Content uploaded by Colen Liesbeth
Author content
All content in this area was uploaded by Colen Liesbeth on Jun 07, 2019
Content may be subject to copyright.
1
Economic Growth, Globalization and Beer Consumption
Liesbeth Colen and Johan Swinnen1
Abstract
We analyze the evolution of beer consumption between countries and over time. Historically, there
have been major changes in beer consumption in the world. In recent times, per capita consumption
has decreased in traditional beer drinking countries while it increased strongly in emerging
economies. Recently, China has overtaken the US as the largest beer economy. A quantitative
empirical analysis studies the relation between economic growth, globalization and beer
consumption. The relationship between income and beer consumption has an inverse U-shape. Beer
consumption initially increases with rising incomes; but at higher levels of income beer consumption
falls. Increased globalization has contributed to a convergence in alcohol consumption patterns
across countries. In countries that were originally beer drinking countries, the share of beer in total
alcohol consumption reduced, while this is not the case in countries which traditionally drank mostly
wine or spirits.
Keywords: beer, consumption patterns, taste convergence, economic growth, globalization, country
and global region trends.
JEL Classification: N30, D12, Q11
1 Liesbeth Colen is associate researcher at LICOS, University of Leuven (KU Leuven) and researcher at the Joint
Research Centre (JRC) of the European Commission. Johan Swinnen is professor of economics and Director of the
Centre for Institutions and Economic Performance (LICOS) at the University of Leuven (KU Leuven). The authors wish
to thank Julian Alston, James Seale, Damiaan Persyn as well as three anonymous referees and the editor of this journal
for very useful comments on earlier versions of this article. The sole responsibility for the content of this article lies with
the authors. The views expressed are purely those of the authors and may not in any circumstances be regarded as stating
an official position of the European Parliament or the European Commission. E-mail: Jo.Swinnen@kuleuven.be for
correspondence.
2
Economic Growth, Globalization and Beer Consumption
1. Introduction
Globally, beer is the most consumed alcoholic drink. There have been major changes in beer
consumption across countries and over time.2 While total beer consumption has increased
significantly over the past fifty years, consumption has declined in “traditional beer drinking
countries” such as the UK, Germany and Belgium. Growth has come from other countries. For
example, Russia has witnessed a dramatic shift from vodka to beer consumption over the past decade
and a half, while countries like Spain and Italy have seen wine consumption decline and beer
consumption increase. Beer consumption in emerging economies has increased rapidly, reflected by
China overtaking the USA in 2003 as largest beer market (Swinnen 2011).
The questions we address in this paper are how does economic growth and globalization
affect these changing patterns of beer consumption? While there have been several studies analyzing
the income elasticities of beer consumption, these studies have typically focused on a single country
or a group of high income countries over relatively short periods of time (see Fogarty (2010) and
Selvanathan and Selvanathan (2007)). In this paper we take a broader and longer time perspective:
beer consumption for more than 80 countries over a period of 50 years. Our dataset includes many
low and middle income countries, some of which have experienced strong growth over the past
decades, and includes times of rapid globalization, reflected in liberalization of borders and growing
trade and foreign investment flows.
2 See Nelson (2005), Poelmans and Swinnen (2011) and Unger (2004) for a more elaborate analysis of the economic
history of beer.
3
The paper is organized as follows. We first document changes in beer, wine and spirits
consumption across countries over the past 50 years, and then discuss conceptual issues and the
literature on the determinants of beer consumption. Next we specify a formal econometric model to
analyze the determinants of beer consumption across times and nations, focusing specifically on the
role of economic growth and globalization. Our final section concludes.
2. Trends in Beer Consumption Since 19603
Globally, beer consumption is more important than wine or other alcoholic beverages (Figure 1). In
volume terms, the gap has grown strongly over the past 50 years. While in the 1960s the global
volume of beer was approximately double that of wine, beer consumption was more than seven
times larger than wine consumption by 2009. Since wine and other alcohol are typically more
expensive than beer, the differences in value terms are smaller than in volume terms (Figure 1).
Between 1960 and 1990 the global value of beer and spirits was roughly equal and approximately
double the value of wine. Over the past twenty years, the value of beer has continued to increase
while that of spirits has stagnated. By 2009, beer accounted for about $168 US billion, and spirits
and wine respectively for about $124 billion and $67 billion. In terms of the volume of pure
alcohol4, beer and spirits are almost equal in recent years (FAOstat, 2014).
For most of the past 50 years5 the US was the largest beer market, consuming around 25
billion liters per year for the past 30 years. In the 1960-1980 period the other large markets were
Germany and the UK. However, consumption of beer has declined significantly in the past 25 years
3 For more details, see Colen and Swinnen (2011).
4 The conversion into liters of pure alcohol is based on the following alcohol contents (% alcohol by volume): beer (5%),
wine (12%), spirits (40%) (WHO 2013). We acknowledge that the alcohol content may have changed considerably over
time, as pointed out by Chikritzhs et al. (2010), but we lack the data to account for this.
5 Note that some countries might not report accurately on alcohol consumption (e.g China during the 1960s, some
muslim countries where alcohol consumption is prohibited), resulting in unreliable numbers.
4
in Germany and the UK and stagnated in the US. Growth in demand is concentrated in the emerging
countries. From close to zero beer consumption as recently as 1980, China overtook the US as the
largest beer market in 2003,6 growing to about 40 billion liters by 2009 (FAOstat, 2014). In Russia
and Brazil beer consumption has increased strongly over the past two decades and today these
countries are larger beer markets than Germany. In all these countries the combination of income
growth and economic liberalization has been associated with a dramatic growth of beer production
and consumption. Also in India, there has been substantive growth in beer consumption in recent
years (Arora et al. 2011).
Clearly total consumption is strongly influenced by population size while per capita beer
consumption also varies across countries. Per capita beer consumption is still the highest in Europe.
The greatest beer drinkers are the Irish and the Czech with about 150 liters per capita per year in
2009 – much more than any other country (FAOstat 2014). Until 2005 Austrians Germans, Belgians
and the British also drank at least 100 liters per capita, although lately this number has come down
considerably. Beer consumption in Belgium and the UK has fallen below 75 liters per capita. The
highest consumption in non-European countries is in Australia and the US (both about 85 liters per
capita). In most of these countries, per capita beer consumption has been declining substantially for
several decades. The maximum consumption per capita was reached in 1974 in Belgium, in 1980 in
the UK, in 1981 in the US, and in 1983 in Germany. The decline is significant: in Germany and
Belgium beer consumption declined from close to 150 liters per capita in the 1970s to respectively
100 and 74 liters per capita by 2009.
Per capita consumption in expanding beer markets is still considerably less: 70, 55 and 26
liters per capita in respectively in Russia, Brazil, and China in 2009. Growth in per capita beer
6 See Bai et al. (2011) for the fascinating story of what happened in the Chinese beer market.
5
consumption has been strongest in Russia, more than quadrupling between 1995 and 2009, stronger
even than in China. In Brazil, growth was strongest in the 1975-1995 period and has slowed since.
Per capita consumption was less than 1 liter in India in 2009.
These trends are also reflected in the share of beer in total (pure) alcohol consumption, which
has been changing significantly between 1965 and 2009 (see Table 1). In many traditional beer
drinking countries such as Belgium, the UK, Germany and Czech Republic, the relative share of beer
in total alcohol consumption is declining and that of wine increasing. In the UK the share of beer fell
from 81 percent to 36 percent, while the share of wine rose from 4 to 33 percent.
Conversely, in traditional wine drinking countries such as Spain and Argentina, the share of
wine is declining and that of beer increasing – and substantially so. In Spain, beer consumption
increased from 11 percent in 1965 to 50 percent in 2009, effectively overtaking wine consumption,
at 21 percent in 2009, compared to 65 percent in 1965. Traditional spirit drinking countries have also
increased beer consumption. In 1965 the Russians, Polish and Chinese consumed most of their
alcohol in the form of spirits, with the beer share beer respectively 15, 28 and 1.5 percent. By 2009,
the share of beer had increased strongly: to 37, 55 and 31 percent of total alcohol consumption
respectively. Countries appear to be converging in their alcohol consumption.
3. Determinants of Beer Consumption: Conceptual Issues, Literature and Descriptive Statistics
Economic theory predicts that demand for beer is a function of the price of beer, the prices of
substitutes and complements, income, product characteristics, and consumption preferences (Stigler
and Becker 1977; Tremblay and Tremblay 2005). Also the fact that beer, like other alcoholic
beverages, is potentially addictive, will affect the demand for beer. For an addictive commodity,
current consumption may depend on past and expected future consumption, and a number of studies
6
have proposed an addiction model to study the demand for alcoholic drinks (Becker and Murphy
1988; Akerlof 1991; Waters and Sloan 1995). Also peer pressure and social status may influence the
consumption of (types of) alcoholic beverages (Lee and Tremblay 1992; Akerlof and Kranton 2000;
George 2011; Deconinck and Swinnen 2015). Finally, concerns over the addictive aspects and
potential negative health effects play a role in consumer decisions (Cutler and Lleras-Muney 2010)
and also government regulations to restrict alcohol use through taxes or limitations on sales will
affect beer consumption (Lee and Tremblay 1992; Okrent 2010; Nye 2011).
There are several studies on beer consumption in specific countries, analyzing its
responsiveness to changes in prices, taxes and income, and the role of government regulations on
consumption and advertising (see e.g. Fogarty 2006, Tremblay and Tremblay 2005; Wagenaar et al.
2009). Our aim is to study how changes in the consumption of beer across the globe and over a long
time period relate to changes in income. As far as we are aware, there are no existing econometric
studies analyzing changes in global beer consumption across many countries, including low and
middle income countries, and over a long time period.
Some studies do provide a cross-country analysis for high income countries. Fogarty (2006,
2010), Gallet (2007), Tremblay and Tremblay (2005) present results from meta-analyses of more
than 150 studies, most of which are on OECD countries. Selvanathan and Selvanathan (2005a, 2007)
estimate income elasticities for a sample of 10 high-income countries.7
Although the income elasticity estimates for the individual countries vary considerably, most
studies indicate that the demand for beer is inelastic (elasticity smaller than one). Based on a meta-
regression analysis, Fogarty (2010) estimates an average income elasticity of 0.69. Selvanathan and
7 Several of these studies also estimate price elasticities. The estimates for the average own-price elasticities are -0.33
(Fogarty 2010) and -0.19 (Selvanathan and Selvanathan 2005a). The estimated cross-price elasticities are small. For
example Selvanathan and Selvanathan (2005a) estimate a cross-price elasticity of 0.03 and 0.16 for wine and spirits
respectively.
7
Selvanathan (2007) estimate the income elasticity for the pooled sample of 10 high-income countries
to be 0.66, with country estimates varying between 0.37 and 1.3. For Japan, they find that beer is a
“luxury good”, i.e. with an income elasticity larger than one. A few studies also find negative
income elasticities (e.g. Nelson (2003) for the US).
The non-linear relation between beer consumption and income
The trends we described above indicate that in middle- and low-income countries that are
experiencing growth, such as China, Russia, Poland and India, beer consumption grows. In rich
countries, however, further income growth corresponded to a reduction in beer consumption per
capita. These observations suggest a non-linear relationship between income and beer consumption.
Figure 2 graphically represents trends in beer consumption against the evolution in real per capita
income in several countries, and clearly suggests an inverted U-shaped relation with income.8
However, very few previous studies discuss or analyze the potential non-linearity of the
relation between beer consumption and income, virtually all the empirical models used in the
literature impose a constant income elasticity. We found only four studies that consider income
elasticity varying with income. The first two analyze beer consumption across states or individuals.
Hogarty and Elzinga (1972) analyze the demand for beer across US states and find that income
elasticities depend on the income level. States with higher income levels have lower income
elasticities compared to poorer states. Gil and Molina (2009) apply a QUAIDS model to the
consumption of alcoholic beverages among Spanish youth in 2000 and, across individuals, they find
an inverted-U shape for the relation of beer consumption with income. A third study, by Lee and
Tremblay (1992), analyzes US beer consumption over time (1953-1983) and allows the income
elasticity to vary with income. They find that the income elasticity declines as US per capita income
8 The pattern is robust for correcting for PPP for those years and countries for which PPP is available.
8
rises: from 0.189 in 1953 to 0.105 by 1983. The fourth study consists of a meta-analysis of demand
elasticity estimates from a total of 141 individual-country studies. Fogarty (2010) plots the estimated
income elasticities of beer from each of the studies against the time period of analysis, and shows
that the estimated income elasticities follow an inverted-U shape9.
Several possible explanations can be offered for the decline of beer consumption with
income growth at higher levels of income. First, there appear to be an increased awareness of and
concerns about the potential negative health effects of alcohol consumption as income rises. As is
the case for other health-related consumption behaviour, such as smoking or obesity (Cutler and
Glaeser 2005; Lakdewalla et al. 2005), higher levels of income result in a higher demand for alcohol,
but at the same time also in an increased demand for health10, thereby reducing the demand for
alcohol. If the health effect becomes relatively more important as income rises, the combined effects
lead to a non-monotonic relationship of beer consumption with income. Similar non-monotonic
relations have indeed been found for smoking and obesity (Cutler and Glaeser 2006; Philipson and
Posner, 2008).
The more information and the stronger the beliefs on the harmful effects of drinking alcohol,
the stronger is the health effect (Cutler and Glaeser 2005). Lee and Tremblay (1992) argue that
throughout the 1970s and 1980s the health risks of alcohol consumption became more apparent. In
response, consumers may have decided to decrease their individual beer consumption. Cutler and
Glaeser (2006) find a negative relation between the share of smokers and income across countries,
and attribute this largely to differences in the information and beliefs on the health consequences of
smoking across countries. Moreover, Cutler and Lleras-Muney (2010) document that higher
9 On average, income elasticities increased by 0.62 between 1904 and 1956 and declined by 0.33 between 1956 and
1994, along with increasing incomes.
10 The increased demand for health with higher incomes is consistent with the higher value of future life with higher
incomes (Grossman 1972; Cutler and Glaeser 2005).
9
educated and richer people are likely to give more importance to the harmful effects of smoking and
drinking, contributing further to the inverted relation of beer consumption with income.
Governments have also responded to the increased health concerns. As countries have grown
richer and more worried about the health and social consequences of alcohol use, many countries
have increased taxes, and imposed limits on advertising and sales of alcohol, and have imposed laws
against drinking and driving (Lee and Tremblay 1992; Fogarty 2010; Okrent 2010). Picone et al.
(2004) and Gallet and Eastman (2007) find that smoking bans have not only reduced smoking, but
have also contributed to lower alcohol consumption.
The declining elasticity at higher income levels may also be related to the fact that there
exists an upper limit to the utility that can be derived from the consumption of certain commodities
(Brown and Deaton 1972). Hence, commodities may move from luxuries at low income levels to
necessities once a saturation level has been reached while income continues to increase (Lee and
Tremblay (1992)). This may also be reflected in a shift in preferences from beer to other alcoholic
beverages as incomes increase. Most studies find indeed that the income elasticities of wine and
spirits are considerably higher than for beer (Selvanathan and Selvanathan 2005a, 2007; Fogarty
2010), and are thus considered more luxurious. Especially in those countries where beer is the
traditional beverage, people may shift alcohol consumption towards wines when they become richer.
The role of globalization in the convergence of alcohol consumption patterns
Globalization is likely to have played a role in this shift from beer to wine in traditional beer-
drinking nations. Traditionally, climate, geology and history have strongly determined the
production (and thus consumption) of particular types of alcoholic drinks in specific parts of the
world. More recently increasing global economic integration has weakened the association between
10
production and consumption. In a detailed analysis on the convergence of wine and beer
consumption for 36 countries, Aizenman and Brooks (2008) show how grape production and latitude
largely explained the relative consumption of wine in 1963, but much less so in 2000, which they
link to globalization processes, and increasing trade in alcoholic drinks (e.g. Persyn et al. 2011;
Anderson 2004; Meloni and Swinnen 2013).
In addition to direct trading links, other forms of international economic interaction, such as
the spread of production technologies, ideas and capital, alter traditional production patterns. For
example, foreign direct investment has led to foreign take-over and a major restructuring of the
brewery sectors in Eastern Europe and Russia during the transition process of the 1990s (Van Herck
et al. 2012). Foreign investors bring capital and the knowledge of brewing, wine or spirits
production to countries were these beverages were traditionally not produced, thereby generating
domestic production (often under licence of a foreign company) of beer or other beverages. As a
result of the lower trade costs and increased domestic production, the relative price of beer to wine
moves closer to its international level in all trading countries, which implies a convergence in the
relative consumption patterns of alcoholic drinks (Leifman 2001).
Finally, cultural globalization contributes to a convergence in taste (Aizenman and Brooks
2008). As a result of increased tourism, migration and international communication, people associate
less with traditional cultural patterns, leading to a homogenization, or at least hybridisation, of life-
styles and taste (Leifman 2001), even though habit persistence may cause such assimilation to new
consumption patterns to be slower than might be expected on the basis of price convergence or
increased exposure and experience (Aizenman and Brooks 2008).
4. Empirical Estimation of the Determinants of Global Beer Demand
11
As mentioned in the introduction, and reflected in the literature review in the previous section, our
focus is on the role of economic growth and globalization on beer consumption. We use data for a
total of 81 countries, over a period of 31 years (1980-2010). The countries cover the world, except
for small island states and the African continent11.
We analyze the level of beer consumption per head, but also the relative importance of beer
with respect to the total consumption of alcohol. Therefore, in a first set of regressions, we use the
logarithm of adult per capita consumption of beer, expressed in liters of pure alcohol. For the second
part of the analysis, we calculate the volume share of beer, wine and spirits in total pure alcohol
consumption. We estimate both the determinants of differences in absolute and relative consumption
between countries and within countries over time and focus specifically on the role of per capita
income and globalization as main variables of interest.
The analysis of the demand for beer in a cross-country setting raises the question whether it
is appropriate to pool data for many countries and over a long time period. The assumption that the
same demand equation holds for all countries and thus imposes constant taste across consumers
worldwide and over time, was proposed by Stigler and Becker (1977)12. In our analysis, we test
whether such a common relation between income and beer consumption can be found across
countries, which form it takes, and what other factors play a role.
11 Taking into account missing observations in some of the explanatory variables for certain countries, we are left with an
average time period of 23 years by country. Although the main results still hold if the African countries are included, the
reliability of both consumption and price indicators are particularly problematic for these countries, so we excluded them
from the regressions. The list of countries used in our analysis can be found in appendix.
12 Comparing estimates for 10 countries, Selvanathan (1991) finds that this assumption of constant taste holds for the
aggregate category of alcoholic beverages, although for the individual alcoholic commodities, including beer, the
assumption is rejected (Selvanathan and Selvanathan, 2007). Based on the meta-analysis of a large set of individual
country studies on beer, Fogarty (2010) finds little support for the idea that the taste for beer varies fundamentally across
most countries. He concludes that the Stigler-Becker hypothesis of constant tastes across consumers in different
countries is an acceptable starting assumption.
12
Empirical model
The basic models of applied demand analysis are the Rotterdam model (Theil 1965) and the
(Quadratic) Almost Ideal Demand System (AIDS) (Deaton and Muellbauer 1980; Banks et al. 1997).
For example, Selvanathan (1991) and Selvanathan and Selvanathan (2005a) use the Rotterdam
model to analyze alcohol consumption in a set of ten countries. However, data limitations (in
particular the lack of good (or any) price data for a wide range of countries) constrain the application
of these models in our case. We tried various potential proxies for prices but all had problems, in
particular for non-OECD countries. Moreover, using Rotterdam or AIDS models with only poor
approximations of price variables would inflate measurement errors in the budget share indicators,
and thus in the estimation procedure.
As in other studies which were confronted with similar problems, we use the double-log
demand equation. For example, Selvanathan and Selvanathan (2005b) also use the double-log model
for the analysis of alcohol consumption patterns across countries. As stated by Alston et al. (2002), it
is well-known that this method cannot satisfy all restrictions from consumer theory (only the
homogeneity restriction can be imposed), but its ease of and straightforward interpretation explains
its continued popularity, especially for demand studies focusing on a single commodity.
We start from the following double-log demand equation where consumption is a function of
income and prices, and where the coefficients can be readily interpreted as respectively the income
elasticity and own-price and cross-price elasticities13. We extend this basic model by including a
second-order term for income to test the potential non-linear relationship with income, and by a
number of variables other than income and prices that may explain the adult per capita consumption
level of beer. This results in the following model:
13 The income variable and price proxies are deflated by a common price deflator. Since this price deflator is an index of
the prices of all other goods, deflating imposes homogeneity and generates the coefficients as Marshallian or
uncompensated income and price elasticities (Alston et al. 2002).
13
ln 𝑄𝑖𝑡 = 𝛼 + 𝛾1𝑙𝑛𝐼𝑖𝑡 + 𝛾2(𝑙𝑛𝐼𝑖𝑡)² + 𝜼′𝑙𝑛𝑷𝑖𝑡 + 𝜷1′𝒙𝑖𝑡 + 𝜷2′𝒛𝑖+ 𝑢𝑖𝑡
(1)
where 𝑄𝑖𝑡 is the amount of beer consumed per capita;
𝛼 the constant term;
𝐼𝑖𝑡 the income level per capita;
𝑷𝑖𝑡 the vector of price proxies for beer, wine, spirits and non-alcoholic beverages;
𝒙it𝒛𝑖 a vector of time-invariant explanatory variables; and
uit the error term, consisting of a time-invariant part 𝑣𝑖, a common time trend 𝜇𝑡,.and
remaining error 𝜀𝑖𝑡.
When analyzing the relative importance of beer consumption with respect to other alcoholic
beverages, we use a similar model. We use the share of beer in total alcohol consumption as left
hand side variable and the same set of explanatory variables is used, except for the raw price
variables, which are replaced by the price of beer relative to wine and to spirits.
We start by focusing on the variation in the level and share of beer consumption between
countries, using the between-effects estimator. Then, we implement the fixed-effects estimator to
explain the evolution of the level and share of beer consumption within countries over time. This
estimator controls for unobserved time-invariant error 𝑣𝑖 using the within transformation, and
provides consistent estimates on the determinants of the evolution of consumption within countries14.
The effect of variables that are invariant over time, such as climate or language, cannot be estimated
in this model15. In order to control for common time shocks, year dummies are included. We note
that simultaneity between demand and prices may cause biased estimates. We searched for a good
instrument to solve this problem, but could not find a satisfactorily solution. Although definitely not
14 The fixed effects model imposes common coefficients across countries, corresponding to our question whether a
common evolution in the beer-income demand exists and whether there is a general effect of globalization on beer
consumption. Alternatively one could estimate individual time series regressions for each country, after which the
estimated coefficients are averaged across countries, as the group of mean group estimators (e.g. Pesaran and Smith
1995) does. We have explored these models, but the time dimension of our panel is too short for these models to be
useful in our case.
15 Using the random effects (RE) model indeed would allow to estimate the effect of time-invariant variables and would
(in case valid) provide more efficient estimates compared to the within-estimator. Yet, an overidentifying restrictions test
(Wooldridge 2002; p.290-291) rejected the validity of the RE model.
14
being perfect instruments, we use lagged prices as instruments in a robustness test, which leads to
similar results.
We further extend the analysis of beer consumption over time by including the lagged level
or share of beer consumption as explanatory variables. In this way we account for habit formation
(Dynan 2000), which is likely relevant in the case of alcohol consumption16.
Data
The dependent variables are both taken from the Global Information System on Alcohol and
Health (WHO 2013). The adult per capita consumption of beer is expressed in liters of pure alcohol
and the share of beer in total pure alcohol consumption17 is calculated from the adult per capita
consumption of beer, wine and spirits. For income, we use GDP per capita from WDI (2013).
Information on the price of beer and its main substitutes is available only for a limited set of
countries. We therefore use a proxy indicator variable for the prices. Based on FAO data on import
values and volumes (FAOstat 2013), we calculate the unit import value for beer, wine, spirits and
non-alcoholic beverages and use this as a proxy. To test its validity, we have compared this proxy
with price indices for 27 European countries between 1996 and 2010 where detailed price data are
available as provided by Eurostat (2013). The correlation coefficients between the percentage
changes in the unit import value for beer, wine, spirits and nonalcoholic beverages and the
corresponding changes in Eurostat price data are 90% or higher over the period 1996-2010.18 This
16 While it is well-known that the inclusion of a lagged dependent variable may lead to bias, this bias is inversely related
to the time dimension of the panel (Nickell 1981) and is likely to be limited in our panel with medium-size time
dimension. As a robustness test, we have implemented the Arellano-Bond estimator which confirms our main results.
17 The share of beer in total alcohol consumption is only considered when the total consumption of alcohol in the country
at that time surpassed at least 1 liter of pure alcohol per year, otherwise it is set to missing. This was done in order to
avoid meaningless changes in the share of beer related to very minor changes in the almost zero consumption of beer,
wine or spirits.
18 Over the period 1996-2010, the correlation was 0.90 for the price of beer, 0.94 for wine, 0.92 for spirits and 0.94 for
non-alcoholic beverages, for a set of 27 European countries.
15
suggests that the indicator performs well for this subset of countries and time. However, the
indicator may well be less representative for prices in other countries, in particular lower income
countries with less trade in beer, or in earlier time periods with less trade.
The income and price variables are deflated to obtain the impact of real changes and to
impose the homogeneity restriction. GDP per capita and prices are deflated by the GDP deflator, but
the results are robust to deflating income and prices by the CPI index instead. Both deflators are
taken from WDI (2013). The 2005 PPP conversion factor (WDI 2013) is used to make prices and
income comparable across countries. Price and income variables are expressed in constant 2005
international dollars.
As a measure of globalization, we implement the widely used KOF Index of Globalization
(Dreher 2006)19. It consists of three indices measuring the economic, social and political dimensions
of globalization, based on actual economic flows, economic restrictions and data on information
flows, personal contact and cultural proximity.
Additional variables capture other potential determinants of beer consumption. Average
precipitation and temperature in a country are indicators for the climatic conditions that affect the
production of different alcoholic beverages. This influences the price and availability of beer in the
absence of international trade and aims to capture the long history of habit formation in drinking
specific beverages. While the production of cereals for beer requires a sufficient level of rainfall and
moderate temperatures, dryer and warmer temperatures favor the cultivation of grapes. Country-
level aggregated climatic data are taken from the Tyndall Centre for Climate Change Research
(Mitchell et al. 2004). In addition, climatic variables might also directly influence demand, by
conditioning taste for more or less, or different types of alcoholic beverages.
19 The index is published and can be downloaded: http://kof.ethz.ch/. We have used the 2014 version. As an alternative
indicator of globalization, we have used merchandize trade as a percentage of GDP (WDI, 2013), which leads to
qualitatively similar results.
16
A set of variables aims to measure the impact of religion and cultural factors on alcohol
consumption. Specifically, we use data on the share of different religions among the countries’
population in the year 1970, taken from the Religion Adherence Database constructed by Barro and
McCleary (2005). Religious adherence may change over time or with income, but no coherent time
series data on religious affiliations are available. We also include three dummy variables, indicating
whether a country has a Romanic, Germanic or other language as official language, taken from
CEPII (2013). This variable reflects the proposition that taste and habits are spread more easily
among populations with the same cultural background and related language. This variable may also
capture effects of colonization. Colonizers may have introduced their favorite alcoholic beverage
into the colonized territory, which may still have an impact on alcohol consumption patterns today20.
A statistical description of the variables is provided in Table A1 of the online appendix.
Results
Table 2 presents the results for the between effects regressions and indicates the determinants of the
large differences across countries in beer consumption and in the share of beer in total pure alcohol
consumption. Tables 3 and 4 show the results of the fixed effects regressions, which aim to explain
the changes in the levels of beer consumption (Table 3) and share of beer in total alcohol
consumption (Table 4) within countries over time. In order to account for potential
heteroskedasticity, autocorrelation, and cross-dependence in our data, we report standard errors that
are robust in each of these dimensions (Driscoll and Kraay 1998) for each of the fixed effects results.
Our results provide strong evidence for an inverted U-shaped relation between income and
per capita beer consumption, both in comparisons across countries and in the evolution over time.
20 We have also tested the inclusion of dummies for having been colonized by a traditional beer, wine or spirits drinking
country, which delivers similar results.
17
From the between effects regression results (Table 2), we see that richer countries consume more
beer. Yet, the second order income coefficient is significantly negative, indicating that countries with
even higher income levels have on average lower per capita beer consumption21. On the one hand,
prices are the result of the demand and supply of beer at country level and including these variables
may cause problems of multicollinearity with country-specific explanatory variables as climate and
religion that explain differences in respectively supply and demand. On the other hand, omitting
prices may cause omitted variable bias. We therefore include a specification without price variables
(1) and one with price variables (2).
The fixed effects regression results (Table 3) show that this non-linear relationship for
income holds not only between countries, but also within individual countries over time,
significantly so for the full sample, and also for Asian countries. 22 Hence, as per capita income in a
country rises, people tend to drink more beer and the income elasticity is positive. Yet once above a
certain income level, beer consumption tends to decline with further increases in income, implying a
negative income elasticity...
Based on the coefficients in Table 3 (2), we plot the relation between the estimated income
elasticity and income level in Figure 3. Note that the long-term impact of the variables in a
specification with a lagged dependent variable (Table 3 (2)) can be obtained by dividing the
estimated coefficients by (1-γ), where γ is the coefficient on the lagged dependent variable. This
results in a long-term impact of 0.74 and 1.51 for the first and second order income variable
respectively, which is slightly lower than the short-run elasticities estimated in the first column. This
is in line with expectations, as habit formation typically slows consumer’s reactions. Based on these
21 The turning point for the between effects model is estimated to be at about 33,000 USD based on the results in column
(1). Four countries in the sample (Switzerland, US, Norway and Belgium&Luxembourg) have an average income higher
than this level and would thus be (on average over the time period concerned) on the downward sloping part of the
(cross-country) income-consumption relationship.
22 The samples of countries in the Middle East and Australia are too small for separate regressions.
18
estimated long-run impacts, we calculate the turning point of the U-shaped relation with income, i.e.
the point where beer consumption starts declining with growing incomes, to be approximately
26,500 US dollars per capita, corresponding to the point where the income elasticity shifts from
positive to negative. This income level was reached by Belgium and the US in the 1970s, by
Germany in the early 1980s, and by the UK halfway the 1980s, and corresponds well to the
evolution shown in Figure 2.
For comparison of our income elasticity estimates with those in the literature, we have also
imposed a constant income elasticity using a linear income specification23. A constant average
income elasticity of 0.97 is estimated. A somewhat lower income elasticity of 0.66 was found by
Selvanathan and Selvanathan (2007) using a pooled demand equation for 10 high-income countries,
while several individual country studies have found an income elasticity of about unity (Clements
and Selvanathan 1991). Given that our non-linear specifications suggest higher income elasticities
for countries with a lower income level and given that the sample of Selvanathan and Selvanathan
(2007) consisted only of high-income countries, it is not surprising that our pooled estimate is larger.
When looking at the consumption of beer relative to other alcoholic beverages across
countries (Table 2), no clear link with income is found. Richer countries seem not to have a typical
preference for beer over other drinks, or the other way around. Yet, when looking at the evolution of
the share of beer over time (Table 4), a similar although less robust non-linear income effect is
found. Hence, the inverted-U shape that was found with respect to total beer consumption can at
least partially be explained by the shift to other alcoholic beverages as income increases above a
certain level, and not only by a decrease in total alcohol consumption when people get richer24.
23 Results available upon request.
24 Table 5 shows the regression with total alcohol consumption as dependent variable instead of beer consumption,
indicating that also for total alcohol consumption an inverted U-shape is found for the relation with income.
19
As mentioned above, increased information and increased concerns over the health risks of
alcohol use may provide an explanation for the lower consumption of beer. Unfortunately the role of
health concerns could not be explicitly measured because no good indicators to capture these health
concerns could be found for a large set of countries and over a longer time period, but as far as
differences across countries are time-invariant, they are captured by the fixed effects. The inclusion
of year dummies controls for the potential global evolution towards increased awareness of the
consequences of alcohol use. But if in certain countries concerns about the health and addictive
effects have been growing faster than in other countries, along with income growth (Cutler and
Glaeser 2006), then the non-linear income relation may reflect some of this effect. If this is the case,
we would expect the non-linear shape to be found equally well when analyzing the consumption of
total pure alcohol. Table 5 shows the results of a regression with total alcohol consumption instead
of beer consumption on the left hand side, and finds that this indeed the case, although the effect
seems less non-linear than for beer consumption. As a result, and as confirmed by the non-linear
shape found for the share of beer, the preference for beer compared to other alcoholic beverages
declines with rising incomes.
Our second important result concerns the impact of globalization on beer consumption. The
globalization variable is not significantly different from zero when analyzing the determinants of
beer consumption across countries (Table 2). When looking at the effect of globalization over time
(Tables 3 and 4), interestingly, the responsiveness of beer consumption to the increase in trade
openness is highly dependent on the countries' traditional pattern of alcohol consumption. When
interacting the globalization variable with the share of beer in total alcohol consumption in the year
1970, globalization is found to increase the level and share of beer consumption for countries with a
low initial share of beer consumption. For countries consuming traditionally more beer, globalization
20
tends to reduce beer consumption. This confirms Aizenman and Brooks’ (2008) finding that
globalization contributes to the convergence of alcohol consumption patterns across countries. This
effect seems to be largely driven by Europe, while being very imprecisely measured for the other
continents (Table 4).
For the other variables the results are largely as expected. In the between effects estimations,
price variables do not always have the expected sign25 (Table 2). In the fixed effects estimations the
results for the price variables, despite our concerns on the quality of the indicators, correspond
largely to what is found in the literature. The own-price elasticity of per capita beer consumption is
negative and significant with a value of -0.09 to -0.16 (Table 3 (1-2)), somewhat lower than the
averages from the meta-studies discussed above. In line with the literature, the estimated cross-price
elasticities are positive but small, and mostly not significantly different from zero. As expected, the
share of beer consumption declines as beer becomes relatively more expensive than wine or spirits
(Table 4).
The remaining explanatory variables do not vary over time and are thus only included in the
between effects specifications. When including these strongly correlated explanatory variables
jointly, the signs and magnitudes of the coefficients remain largely the same, but estimates become
more imprecise. The climatic variables indicate that beer consumption is higher in countries with
lower temperatures and more rainfall (expressed in 1,000 ml per year) (Table 2), corresponding to
the ideal growing conditions for barley. For the share of beer in total alcohol consumption we find
unexpectedly a positive relation with temperature when price variables are included, which may
25 This may be due to strong multicollinearity between the three price variables, or to the extrapolation of the 2005 PPP
conversion rate over a long time period. Since PPP conversion rates are not provided annually, but are only constructed
for specific years, we have used the 2005 PPP conversion factors and extrapolated these over the entire time period in
order to allow for cross-country comparison. Since consumption baskets may change considerably over time, one needs
to be cautious with interpreting the results of such extrapolation for large data series.
21
indicate that some other country specific characteristics may be captured through this variable. Not
surprisingly, also religion plays an important role in determining beer consumption. Countries with a
high share of Muslims consume less beer and beer consumption is larger in countries with a larger
share of Catholics and Protestants in the country (Table 2), though the latter coefficients are only
marginally different from zero. Catholics and especially Protestants also seem to have a preference
for beer over other alcoholic beverages, as shown in Table 5. Although coefficients are estimated
very imprecisely, the signs of the coefficients suggest that in countries speaking a Romanic
language, beer consumption is lower and less popular than other alcoholic beverages, compared to
countries speaking a Germanic language (Table 2).
Robustness tests
As indicated above, we performed a number of robustness tests to test the sensitivity of our
results to using the lagged values of prices as instruments, to using the Arellano-Bond estimator to
avoid potential bias related to the inclusion of the lagged dependent variable, to deflating income and
prices by the CPI index instead of the GDP deflator, to splitting up the sample in a group of low and
a group of high income countries, and to using the longer (50 years) data series of non-PPP
converted income indicators. For reasons of space, the detailed estimations results are not included
in the paper. Results of the first two robustness tests are provided in an online appendix (Table A2),
and remaining tests are available from the authors. In all cases our conclusions regarding the non-
linear relation with income and the role of globalization hold throughout these different
specifications.
Conclusion
22
We study the evolution of beer consumption across countries and over time. A historic
overview of the evolution of beer consumption in the world indicates that consumption of beer has
changed importantly over time. Over the past 50 years consumption patterns of beer have changed
substantially, with decreasing consumption in the traditional beer drinking countries and strong
growth in emerging economies.
We analyzed the determinants for beer consumption and estimated an empirical model to
explain beer drinking. Our first empirical result is that the relationship between income and beer
consumption is non-linear. Beer consumption initially increases with rising incomes, but at higher
incomes beer consumption falls with further income growth. Second, we find that in traditional beer
drinking countries, the share of beer in total alcohol consumption declines with opening of trade and
increasing globalization, while the reverse happens in non-beer drinking countries (at least in
Europe). These findings are consistent with the idea that there is convergence in the consumption
patterns of alcoholic beverages, as suggested in the literature.
Finally, other factors that can explain the different rates of beer consumption among
countries are the price of beer and other alcoholic beverages, climatic conditions, the importance of
different religions in the country, and strong cultural or historical ties with beer producing countries.
References
Aizenman, J. and E. Brooks (2008). Globalization and taste convergence: the cases of wine and beer.
Review of International Economics 6: 217-233.
Akerlof, G. A. and R. E. Kranton (2000). Economics and identity. Quarterly Journal of Economics
115: 715-753.
Akerlof, G.A. (1991). Procrastination and obedience. The American Economic Review, 81/2: 1-19.
Alston, J.M., Chalfant, J.A. and N.E. Piggott (2002). Estimating and testing the compensated
double-log demand model. Applied Economics 34: 1177-1186.
Anderson, Kym (ed.) (2004). The World Wine Markets. Globalization at Work. Edward Elgar.
23
Arora, A., Bhaskar, A., Minten, B. and A. Vandeplas (2011). Opening the beer gates: how
liberalization caused growth in India’s beer market. In: Swinnen, J. F. M. (ed.), The Economics
of Beer, Oxford: Oxford University Press.
Bai, J., Huang, J., Rozelle, S. and M. Boswell (2011). Beer battles in China: the struggle over the
world’s largest beer market. In: Swinnen, J. F. M. (ed.), The Economics of Beer, Oxford: Oxford
University Press.
Banks, J., Blundell, R. and A. Lewbel (1997). Quadratic Engel curves and consumer demand.
Review of Economics and Statistics 79/4: 527-539.
Barro, R.J. and R.M. McCleary (2005). Which countries have state religions? Quarterly Journal of
Economics 120/4: 1331-1370.
Becker, G. S. and K. M. Murphy (1988). A theory of rational addiction. The Journal of Political
Economy 96/4: 675-700.
Brown, A. and A. Deaton (1972). Models of consumer behavior, The Economic Journal, 82/328:
1145-1236.
CEPII (2013). Language database. http://www.cepii.fr/CEPII/en/bdd_modele/bdd.asp, last accessed
October 2013.
Chikritzhs, T.N., Allsop, S., Moodie, R. and W. Hall (2010). Per capita alcohol consumption in
Australia: Will the real trend please step forward? Medical Journal of Australia 193/10: 594-597.
Clements, K.W. and S. Selvanathan (1991). The economic determinants of alcohol consumption.
Australian Journal of Agricultural Economics 35/2: 209-231.
Colen, L. and J.F.M. Swinnen (2011). Beer Drinking Nations. The Determinants of Global Beer
Consumption. AAWE Working Paper No. 79, American Association of Wine Economists,
http://www.wine-economics.org/workingpapers/AAWE_WP79.pdf
Culter, D.M. and E.L. Glaeser (2006). Why do Europeans smoke more than Americans? NBER
Working Paper 12124.
Cutler, D.M. and E.L. Glaeser (2005). What explains differences in smoking, drinking and other
health-related behaviors? American Economic Review 95/2: 238-242.
Cutler, D.M. and Lleras-Muney A. (2010). Understanding differences in health behaviors by
education, Journal of Health Economics 29/2: 1-28.
Deaton, A. and J. Muellbauer (1980). An almost ideal demand system. American Economic Review
70/3: 312-326.
Deconinck, K. and J.F.M. Swinnen (2015). Peer effects and the rise of beer in Russia. Food Policy
51: 83-96.
Dreher, Axel (2006): Does Globalization Affect Growth? Evidence from a new Index of
Globalization, Applied Economics 38, 10: 1091-1110.
Driscoll, J. and Kraay, A. (1998), Consistent covariance matrix estimation with spatially dependent
panel data, Review of Economics and Statistics 80(4): 549-560.
Dynan, K. E. (2000). Habit formation in consumer preferences: Evidence from panel data, The
American Economic Review, e90(3): 391-406.
24
Eurostat (2014). Purchasing power parities (PPPs), price level indices and real expenditures for
ESA95 aggregates. http://epp.eurostat.ec.europa.eu/portal/page/portal/eurostat/home, last
accessed October 2013.
FAOstat (2014). Statistics of the Food and Agriculture Organization of the United Nations.
http://faostat.fao.org, last accessed October 2013.
Fogarty, J. (2006). The nature of demand for alcohol: understanding elasticity. British Food Journal
108/4 :316-332.
Fogarty, J. (2010). The demand for beer, wine and spirits: A survey of the literature. Journal of
Economic Surveys 24/3: 428-478
Gallet, A. (2007). The demand for alcohol: a meta-analysis of elasticities. Australian Journal of
Agricultural and Resource Economics 51/2: 121-135.
Gallet, C.A. and H.S. Eastman (2007). The impact of smoking bans on alcohol demand. The Social
Science Journal 44/4: 664-676.
George, L. (2011) The growth of television and the decline of local beer. In: Swinnen, J.F.M. (ed.),
The Economics of Beer, Oxford: Oxford University Press.
Gil, A.I. and J.A. Molina (2009). Alcohol demand among young people in Spain: an addictive
QUAIDS. Empirical Economics 36:515-530.
Grossman, M. (1972). On the concept of health capital and the demand for health, Journal of
Political Economy 80/2: 223-55.
Hogarty, T.F. and K.G. Elzinga (1972). The demand for beer. Review of Economics and Statistics
54/2: 195-198.
Lakdawalla, D., Philipson, T. and J. Bhattacharya (2005). Welfare-enhancing technological change
and the growth of obesity. American Economic Review 95/2: 253-257.
Lee, B. and V.J. Tremblay (1992). Advertising and the U.S. market demand for beer. Applied
Economics 24/1: 69–76.
Leifman, H. (2001). Trends in population drinking. In: T. Norström (ed.), Alcohol in Post-War
Europe: Consumption, Drinking Patterns, Consequences and Policy Responses in 15 European
Countries, Stockholm: National Institute of Public Health, 49–81.
Meloni, G. and J.F.M. Swinnen (2013). The political economy of European wine regulation. Journal
of Wine Economics 8/3: 244–284.
Mitchell, T., T.R. Carter, P. Jones, and M. Hulme (2004). A comprehensive set of high-resolution
grids of monthly climate for Europe and the globe: the observed record (1901-2000) and 16
scenarios (2001-2100). Tyndall Centre Working Paper 55, Tyndall Centre for Climate Change
Research, http://www.cru.uea.ac.uk/cru/data/hrg/.
Nelson, J. (2003). Advertising bans, monopoly, and alcohol demand: testing for substitution effects
using state panel data. Review of Industrial Organization 22/1: 1–25.
Nelson, M. (2005). The Barbarian’s Beverage. A History of Beer in Ancient Europe. UK: Routledge.
Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica, 49(6): 1417-1426.
25
Nye, J. V.C. (2011). Brewing nation: war, taxes and the growth of the British beer industry in the
18th and 19th centuries. In: Swinnen, J.F.M. (ed.), The Economics of Beer, Oxford: Oxford
University Press.
Okrent, D. (2010). Last Call: The Rise and Fall of Prohibition. New York: Scribner, Simon &
Schuster.
Persyn, D., Swinnen, J.F.M. and S. Vanormelingen (2011). Belgian beers: where history meets
globalization. In: Swinnen, J.F.M. (ed.), The Economics of Beer, Oxford: Oxford University
Press.
Pesaran, M. H. and R.P. Smith (1995). Estimating long-run relationships from dynamic
heterogeneous panels. Journal of Econometrics, 68(1): 79-113.
Philipson J.P. and R.A. Posner (2008). Is the obesity epidemic a public health problem? A review of
Zoltan J. Acs and Alan Lyles's "Obesity, business and public policy". Journal of Economic
Literature 46/4 974-982.
Picone, G.A., Sloan, F. and J.G. Trogdon (2004). The effect of tobacco settlement and smoking bans
on alcohol consumption. Health Economics 13: 1063-1080.
Poelmans, E. and J.F.M. Swinnen (2011). From monasteries to multinationals (and back): a
historical review of the beer economy. Journal of Wine Economics 6/2: 196-216
Selvanathan, E. A. (1991) Cross-country alcohol consumption comparison: an application of the
Rotterdam demand system. Applied Economics 23/10: 1613-1622.
Selvanathan, S. and E.A. Selvanathan (2005a). Empirical regularities in cross-country alcohol
consumption. The Economic Record 81/255: 128-142.
Selvanathan, S. and E.A. Selvanathan (2005b). Demand for beer, wine and spirits (Chapter 9). In:
Selvanathan, S. and E.A. Selvanathan (Eds.) The demand for alcohol, tobacco and marijuana.
International evidence. Ashgate publishing, pp. 211-242.
Selvanathan, S. and E.A. Selvanathan (2007). Another look at the identical tastes hypothesis on the
analysis of cross-country alcohol data. Empirical Economics 32/1: 185-215.
Stigler, G. J. and G. S. Becker (1977). De gustibus non est
disputandum. American Economic Review 67/2: 76-90.
Swinnen, J.F.M. (2011). The Economics of Beer, Oxford: Oxford University Press.
Theil, H. (1965). The information approach to demand analysis. Econometrica 33/1: 67-87.
Tremblay, V. J. and C.H. Tremblay (2005). The US brewing industry: Data and economic analysis.
MIT Press Books.
Unger, R.W. (2004). Beer in the Middle Ages and the Renaissance, University of Pennsylvania
Press.
Van Herck, K., J.F.M. Swinnen and K. Deconinck (2012). “How the East Was Won: Supply Chain
Restructuring in the Eastern European Beer Market”. Agrarwirtschaft. German Journal of
Agricultural Economics 61(4):213-222.
26
Wagenaar, A.C., Salois, M.J. and K.A. Komro (2009). Effects of beverage alcohol price and tax
levels on drinking: a meta-analysis of 1003 estimates from 112 studies. Addiction 104/2: 179-
190.
Waters, T.M. and F.A. Sloan (1995). Why do people drink? Tests of the rational addiction model.
Applied Economics 27/8: 727-736.
WDI (2013). World Development Indicators, http://databank.worldbank.org, last accessed October
2013.
WHO (2013). World Health Organization, Global Information system on Alcohol and Health,
http://apps.who.int/globalatlas/default.asp, last accessed October 2013.
27
Figure 1. Global consumption of beer, wine and spirits in (A) volume (billion liters) (a), (B)
volume of pure alcohol (billion liters) (b), and (C) value (billion US dollars) (c), 1961- 2009
Source : FAOstat (2014)
(a) Data on volumes in kg have been converted to liters, assuming 1 kg of liquid equals 1 liter.
(b) Volumes of pure alcohol are calculated using the conversion rates from WHO(2013): 5% for beer,
12% for wine, 40% for spirits.
(c) Values are calculated using the average of global import and export unit prices (derived from
trade volumes and values) multiplied by volume.
0
30
60
90
120
150
180
210
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
2006
2009
Billion litres
Volume
Beer
Wine
Spirits
0
1
2
3
4
5
6
7
8
9
10
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
2006
2009
Billion litres
Volume (pure alcohol)(c)
Beer
Wine
Spirits
0
30
60
90
120
150
180
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
2006
2009
Millions
Value
Beer
Wine
Spirits
28
Figure 2. Beer consumption per capita vs. income per capita (1,000 constant US dollars) in selected countries (1961-2009).
Source: FAOSTAT (2013) and WDI (2013).
0
20
40
60
80
100
120
140
160
0 5 10 15 20 25 30 35 40 45 50
Liter per cap
GDP per cap (1,000 constant USD)
China India Russia Brazil USA
Belgium Germany United Kingdom Japan Spain
29
Figure 3. Estimated income elasticity by level of income
Based on regression results in Table 3, model (2).
-1
0
1
2
010 20 30 40 50 60
Income elasticity
GDP per cap, PPP (1,000 international dollars)
30
Tables
Table 1. Share of beer, wine and spirits in total pure alcohol consumption (Source: WHO, 2013)
1965
2009
beer
wine
spirits
beer
wine
spirits
USA
45.8
10.6
43.6
51.1
16.6
32.3
Germanya
57.1
17.3
25.5
54.7
26.8
18.5
UK
81.0
4.3
14.7
36.3
32.8
30.9
Ireland
74.9
5.0
20.1
51.1
23.5
25.4
Belgium-Luxembourg
64.0
23.7
12.3
42.5
39.7
17.8
France
11.3
74.4
14.3
18.6
56.8
24.6
Argentina
4.1
83.0
12.9
38.8
51.7
9.5
Spain
11.0
65.4
23.6
49.9
20.9
29.2
Brazil
66.4
0.0
33.7
57.3
4.0
38.6
Poland
27.7
12.2
60.1
55.1
9.4
35.5
USSR/Russiab
14.6
17.1
68.3
37.3
11.2
51.5
India
0.0
0.0
1.0
6.9
.0.0
93.1
China
1.5
0.0
98.5
31.4
2.0
66.6
The bold numbers indicate which beverage has the highest share in total alcohol consumption.
a Germany refers to the average of FRG and GDR until 1990.
b The 1961 numbers are for the former Soviet Union, the 2005 numbers are for the Russian Federation.
31
Table 2. Determinants of (relative) beer consumption across countries: between effects model with beer
consumption per capita (1 and 2) and share of beer in total alcohol consumption (3 and 4) as dependent
variables
(1)
(2)
(3)
(4)
Beer cons/cap(ln)
Beer cons/cap(ln)
Share of beer
Share of beer
GDP/cap(ln)
2.36***
1.93***
0.089
0.13
(0.38)
(0.42)
(0.086)
(0.093)
GDP/cap(ln)^2
-0.45***
-0.26**
-0.020
-0.029
(0.093)
(0.11)
(0.021)
(0.025)
Globalization index
0.013
0.0059
0.0022
0.0042
(0.015)
(0.014)
(0.0035)
(0.0033)
Average temperature
-0.023
-0.0054
0.0043
0.0088**
(0.015)
(0.016)
(0.0035)
(0.0038)
Annual precipitation
0.093
0.088
0.052
0.047
(0.20)
(0.19)
(0.044)
(0.042)
% Catholics
0.87*
0.45
0.099
0.0090
(0.47)
(0.46)
(0.11)
(0.11)
% Protestants
0.62
-0.16
0.22
0.16
(0.59)
(0.58)
(0.14)
(0.14)
% Orthodox
0.64
0.34
-0.051
-0.092
(0.59)
(0.58)
(0.14)
(0.14)
% Jews
-0.62
-1.23
0.065
0.046
(1.04)
(1.00)
(0.25)
(0.23)
% Muslims
-1.81***
-1.52***
0.015
0.17
(0.47)
(0.52)
(0.11)
(0.12)
Romanic language
-0.34
-0.083
-0.081
-0.021
(0.28)
(0.26)
(0.066)
(0.064)
Germanic language
0.14
0.10
0.027
0.048
(0.27)
(0.26)
(0.065)
(0.064)
Price beer(ln)
0.20
(0.32)
Price wine(ln)
-0.92***
(0.24)
Price spirits(ln)
0.40
(0.27)
Price non-alc(ln)
0.23
(0.32)
Price beer/Price wine(ln)
0.20***
(0.055)
Price beer/Price spirits(ln)
-0.16**
(0.062)
Observations
2097
1674
2130
1758
Nr of countries
85
81
86
83
Average time dimension
24.7
20.7
24.8
21.2
Adjusted R2
0.77
0.80
0.14
0.28
RMSE
0.77
0.69
0.19
0.17
Standard errors in parentheses.
* p < 0.10, ** p < 0.05, *** p < 0.01
32
Table 3. Determinants of levels of beer consumption over time: fixed effects model with beer
consumption per capita as dependent variable.
Full sample
By continent
Europe
America
Asia
(1)
(2)
(3)
(4)
(5)
Beer
cons/cap(ln)
Beer
cons/cap(ln)
Beer
cons/cap(ln)
Beer
cons/cap(ln)
Beer
cons/cap(ln)
L.Beer cons/cap(ln)
0.67***
0.68***
0.77***
0.58***
(0.042)
(0.085)
(0.057)
(0.074)
GDP/cap(ln)
1.77***
0.50***
0.71
0.36
0.64***
(0.11)
(0.097)
(0.44)
(0.24)
(0.20)
GDP/cap(ln)^2
-0.26***
-0.078***
-0.11
-0.078*
-0.11***
(0.018)
(0.015)
(0.077)
(0.042)
(0.030)
Price beer(ln)
-0.16***
-0.092***
-0.11***
-0.019
-0.22***
(0.024)
(0.016)
(0.032)
(0.024)
(0.056)
Price wine(ln)
0.061**
0.030*
-0.016
0.028
0.081*
(0.025)
(0.016)
(0.014)
(0.017)
(0.043)
Price spirits(ln)
-0.0012
0.025**
0.053***
0.0061
0.017
(0.020)
(0.011)
(0.015)
(0.017)
(0.048)
Price non-alc(ln)
0.040**
0.011
0.018
0.013
0.022
(0.019)
(0.010)
(0.022)
(0.011)
(0.042)
Globalization index
0.019***
0.0063***
0.0046
0.00083
0.0084
(0.0021)
(0.0017)
(0.0038)
(0.0028)
(0.0052)
Globalization*Original beer
share
-0.049***
-0.015***
-0.010
-0.010***
-0.029**
(0.0062)
(0.0026)
(0.0064)
(0.0038)
(0.013)
Observations
1674
1672
684
506
340
Nr of countries
81
81
35
20
20
Average time dimension
20.7
20.6
19.5
25.3
17
Adjusted R2
0.47
0.72
0.78
0.69
0.71
RMSE
0.27
0.19
0.17
0.14
0.30
Driscoll-Kraay standard errors in parentheses. Year dummies included.
* p < 0.10, ** p < 0.05, *** p < 0.01
33
Table 4. Determinants of relative beer consumption over time: fixed effects estimation with the share of
beer in total alcohol consumption as dependent variable
Full sample
By continent
Europe
America
Asia
(1)
(2)
(3)
(4)
(5)
Share of beer
Share of beer
Share of beer
Share of beer
Share of beer
L.Share of beer
0.78***
0.74***
0.77***
0.73***
(0.044)
(0.051)
(0.070)
(0.090)
GDP/cap(ln)
0.11***
0.033**
0.021
0.049*
0.042*
(0.034)
(0.014)
(0.040)
(0.028)
(0.021)
GDP/cap(ln)^2
-0.029***
-0.011***
-0.0057
-0.014**
-0.0052
(0.0088)
(0.0031)
(0.0074)
(0.0058)
(0.0042)
Price beer/Price wine(ln)
-0.0050
0.0026
0.0024
0.00034
0.012
(0.0063)
(0.0028)
(0.0042)
(0.0035)
(0.0084)
Price beer/Price spirits(ln)
-0.018***
-0.010***
-0.018***
0.0043
-0.0100
(0.0054)
(0.0030)
(0.0055)
(0.0029)
(0.0082)
Globalization index
0.0060***
0.0014***
0.0021***
0.0014
0.0012
(0.00086)
(0.00031)
(0.00063)
(0.00087)
(0.00078)
Globalization*Original beer
share
-0.014***
-0.0032***
-0.0050***
-0.0031
-0.00028
(0.0020)
(0.0010)
(0.0014)
(0.0021)
(0.0023)
Observations
1601
1600
688
543
246
Nr of countries
75
75
35
20
15
Average time dimension
21.3
21.3
19.7
27.1
16.4
Adjusted R2
0.29
0.73
0.70
0.74
0.79
RMSE
0.060
0.037
0.037
0.040
0.033
Driscoll-Kraay standard errors in parentheses. Year dummies included.
* p < 0.10, ** p < 0.05, *** p < 0.01
Table 5. Determinants of total alcohol consumption over time: fixed effects model with alcohol
consumption per capita as dependent variable (full sample).
(1)
(2)
Total alc cons/cap(ln)
Total alc cons/cap(ln)
L.Total alc cons/cap(ln)
0.71***
(0.049)
GDP/cap(ln)
0.60***
0.16***
(0.066)
(0.050)
GDP/cap(ln)^2
-0.062**
-0.0082
(0.024)
(0.0093)
Price beer(ln)
-0.039*
-0.0077
(0.022)
(0.015)
Price wine(ln)
0.052***
0.038***
(0.0086)
(0.0083)
Price spirits(ln)
-0.053***
-0.037***
(0.012)
(0.012)
Price non-alc(ln)
0.036**
0.0042
(0.013)
(0.0067)
Globalization index
0.00021
0.00052
(0.0014)
(0.00084)
Observations
1689
1684
Nr of countries
83
82
Average time dimension
20.3
20.5
Adjusted R2
0.11
0.61
RMSE
0.19
0.13
Driscoll-Kraay standard errors in parentheses. Year dummies included.
* p < 0.10, ** p < 0.05, *** p < 0.01
34
ONLINE APPENDIX
List of countries:
Albania, Argentina, Armenia, Australia, Austria, Azerbaijan, Belarus, Belgium-Luxembourg, Bolivia, Bosnia
and Herzegovina, Brazil, Bulgaria, Canada, Chile, China, Colombia, Croatia, Cyprus, Czechoslovakia,
Denmark, Dominican Republic, Ecuador, El Salvador, Estonia, Finland, France, Georgia, Germany, Greece,
Guatemala, Haiti, Honduras, Hungary, India, Indonesia, Israel, Italy, Japan, Kazakhstan, Kyrgyzstan, Lao
P.D.R., Latvia, Lebanon, Lithuania, Macedonia FYR, Mexico, Moldova, Mongolia, Nepal, Netherlands, New
Zealand, Nicaragua, Norway, Pakistan, Panama, Paraguay, Peru, Philippines, Poland, Portugal, Republic of
Korea, Romania, Russian Federation, Slovenia, Spain, Sri Lanka, Sweden, Switzerland, Syrian Arab
Republic, Tajikistan, Thailand, Turkey, Turkmenistan, Ukraine, United Kingdom, United States of America,
Uruguay, Uzbekistan, Venezuela RB, Vietnam, Yemen
Table A1. Description of variables
Variables
Mean
Std. Dev.
Min
Max
Continuous variables
Beer cons/cap (in l of total pure alc.)
2.68
2.09
0.01
9.00
Share beer (in total alcohol consumption)
0.38
0.20
0.00
1.00
GDP per capita (PPP, 1,000 USD)
12.65
10.63
0.52
49.10
Price beer (USD/l)
0.76
0.40
0.12
5.80
Price wine (USD/l)
2.30
1.72
0.12
30.33
Price spirits (USD/l)
4.46
3.13
0.18
42.36
Price non-alcoholic beverages
0.75
0.51
0.01
6.66
Globalization index
58.0
16.7
18.1
91.7
Average temperature
14.0
8.4
-5.4
26.9
Annual precipitation
1.13
0.67
0.16
2.71
% Catholics
41%
39%
0%
97%
% Protestants
14%
26%
0%
97%
% Orthodox
8%
19%
0%
94%
% Jews
2%
11%
0%
85%
% Muslims
9%
21%
0%
100%
% Other
2%
6%
0%
39%
Original beer share (1970)
32%
23%
0%
100%
Dummy variables
Romanic language
39%
49%
-
-
Germanic language
21%
41%
-
-
Other language
45%
50%
-
-
Asia
21%
41%
-
-
Europe
41%
49%
-
-
America
30%
46%
-
-
Middle East
5%
21%
-
-
Oceania
4%
19%
-
-
35
Table A2. Robustness tests: Columns (1) and (2): Using lagged prices as instruments for current prices.
Columns (3) and (4): Arellano-Bond estimation to account for bias related to the inclusion of the lagged
dependent variable
(1)
(2)
(3)
(4)
Beer cons/cap(ln)
Share of beer
Beer cons/cap(ln)
Share of beer
L.Beer cons/cap(ln)
0.92***
(0.011)
L.Share of beer
0.89***
(0.027)
GDP/cap(ln)
1.71***
0.013
0.22***
0.038**
(0.080)
(0.022)
(0.054)
(0.018)
GDP/cap(ln)^2
-0.25***
-0.0029
-0.040***
-0.0092**
(0.018)
(0.0049)
(0.013)
(0.0043)
Globalization index
0.016***
0.0076***
0.00063
0.00053
(0.0021)
(0.00057)
(0.0017)
(0.00042)
Globalization*Beer share
-0.058***
-0.019***
0.00049
-0.00093
(0.0051)
(0.0013)
(0.0034)
(0.00077)
Price beer(ln)
-0.25***
-0.086***
(0.044)
(0.020)
Price wine(ln)
0.13***
-0.015
(0.031)
(0.016)
Price spirits(ln)
0.010
0.058***
(0.030)
(0.017)
Price non-alc(ln)
0.13***
0.036**
(0.032)
(0.015)
Price beer/Price wine(ln)
-0.0048
0.0087
(0.0087)
(0.0054)
Price beer/Price spirits(ln)
-0.027***
-0.017***
(0.0089)
(0.0054)
Year dummies
No
No
Yes
Yes
Observations
1610
1706
1672
1752
Nr of countries
81
80
81
82
Average time dimension
19.9
21.3
20.6
21.4
Robust standard errors in parentheses.
* p < 0.10, ** p < 0.05, *** p < 0.01
