other anecdotes of brewers learning how to brew beer at home, illustrating the importance of understanding the impact that this change in legislation has had on the industry.
A relevant consideration missing from our analysis is to what extent such growth spilled over into input industries. Historically, the majority of hops have come from the Pacific Northwest and malts from the Inland Northwest and Canada and varieties have been controlled by large brewery demand. But over the past decade, states are increasing their acreage and variety in hops and malting barley. In addition, technological innovation in fermentation has been growing to suit the needs of a wider range of brewers; specifically, designing brewing systems for smaller-scale breweries. Altogether, such factors can have a significant impact on state growth.
In addition to economic growth, there is also potential gain of consumer welfare. If variety is a normal good, the growth of more breweries is likely to provide more choices to consumers as each brewery must find a way to compete in a heterogeneous product market.
The craft brewing industry is a noteworthy example of a cottage industry gone commercial. Indeed, very few modern industries have demonstrated the ability to scale from the home-production level like brewing beer. Our analysis shows how removing restrictions that prohibit the development of human capital can contribute to the development of a commercially viable industry. While one cannot draw the conclusion that the mere legalization of homebrewing was the main driver for the existence of the beer brewing industry as it is today, one can say that it would not exist in its current fashion without such political action.
The implications of our research may extend to other industries that rely on learning-by-doing and whose effects extend beyond political boundaries via spillovers. While we provide an initial estimate of the impact of a policy change on economic growth, there are opportunities for further studies to elaborate. Specifically, the rate of growth associated with learning-by-doing is likely to be scale dependent. Finer-level data identifying the size of operations (i.e., scale) and extent of operations (i.e., scope) could provide greater insight. At least one paper (Berning and McCullough 2017) suggests that craft brewers may exhibit economies of scale and scope through their product line offering. This has further implications for changes in policy for industry development and consumer welfare through increased product offerings.
Additionally, the level of data aggregation to the state level misses economic interdependencies that are important for the growth of many industries. In particular, the interdependence of the upstream and downstream supply chain would suggest other economic spillovers associated with the expansion of breweries, for instance, for hop and grain suppliers or for beer distributors. In addition, firm-level data would allow us to measure more directly factors associated with learning-by-doing and learning spillovers. Finally, we do not address in this paper the appropriate lag of the impact of legalizing homebrewing. Presumably, our impact estimates could be more precisely measured with a better understanding of how long it takes homebrewing knowledge to generate commercial enterprise and for those enterprises to grow from small to medium size.
TABLE A1 Homebrewing Legislation Timeline Year of Year of State Legislation State Legislation Indiana 1933 Texas 1983 Massachusetts 1933 Minnesota 1985 Oregon 1933 Colorado 1986 Rhode Island 1933 Alaska 1989 Washington 1933 New Jersey 1991 Wisconsin 1934 Georgia 1993 Nebraska 1935 Arkansas 1995 Kansas 1949 Missouri 1995 Iowa 1971 North Dakota 1995 Hawaii 1972 Connecticut 1996 Virginia 1972 South Carolina 1996 North Carolina 1973 Michigan 1997 Maryland 1977 South Dakota 1997 Arizona (a) 1978 Tennessee 1997 California 1978 Delaware 1998 Kentucky (b) 1978 Illinois 1998 Maine (b) 1978 New Hampshire 1998 Montana (a) 1978 Pennsylvania 1998 Nevada (a) 1978 Vermont 1998 New York (a) 1978 Idaho 1999 Ohio (b) 1978 Utah 2009 West Virginia (b) 1978 Louisiana 2010 Wyoming 1979 Oklahoma 2010 Florida 1980 Alabama 2013 New Mexico (a) 1981 Mississippi 2013 (a) Active "Default to Federal Law" states that explicitly defer to federal law in their state statutes. (b) Passive "Default to Federal Law" states that have no reference to homebrewing in their state statutes.
Alston, J. M. M. A. Andersen, J. S. James, and P. G. Pardey. "The Economic Returns to US Public Agricultural Research." American Journal of Agricultural Economics, 93(5), 2011, 1257-77.
Andersen, M. A., and W. Song. "The Economic Impact of Public Agricultural Research and Development in the United States." Agricultural Economics, 44(3), 2013, 287-95.
Angrist, J. D., and J. S. Pischke. Mostly Harmless Econometrics. Princeton, NJ: Princeton University Press, 2009.
Arrow, K. J. "The Economic Implications of Learning by Doing." Review of Economic Studies, 29(3), 1962, 155-73.
Beer Institute. "Brewers Almanac, Electronic Database." 2013. Accessed January 2013. http://www.beerinstitute.org/multimedia/brewers-almanac/.
Berning, J., and M. P. McCullough. "Product Line Extension among New England Craft Breweries." Agricultural and Resource Economics Review, 46(1), 2017, 73-86.
Brewers Association. "Number of Breweries, Web Page." 2015. Accessed November 2015. https://www.brewersassociation.org/statistics/number-of-breweries/.
--. "Craft Beer Industry Market Segments, Web
Page." 2016. Accessed January 2016. https://www.brewersassociation.org/statistics/market-segments/.
--. "Economic Impact, Web Page." 2017. Accessed
January 2017. https://www.brewersassociation.org/statistics/economic-impact-data/.
Browning. J. R. "Anti-Miscegenation Laws in the United States." Duke Bar Journal, 1(1), 1951,26-41.
Burgdorf, J. E. "Essays on Mandated Vertical Restraints." All Dissertations, Paper 1659, 2016.
Corzine, N. M. "Right at Home: Freedom and Domesticity in the Language and Imagery of Beer Advertising 1933-1960." Journal of Social History. 43(4), 2010, 843-66.
Debarsy, N., and C. Ertur. "Testing for Spatial Autocorrelation in a Fixed Effects Panel Data Model." Regional Science and Urban Economics, 40, 2010, 453-70.
Driscoll, J. C, and A. C. Kraay. "Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data." Review of Economics and Statistics, 80(4), 1998,549-60.
Elzinga, K. G., C. H. Tremblay, and V. J. Tremblay. "Craft Beer in the United States: History. Numbers, and Geography." Journal of Wine Economics, 10(3), 2015, 242-74.
Foster, A. D., and M. R. Rosenzweig. "Learning by Doing and Learning from Others: Human Capital and Technical Change in Agriculture." Journal of Political Economy. 103, 1995, 1176-209.
Hanson, J. L., M. P. McCullough, and J. Berning. "New West, Brew West: Home Brewing an Industry in the West." Journal of the American West, 55(2), 2016, 49-59.
Irwin. D. A., and P. J. Klenow. "Learning-by-Doing Spillovers in the Semiconductor Industry." Journal of Political Economy, 102, 1994, 1200-27.
Levin, A., C. F. Lin, and C. S. Chu. "Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties." Journal of Econometrics, 108,2002, 1-24.
Levitt, S. D., J. A. List, and C. Syverson. "Toward an Understanding of Learning by Doing: Evidence from an Automobile Assembly Plant." Journal of Political Economy. 121(4), 2013, 643-81.
Lucas, R. E. "On the Mechanics of Economic Development." Journal of Monetary Economics, 22(1), 1988, 3-42.
McMorrow, P. "Vinnie Cilurzo of Russian River Brewing Company." BeerAdvocale Magazine #9. (2007).
Accessed June 2016. http://www.beeradvocate.com/mag/1338/vinnie-cilurzo-russian-river-brewing-company/.
Moretti, E. "Workers' Education, Spillovers, and Productivity: Evidence from Plant-Level Production Functions." American Economic Review, 94(3), 2004, 656-90.
National Institute of Alcohol Abuse and Alcoholism. "Alcohol Policy Information System, Electronic Database." 2018. Accessed January 2018. http://www.alcoholpoIicy.niaaa.nih.gov/.
Redding, J. "2012 Industry Review." The New Brewer, 30(3), 2013.
Romer. P. M. "Endogenous Technological Change." Journal of Political Economy, 98(5.2), 1990, S71-102.
Stanig, P. "Regulation of Speech and Media Coverage of Corruption: An Empirical Analysis of the Mexican Press." American Jouranl of Political Science, 59(1), 2015, 175-93.
Swindell, B. "Pliny Release Brings Economic Windfall to Sonoma County." The Press Democrat, February 5, 2015.
Talking Points. "American Homebrewers Association, Web Page." 2015. Accessed July 2015. https://www.homebrewersassociation.org/homebrewing-rights/talking-points/.
Thornton, R. A., and P. Thompson. "Learning from Experience and Learning from Others: An Exploration of Learning and Spillovers in Wartime Shipbuilding." American Economic Review, 91(5), 2001, 1350-68.
Yu, J., R. de Jong, and L. Lee. "Estimation for Spatial Dynamic Panel Data with Fixed Effects: The Case of Spatial Cointegration." Journal of Econometrics, 167, 2012, 16-37.
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Appendix S1. States with legalized homebrewing and number of breweries 1970-2012
MICHAEL MCCULLOUGH, JOSHUA BERNING and JASON L. HANSON
McCullough: Associate Professor, Department of Agribusiness, California Polytechnic State University, San Luis Obispo. CA 93407. Phone 805-756-5009, Fax 805-756-5040. E-mail firstname.lastname@example.org
Boning: Associate Professor, Department of Agricultural and Resource Economics, Colorado State University, Fort Collins, CO 80523. Phone 970-491-6325, Fax 970-491-2067, Eemail@example.com
Hanson: Chief Creative Officer & Director of Interpretation and Research, History Colorado, Denver, CO 80203. Phone 303-866-4762, Fax 303-866-2796, E-mail firstname.lastname@example.org
ABV: Alcohol by Volume
AHA: American Homebrewers Association
APIS: Alcohol Policy Information System
ATE: Average Treatment Effect
BEA: Bureau of Economic Analysis
GDP: Gross Domestic Product
LATE: Local Average Treatment Effect
NAICS: North American Industry Classification System
SIC: Standard Industrial Classification
(1.) While various breweries have their own requirements for becoming a brewmaster, we are only referring to a brew-master as someone who is in charge of brewing beer at a commercial brewery.
(2.) We do not explore the length of the lag, which is discussed more in Alston et al. (2011). We discuss this more in the results section.
(3.) It would be beneficial to more directly measure changes in the homebrewing industry, e.g., the number of homebrewing clubs, members, and shops offering products. However, the American Homebrewers Association has limited data on membership since 2002.
(4.) Data before 1998 were acquired from the National Archives Catalog at research.archives.gov
(5.) A standard barrel (bbls) is 31 US gallons.
TABLE 1 Definition of Industry Classification System Classification System Code Definition SIC 2082 Establishments primarily engaged in manufacturing malt beverages. Establishments primarily engaged in bottling purchased malt beverages are classified in Industry 5181. NAICS 312,120 This industry comprises establishments primarily engaged in brewing beer. ale. malt liquors, and nonalcoholic beer. Source: U.S. Census Bureau. TABLE 2 U.S. Census Regions Region States Far West AK, CA, HI, NV, OR, WA Great Lakes IL, IN, MI, OH, WI Midwest DE, MD. NJ, NY. PA New England CT, ME, MA, NH, RI Plains IA, KS, MN, MO, NE, ND, SD Rocky Mountains CO, ID, MT, UT, WY Southeast AL, AR, FL, GA, KY, LA, MS, NC, SC, TN. VA, WV Southwest AZ, NM, OK, TX TABLE 3 Fixed Effects Estimation Balanced Panel Variables 1 2 3 Dependent Variable yt yt yt Legal 0.372 (**) 0.37 0.363 Time since legal -0.00444 -0.00227 -0.009 Brewpub -0.657 (***) -0.0087 0.0771 Keg registration 0.464 (***) 0.523 0.204 Sunday sales ban -0.532 (***) 0.285 0.036 Producer ABV -0.980 (***) 1.266 (**) 0.15 Self-distribution 0.966 (***) -0.423 -0.738 Franchise -0.454 (***) -0.464 -0.439 Spillover 0.627 (*) GDP 0.0889 -3.515 (**) -3.714 (**) Constant 0.696 5.123 (**) 6.092 (***) Observations 2,150 2,150 2,150 R-squared-within 0.282 0.293 0.323 Between 0.197 0.0663 Overall 0.0453 0.0774 State fixed effects No Yes Yes Time fixed effects Yes Yes Yes Clustered SEs No Yes Yes Variables 4 Dependent Variable (yt-yt-1) Legal -0.0213 Time since legal 0.00159 Brewpub 0.00725 Keg registration 0.00424 Sunday sales ban -0.117 Producer ABV 0.257 (**) Self-distribution -0.145 (***) Franchise -0.0243 Spillover 0.0251 GDP -0.375 (***) Constant 1.979 (***) Observations 2,100 R-squared-within 0.135 Between 0.185 Overall 0.081 State fixed effects Yes Time fixed effects Yes Clustered SEs Yes Note: Clustered standard errors computed with 1,000 bootstrap iterations. (***) p < .01; (**) p < .05; (*) p < . 1. TABLE 4 Instrumental Variables Estimation with Panel Fixed Effects Variables Dependent Variable yt 1 2 3 Antimisc laws 0.0393 (***) 0.0392 (***) Legal 7.148 (***) Time since legal 0.0798 Brewpub -0.761 Keg registration 0.522 Sunday sales ban 0.553 Producer ABV -1.592 (*) Self-distribution 0.439 Franchise 0.344 Spillover 0.769 GDP -0.0652 Observations 2,150 2,100 (AP) F-test 70.95 (***) 69.62 (***) F-test 14.0 (***) Variables Dependent Variable (yt-yt-D 4 Antimisc laws Legal 0.963 (***) Time since legal 0.0147 (***) Brewpub -0.114 Keg registration 0.0126 Sunday sales ban -0.453 (***) Producer ABV -0.0879 Self-distribution 0.0342 Franchise -0.072 Spillover 0.337 GDP 0.0367 Observations (AP) F-test F-test 1.549 (***) Notes: (1, 2, 3) Clustered standard errors computed with 1,000 bootstrap iterations. (4) Conventional standard errors. (***) p <.01; (*) p<.1. TABLE 5 Instrumental Variables Estimation with Panel Fixed Effects by Brewery Employment Size Variables Dependent Variable Small Medium 1 2 3 Antimisc laws 0.0393 (***) Legal 5.640 (***) 1.504 (**) Time since legal 0.0848 (**) -0.00416 Brewpub -0.566 -0.215 Keg registration -0.0188 0.451 Sunday sales ban 0.437 -0.105 Producer ABV 0.606 0.379 Self-distribution -0.963 -0.106 Franchise -0.107 0.00271 Spillover 0.458 0.0678 GDP -1.266 (*) -0.291 Observations 2,150 2,150 2,150 (AP) F-test 70.95 (***) F-test 15.22 (***) 15.18 (***) Dependent Variable Large 4 Antimisc laws Legal 0.00427 Time since legal -0.000849 Brewpub 0.0203 Keg registration 0.00762 Sunday sales ban 0.0116 Producer ABV -0.216 Self-distribution -0.0175 Franchise 0.0388 Spillover -0.00437 GDP -0.0355 Observations 2,150 (AP) F-test F-test 50.3 (***) Note: Clustered standard errors computed with 1.000 bootstrap iterations. (***)p<.01; (**) p<.05; (*) p<.1. TABLE 6 Instrumental Variables Estimation with Panel Fixed Effects by Production Type Variables Dependent Variable Total Production Craft Production 1 2 3 Antimisc laws 0.0393 (***) 0.0404 (**) Legal -0.0144 Time since legal -0.00355 (**) Brewpub -0.0112 Keg registration -0.008 Sunday sales ban 0.0411 (*) Producer ABV 0.0122 Self-distribution 0.00974 Franchise 0.0250 (*) Spillover -0.0108 GDP 0.102 (***) Observations 2,150 1,700 (AP) F-test 70.95 (***) 88.98 (***) F-test 160.6 (***) Variables Dependent Variable Craft Production 4 Antimisc laws Legal 8.517 (**) Time since legal 0.242 (**) Brewpub -1.881 (**) Keg registration -1.023 Sunday sales ban -4.102 (*) Producer ABV -0.55 Self-distribution -1.349 Franchise -0.344 Spillover 0.255 GDP -1.626 Observations (AP) F-test F-test 15.30 (***) Note: Clustered standard errors computed with 1,000 bootstap iterations. (***) p<.01; (**) p <.05; (*) p<. 1.
|Printer friendly Cite/link Email Feedback|
|Author:||Mccullough, Michael; Berning, Joshua; Hanson, Jason L.|
|Publication:||Contemporary Economic Policy|
|Date:||Jan 1, 2019|
|Previous Article:||HOW EFFECTIVE ARE EXPERT TV HOSTS AT SAVING FAILING BUSINESSES?|
|Next Article:||ECONOMIC FREEDOM AND INCOME LEVELS ACROSS U.S. STATES: A SPATIAL PANEL DATA ANALYSIS.|