I. Introduction
Wine and alcoholic beverages are important goods in the European Union (EU) food market, representing billions of dollars in sales, exports, and imports. In 2011, the total EU expenditure on wine was $97 billion, expenditure on beer was $157 billion, and expenditure on spirits was $107 billion (Euromonitor, 2012). The EU is a major producer of wine, but also a major importer of wine. EU wine imports from the United States were $489 million in 2012, and total EU wine imports (external trade) were $3.2 billion. The EU has also been importing increasing amounts of wine, in quantity terms. The effects of recent trends on wine consumption may have ramifications for countries that export wine to the EU.
The EU has experienced a number of structural and financial changes that may affect its demand for wine and alcoholic beverages. We observe that one potential reason for changes in the demand for wine worldwide might be research on the potential health properties, including both costs and benefits, which have been publicized. This research may affect consumer tastes and the position of wine and other alcoholic beverages in the average consumer's consumption bundle. Second, the EU now has a number of new members, many of which are middle-income countries from Eastern Europe. These countries may have more rapidly rising incomes, leading to greater consumption of goods generally regarded as luxuries. The increased trade that has characterized the EU has also in some cases led to convergence in consumption patterns across countries. Additionally, several EU countries have experienced negative growth as a result of the recent recession, which would result in reduced consumption of all goods, especially those generally regarded as luxuries. Finally, wine may also be considered almost a staple for some EU countries, given their history of wine production and consumption.
This research calculates income and price elasticities of demand for wine, as well as some other alcoholic beverages, for all EU countries and then for various subgroups of EU countries. The research will (1) characterize the demand for wine and other alcoholic beverages and (2) determine whether the EU-27 can be characterized as a single large market or differentiated based on whether the countries are more recent members with somewhat lower incomes, differences in consumption patterns, characterization as “mature” markets for wine, and differences in the relative importance of each determinant of demand. The work focuses on wine, but demand for alcoholic spirits and beer will also be evaluated.
Section II of this work provides background on the broad trends in EU wine and alcoholic beverage markets. Section III discusses the methodology used to decompose these changes in demand into income effects, price effects, and trend effects. Section IV discusses the results, which characterize the demand functions for different subgroups of countries within the EU. Discussion and conclusions are presented in Sections V and IV.
II. Background
A. Trends in EU Wine Consumption: Volume and Expenditure
Overall wine consumption volume in the EU has declined slightly, whereas expenditure on wine and other alcoholic beverages has experienced increases and decreases over the past decade. Overall wineFootnote 1 consumption (measured in volume) in EU countries fell by 4.1% between 2001 and 2011 (Euromonitor, 2012). The rate of change of quantity consumed has been small.
The expenditures on wine, however, have been somewhat different. From 1998 to 2012, the total value of wine sales changed little in inflation-adjusted terms. However, this small change masks several marked changes in trends over that period, with a slight decline from the late 1990s to 2000, a marked rise from 2001 to 2008, and a decline from 2008 to 2012. Alcoholic beverage expenditures, beer expenditures, and spirits expenditures overall exhibit the same trends; beer expenditures exhibit more of an overall decline.
B. Consumption Patterns across Different Countries within the EU
Consumption patterns, however, differ greatly across countries within the EU. This overall decline in quantity masks a wide variety of trends within individual countries. Table 1 illustrates the differences across countries in the allocation of budgets to different alcoholic beverages. Table 2 gives some indication of differences in prices and quantities of wine and other alcoholic beverages consumed in individual EU countries, and Table 3 illustrates differences in the rate of change of expenditure on different beverage categories. Wine consumption has differed greatly for the Eastern European countries that have recently acceded to the EU, and the Western European countries that made up the EU-15, plus Malta and Cyprus. The former group saw a 34% increase in overall wine consumption (in quantity terms), although it constitutes less than 10% of the EU's total wine market in volume.
The Western European group, in contrast, experienced a 6% volume decline in wine consumption overall. However, declines in Italy, Spain, and France, which are both major wine producers and consumers, were mostly responsible for the overall declines, whereas markets like the Netherlands, the United Kingdom, Ireland, Sweden, and Finland experienced large increases (see Table 3). More modest income growth in the Western EU countries may lead to more modest increases in the demand for wine. However, the variation across this group of wealthy countries suggests that the variation may also correlate with the maturity of the market for wine, with less or negative growth in more mature markets. These trends are mimicked somewhat in the consumption of alcoholic spirits and beer, with modest declines overall and a large variation across countries.
When consumption is measured in expenditure, the data follow a similar pattern. The Eastern European countries have experienced large increases in wine consumption over the period in question, with growth rates in nominal expenditure in the triple digits in many cases. (Table 1). Large increases in beer and spirits consumption are also evident.
Table 1 Share of Alcoholic Beverages in Total Alcoholic Beverage Expenditure
Sources: Euromonitor (2012, 2013), with Economic Research Service (ERS) calculations.
High wine consumption countries in Western Europe, in contrast, have seen low, single-digit increases in nominal wine expenditure, with much larger increases in beer expenditure. Other Western European countries, however, have experienced double-digit increases in wine expenditure, with less growth or even declines in beer. Growth in spirits has varied.
C. Reasons for Across-Country Differences in Demand
There are several potential reasons for the changes in alcoholic beverage consumption patterns and for their differences across the EU. These countries may be experiencing increases in demand for wine due to increases in income. The average Eastern European country had a 323% increase in nominal per capita income between 1997 and 2011, compared with a 100% increase for the Western European countries. Wine is often conceptualized as a luxury good, so that countries with large increases in income may enjoy greater consumption. Additionally, the responsiveness of wine budget shares to income and expenditures may be lower for the higher-income countries of Western Europe than for more middle-income Eastern European countries. Seale, Regmi, and Bernstein (Reference Seale, Regmi and Bernstein2003) found that expenditure elasticities for the broad food category of beverages plus tobacco were an average of 0.836 for middle-income countries and 0.44 for high-income countries.
Additionally, most of the EU countries have had an initial increase in alcoholic beverage consumption followed by a decline from 2008 during the recent downturn. Although overall alcoholic beverage expenditure followed this pattern in most countries, some experienced very little downturn in wine or alcoholic beverage consumption due to the recession, and others experienced large downturns. Some countries also saw an increase in per capita wine expenditure, despite an overall decline in nominal per capita expenditure on alcoholic beverages.
Prices have historically varied across the EU. Countries vary in their comparative advantage in the production of wine, which has probably affected historic consumption patterns, so that Mediterranean grape-growing countries have had access to inexpensive wine for centuries, whereas in Finland, wine would have been relatively expensive until the shipping innovations of the last century. With free trade among EU countries, the improved access to imports should result in smaller price differences across countries. However, there are also large differences in taxation regimes for alcoholic beverages across the countries of the EU (Anderson, Reference Anderson2014). These differences may also affect prices.
Convergence in consumption patterns, in which countries’ dietary patterns and budget shares begin to resemble each other's over time, has been noted in other research on the EU. Convergence in consumption may come from the convergence of prices as trade barriers fall, but it may also stem from demonstration effects. Elsner and Hartmann (Reference Elsner and Hartmann1998) found evidence of convergence between Eastern European and Western European countries for some food stuffs. Gil, Gracia, and Pérez y Pérez (Reference Gil, Gracia and Pérez y Peréz1995) found convergence in diet across the Western European members of the EU. Alternatively, De Mooij (Reference De Mooij2003) has found divergence across Western European countries in a variety of consumer goods. The decreases in wine share in a few high wine consumption countries and the large increases in Eastern Europe and some countries that historically have consumed less wine and more beer suggest that convergence may be a possibility.
Taste for wine may also be changing. Beer consumption, for example, has been falling in the United States due to the shift toward wine and spirits (Yglesias, Reference Yglesias2013). Beer consumption fell from 84.4 liters per capita in 2000 to 76.6 liters per capita in 2014 (Euromonitor, 2015). In addition, recent research on the health effects of wine has received a great deal of press coverage, which may loom large in the mind of the consumer (Oaklander, Reference Oaklander2015; Wade, Reference Wade2006). However, evaluating the total costs and benefits of wine consumption in particular and the costs and benefits of alcohol consumption in general is a complex task, and the benefits may not outweigh the costs (Bakalar, Reference Bakalar2014; Fekjaer, Reference Fekjaer2013).
III. Characterizing the Demand for Wine in the EU
A. Approach and Models
The background information suggests that different countries in the EU may have differing levels of per capita wine consumption. These differences may be due to differences in income, differences in historical costs of wine production (particularly when compared with the costs of beer or spirits consumption), differences in taxation, and across-country differences in taste parameters mentioned previously. These differences may mean that we should not attempt to characterize the EU as a monolithic market for wine. Rather, the EU may best be characterized as several markets with varying characteristics. This work uses several techniques to determine how EU countries might best be grouped into submarkets and then tests for significant differences in demand across the submarkets.
This work will first group countries with similar characteristics into alcoholic beverage submarkets by (1) looking at consumption patterns and (2) determining the differing effects of prices, incomes, and trends for domestic wine demand using a standard model of demand function estimation. These models will calculate the price and income elasticities of demand for wine. Second, these parameter values will be compared across subgroups of countries to determine whether there are significant differences across countries.
We use the first-differences linear almost ideal demand system (abbreviated L-AIDS) model for both phases of the work. The AIDS model, developed by Deaton and Muellbauer (Reference Deaton and Muellbauer1980), models consumption by representing the share of a good in total expenditure as a function of the prices of the good and its substitutes and total expenditure for the good and its substitutes, with the latter adjusted by a weighted price index. The first-differences version of the model regresses the changes in the log-linear values of the per capita expenditure share of the good in question on the log-linear values of the price of the good, the prices of substitutes, and total expenditure on the good, plus its substitutes, compared to an index of prices.
The standard first-differences L-AIDS model equation is
$$d\; w_i = b_i (d\; \ln X - {\rm d\; ln}\;P) + {\rm \Sigma} _j^n \; c_{ij} \; ln\; d\; p_j, $$
where w is the share of a good, X is the total expenditure on the good and its substitutes, and P is a price index for the goods, following Neves (Reference Neves1994).
We do this for the cross section of EU countries for the period 1998–2012. For the first-differences models, we use the intercept as a measure of the time trends. We use seemingly unrelated regression (SUR) (after Alston, Foster, and Green, Reference Alston, Foster and Green1994; Henneberry, Reference Henneberry2010; Mehrara and Ahmadi, Reference Mehrara and Ahmadi2011; Neves, Reference Neves1994; SAS Institute, 2013) to calculate the parameters, using the restrictions that the sum of the price coefficients must add up to zero and the symmetry condition that the cross-price coefficients for wine and beer must be equal. The spirits equation was the omitted equation in the system. The parameters are then used (after Alston, Foster, and Green, Reference Alston, Foster and Green1994; Neves, Reference Neves1994) to calculate the uncompensated price elasticities and the expenditure elasticities (see Appendix).
We use a single-stage model of consumption that focuses on alcohol expenditure. Expenditures on beer and alcoholic spirits are considered as potential substitutes for wine, and the sum of expenditures on those three categories (adjusted by a price index) is the expenditure variable. Although a reasonable level of correlation between alcoholic beverage expenditure and income can probably be assumed, if the share of alcoholic beverage expenditure in the total budget changes, our expenditure elasticity will not be equal to a true income elasticity.
In first-differences models, the magnitude of time trends is reduced (Madden, Reference Madden1993). They only reflect whether expenditure shares change at an increasing or decreasing rate. However, if the model includes an intercept term, the original time trend is reflected in that (Madden, Reference Madden1993), and we do use an intercept here to document the trends. Some of the causes of changes in wine consumption, such as changes in tastes over time and or/convergence, may be best observed in the time trends, as separate from price/cost effects or income effects.
B. Data
Quantity per capita and price data for wines, beer, and spirits for the period 1998–2012 were taken from the Euromonitor database, 2012–2013. Prices are given in constant U.S. dollars, and quantity is given in liters per capita. Euromonitor records sales and market sizes for a number of processed goods in many individual countries from a variety of official and private sources. Their retail selling price is used for market valuation here. Unit prices were used; thus prices represent an average across all wine consumed.
C. Phase 1: Monolithic Market?
We group the EU economies into different configurations to determine whether the EU actually consists of several different markets for wine. Determining the appropriate groupings is the first stage in the process. Fogarty (Reference Fogarty2008) also classifies countries on the basis of shares of alcoholic beverages consumed. Nelson (Reference Nelson2013), in a meta-analysis of determinants of wine and alcohol demand elasticities, includes binary variables for subgroupings of countries, with European country groupings of Nordic, United Kingdom–Ireland, and a group consisting of Cyprus, Czech Republic, France, Italy, Poland, Romania, and Spain.
In order to determine which countries were like countries, we examined consumption levels for each beverage, initial beverage shares of expenditure, changes in beverage shares, and growth in per capita expenditures for each of the alcoholic beverage categories (Tables 1–3).
Table 2 Mean Prices and Per Capita Consumption, Alcoholic Beverages, 1998–2012
Sources: Euromonitor (2012, 2013).
Table 3 Nominal Expenditure Increase, 1998–2012 (in percent)
Sources: Euromonitor (2012, 2013), with Economic Research Service (ERS) calculations.
We also applied the first-differences L-AIDS model to individual countries (Table 4). Each country only has 14 observations, so these individual country results are not as robust as results involving a larger panel sample with several countries. They do, however, provide some indication of how countries’ consumption patterns differ. Fogarty (Reference Fogarty2008) and Nelson (Reference Nelson2013) use a meta-analysis approach to evaluate the parameters of alcohol demand across studies. Both articles consider studies for a number of countries, including some EU countries, and both determine average values of income and price elasticities for various alcoholic beverages, while also pointing out that some EU countries or groups of EU countries have different elasticity values than the averages across all countries.
Table 4 Elasticities for Individual Countriesa
Notes: a Because the parameters are based on a short time series, many are not necessarily significant. b The asterisk (*) indicates that the expenditure parameter was significant at the 10% level. c Calculated parameter. d The asterisk (*) indicates that the own-price parameter was significant at the 10% level. e Significance evaluated at the 10% level; ns, not significant
D. Phase 2: Verify Groupings and Characterize the Demand
Here, we characterize the demand for wine and other alcoholic beverages for the whole EU sample and for each group. We apply the model to the full sample of 27 EU countries. We look at a simple SUR regression of the model on the sample as a whole, adding interactive slope and intercept terms for each of the four major subgroupings (Tables 4 and 5). Thus, equation (1) becomes
$$\eqalign{d\; w_i & = b_i (d\; \ln X - {\rm d\; ln}\;P) + {\rm \Sigma} _j^n \; c_{ij} \; ln\; d\; p_j \; + {\rm \Sigma} _{k\;} ^m s_k + {\rm \Sigma} _{k\;} ^m s_k b_i (d\; \ln X - {\rm d\; ln}\;P) \cr & \quad + {\rm \Sigma} _{k\;} ^m s_k {\rm \; \Sigma} _j^n \; c_{ij} \; ln\; d\; p_{\,j,} \;}$$
where k indexes the m subgroups of countries, and s is a binary variable indicating membership in country group k. The significance of the interactive terms will give some indication of the validity of the country subgroupings. Fogarty (Reference Fogarty2008) uses a country binary/interactive term approach as well.
Table 5 Descriptive Statistics for Stage 2, European Union–Wide Regression
Note: Variables are not in differenced log-linear form, with the exception of the expenditure variable.
Using panel data introduces two kinds of variations into the data. There are across-country effects, and there are effects that take place in similar countries over time. The elasticities may measure variation across countries due to historical costs of wine production and consumption or historical preferences, rather than what is happening over time. The full sample regression captures the sum of these two types of effects. Using first-differences models reduces the across-country effects, as the initial magnitudes matter less in the difference variables. Also, the cross-country effects of initial starting values, which for alcohol consumption vary quite a bit across countries, are reduced, as only difference over time are compared across countries. Grouping like countries, which have similar characteristics, also reduces some of the effects of differences in historical starting points. Classifying countries by consumption pattern does illustrate the diversity within the EU, however, and gives some indication of the kind of market for beverages that will be found in each country.
IV. Results
A. Phase 1
The following four groupings collect together sets of countries that have several similarities in their demand functions for wine. These groupings are not the only way that the EU market for wine might be divided, but they do highlight the variation across the EU in wine market attributes.
The mature wine markets of France, Italy, Spain, and Portugal form one group. In these countries, which are historically producers and significant consumers of wine, we note the following findings:
-
• Wine share has fallen (see Table 1).
-
• Individual per capita expenditure elasticities for wine are between 0.5 and 1.04 (see Table 4). We expect income elasticities for wine to be smaller in these countries.
-
• Individual price elasticities for wine are generally between 0 and −0.5 for these countries, suggesting a more inelastic demand. We would expect price elasticities to be smaller, as wine behaves more like a staple good in these markets.
-
• Individual expenditure elasticities for beer are greater than or equal to 1.
Austria, Cyprus, Luxembourg, Malta, Hungary, and Slovenia (Central European countries) form a second group. The shares and levels of wine expenditure vary quite a bit for this group (Tables 1 and 2). In these countries, results are as follows:
-
• Wine consumption and consumption share have risen (Table 1), with the exception of Slovenia, and Malta. In Slovenia, wine expenditures have risen by more than alcoholic beverage expenditure (Table 3), although beer expenditure has risen faster.
-
• The expenditure coefficients for wine are positive and significant for each country, and the expenditure elasticities for wine are high, with Luxembourg's 1.36 the smallest and the rest well over 1.5 (Table 4). The demand curves for countries in this group exhibit characteristics of an expenditure-driven change in demand for wine.
-
• The expenditure coefficients for beer are negative and significant.
-
• For all but Austria and Cyprus, the intercept or time trend variable are significant as well, so there may be an additional time-linked change in tastes.
-
• Price coefficients in the wine equation are significant for Austria, Cyprus, and Luxembourg.
Belgium, Denmark, Finland, Ireland, the Netherlands, Sweden, and the United Kingdom (Northern European countries) form another group.
-
• These countries have also increased the share of wine in their alcoholic beverage expenditure and decreased the share of beer (Table 1). The initial wine shares of the countries in this group are all less than 0.35 and are less than 0.3 for all but Sweden.
-
• Expenditure coefficients are not significant for this group (negative and significant for Denmark), although Ireland and Belgium's expenditure elasticities for wine are actually greater than Luxembourg's, as well as those of the other countries in the group, despite the coefficients’ lack of significance (Table 4).
-
• In these countries, time trends are positive and significant for wine, and generally negative and significant for beer (the beer trend is not significant for Sweden). The demand characteristics of this group, therefore, seem to be mostly linked to time/preference trends.
-
• Price parameters are significant for wine for Denmark, the Netherlands, and the United Kingdom.
Bulgaria, Czech Republic, Estonia, Latvia, Poland, and Slovakia (Eastern European 6) form another group of wine-consuming countries.
-
• All of these countries have had significant increases in wine expenditure that surpass the growth rate of alcoholic beverage expenditure (Table 3). Wine and beer shares are increasing at the expense of the budget share for alcoholic spirits, although beer share has fallen slightly for Slovakia (Table 1).
-
• Their income parameters in the first-differences regressions are either not significant or negative and significant, and individual country expenditure elasticities are low (less than 0.7).
-
• In addition, this group also has a significant time trend of increasing wine consumption over time.
-
• All countries except for the Czech Republic and Estonia have statistically significant and positive wine price coefficients.
Germany, Greece, Romania, and Lithuania did not seem to fit well with any of the other countries because of the absence of significant time trends and income parameters or differing consumption patterns. For Germany and Greece, only the price parameters were significant in the wine equation. For Romania, even the price parameters were not significant. Lithuania's wine equation lacked a significant time trend, and the main growth in expenditure has been in spirits. They are therefore left out of the subgroups analysis.
B. Phase
Tables 5 and 6 provide some summary statistics for the sample of all 27Footnote 2 EU countries, and Table 7 provides the coefficients resulting from the SUR.
Table 6 Correlation Matrix, Independent Variables, Second Stage of Results
Table 7 Seemingly Unrelated Regression Results with Interactive Terms
Note: Asterisks (***, **, and *) indicate significance at the 1%, 5%, and 10% levels, respectively.
(1) Own-Price Elasticities
Own-price elasticities for wine and beer for the full sample are less than 1 (Tables 4 and 5 provide coefficients). The elasticity of wine is somewhat larger in magnitude than beer. These results imply that for the sample as a whole, wine and beer are not luxury goods. Spirits, however, are income sensitive. Own-price elasticities for wine for the four subgroups are generally negative as expected. The own-price parameter interactive terms for wine and beer for the four subgroups were generally not significant.
(2) Expenditure Elasticities
Expenditure elasticities for the full sample indicate that spirits have the highest expenditure elasticity, by a very large margin. Wine is next with a value greater than 1, and the value for beer is less than 1, probably reflecting the status of beer as more of a staple.
The interactive terms for expenditure for wine for the four subgroups differ markedly. The expenditure term for the Central European countries, with increasing wine consumption, is positive and significant, as we would expect from the results for the individual countries in that group. The Northern European countries have a negative and significant expenditure term, suggesting that the income elasticity would be lower than that of the sample as a whole. The term for the traditional wine consumers and producers is negative, but not significant. Surprisingly, the six Eastern European countries have the lowest expenditure elasticities for wine.
The interactive term for expenditure on beer (Table 7) was positive and significant for the traditionally high wine producers and consumers. Interestingly, the Northern European countries, with very high initial levels of beer consumption, also have a positive and significant expenditure interactive term. The interactive terms for the expenditure for Central and Eastern European countries in the beer equation were not significant.
Expenditure elasticities for spirits are quite high; in most cases, the values are greater than 2.0. This probably reflects the status of spirits as a good that consumers treat as a luxury.
(3) Cross-Price Elasticities
Here, we focus our analysis on the elasticities for changes in wine demand in response to changes in the prices of beer and spirits and the cross-price elasticity for beer demand in response to a change in wine prices. We would expect that both beer and spirits would be substitutes for wine. However, for the sample as a whole, they were weakly complementary.
The interactive cross-price terms were not significant in the beer equation. For the traditionally high wine consumers and the Northern European group, the same was true for the wine equation. However, for the wine equation, the cross-price term for beer was negative and significant for the Central European countries and positive and significant for the Eastern European countries, whereas the opposite was true for the cross-price term for spirits. This suggests that wine might be more complementary to beer in the Central European countries and more substitutable in the Eastern European ones, whereas the reverse is true for spirits.
(4) Time Trends
Time trends were measured by using the intercept in the difference model. For the model as a whole, the wine equation exhibits a significant positive time trend for wine, whereas the beer time trend is not significant.
The four subgroups exhibit marked differences over time, as measured by the significance of the binary variable for each group. Only the binary variable for the traditionally high wine producers was not significant, and it was also negative. The binary variables for the Eastern, Central, and Northern European groups were all positive and significant in the wine equation, indicating a positive time trend for wine consumption even greater than that of the sample as a whole.
For beer, once again, the traditionally high wine producer binary is not significant. All of the other group binaries are, but they vary in sign. For Eastern European countries, the binary is positive and significant; thus, the taste for beer is increasing over time. For the Northern and Central European countries, however, the coefficient is negative and significant, indicating a declining taste for beer.
V. Discussion
The own-price elasticities of wine and beer demand for the EU as a whole suggest that these goods are not luxury goods, as their own-price elasticities are less than 1. Although that may be surprising when evaluated against a standard conception of alcoholic beverages as nonnecessities, this finding is well supported in the literature. Nelson (Reference Nelson2013), in a metastudy of research on the elasticities of alcoholic beverages, found price elasticity averages of −0.30 for beer, −0.45 for wine, and −0.55 for spirits. Gallet (Reference Gallet2007), in another meta-analysis, reported that average price elasticities for alcoholic beverages were −0.83 for beer and just over 1 for wine and spirits. He also found that elasticities for alcoholic beverages were sensitive to demand and econometric specifications, as did Fogarty (Reference Fogarty2008). Fogarty finds that the market for alcoholic beverages as a whole behaves as a necessity, with mean own-price elasticities of −0.65 for wine, −0.44 for beer, and −0.73 for spirits. Interestingly, the literature seems to find that spirits also behave as a necessity, in contrast to our results here.
The calculated expenditure elasticities for the group as a whole are less than 1 for both wine and beer. This corroborates the research of Seale, Regmi, and Bernstein (Reference Seale, Regmi and Bernstein2003), which reports income elasticities of less than 1 for alcoholic beverages and tobacco. This is also not terribly far off from the work of Nelson (Reference Nelson2013), in which average income elasticities are 0.5 for beer, 1.0 for wine, and 1.0 for spirits. We should also note that as this model is a one-stage formulation, we measure the elasticity of alcoholic beverage expenditure, as opposed to household income as a whole. This may measure more than one potential change. Alcoholic beverage expenditure presumably rises with income. Thus, the change of wine shares with respect to alcoholic beverage expenditure is probably correlated with the change of wine shares with income. However, if the share of alcoholic beverages in the overall budget rises as income rises, we may understate the effect of a change in income on the expenditures on each of the beverages.
Our results indicate that wine and beer and wine and spirits are not substitutes in the EU but, rather, are weakly complementary. There may be a greater or lesser tendency to consume alcoholic beverages as a group, so the demand may be more likely to move together. There may also be a joint marketing aspect to these goods within the EU as a whole.
We do see evidence that there are differences in demand functions across the various subgroupings of countries. Their overall consumption pattern over time indicates that the high wine consumption countries are slowly moving into beer (or spirits) and away from wine. The group of traditional wine producers may be the least different from the sample as a whole (Fogarty [Reference Fogarty2008] found that wine price elasticity for France was less negative than for other countries). This may be due to the fact that their changes in wine consumption are very large and may be driving the results for the sample as a whole. The only statistically significant difference from the whole sample is the expenditure term for beer, and this difference would give this group a higher expenditure elasticity for beer. Most of the traditionally high wine consumption countries have had markedly rising shares for beer, and beer may be more responsive to expenditure in those countries, as there, it is a more novel addition to the consumption basket. Interestingly, however, there is no time trend suggesting a changing taste for beer. Rather, an income change seems more likely to be accompanied by a consumption increase for beer. The rising beer consumption and declining wine consumption in these countries could be evidence of convergence.
In the Central European countries, we see rising wine expenditures and, for some countries, rising wine shares. Wine expenditure in these Central European countries has generally increased faster than alcoholic beverage expenditure. This is true even in countries where wine share was already quite high compared with the rest of the sample. Our results suggest that a Central European demand system would have a larger expenditure elasticity for wine, a greater tendency to purchase wine in response to a change in beer price, and more substitutability between wine and spirits. The demand for wine described by the model behaves more like the demand curve for a discretionary purchase, with greater income and price sensitivity, albeit mostly cross-price sensitivity.
The positive significance of the binary for this group in the wine regression suggests that the effect of the time trend is greater in Central Europe. This suggests that there is an increasing demand for wine in these countries independent of price and income trends. The time trend captures any changes in tastes that are consistent across time. These changes can include demonstration effects that can come from exposure to new wines via trade, media reports, or advertising or increasing desire for variety in the diet. Taste for wine, as measured by time trends, is increasing, whereas taste for beer is decreasing. Fogarty (Reference Fogarty2008) also finds evidence of time trends in wine demand. For these Central European countries, changes in tastes, income, and prices of substitutes are driving the demand for wine.
Northern European countries have increasing wine shares as well, but starting from a lower average level than the Central European countries. The binary interactive terms for this group indicate a smaller income elasticity for wine and a larger one for beer, despite the prevalence of beer in the consumption basket, usually a characteristic of a staple. This also contrasts with the results of Fogarty (Reference Fogarty2008), who found a smaller income coefficient for beer for Belgium and Finland and larger coefficients for wine for Finland and the United Kingdom, and agrees partially with the results of Nelson (Reference Nelson2013), who found smaller income coefficients for Nordic countries but larger ones for the United Kingdom and Ireland compared with the world as a whole. Nelson also found higher price elasticities for Nordic countries and smaller ones for the United Kingdom–Ireland. Fogarty and Nelson did not include all the EU countries, however, nor were the countries grouped as they have been here. The magnitudes in this research appear to be largely explained by the time trends that are significant and positive for wine and largely negative and significant for beer. These results suggest that the time trends may be more definitively determinant of demand than incomes or prices for this group. They may also be evidence of some convergence in alcoholic beverage consumption for this group, as these countries move out of beer and into wine.
Eastern European countries would also have lower income elasticities for wine. The wine expenditure elasticities are quite small. This seems quite surprising, as one might imagine that as incomes rise, middle-income countries such as those of Eastern Europe might increase their wine consumption, along with their consumption of other goods commonly imagined as luxuries. Indeed, the results for Seale, Regmi, and Bernstein (Reference Seale, Regmi and Bernstein2003) indicate that we would expect higher income elasticities for alcoholic beverages for lower-income countries than for higher-income countries. Over this period, however, incomes have been falling in many cases for Eastern European countries, while wine expenditures continued to increase. Time trends for both wine and beer are both positive and significant, suggesting that these countries may simply be shifting tastes to both wine and beer from alcoholic beverages. These trends were not disrupted by income fluctuations, suggesting that demand for wine in these countries is more sensitive to changes in taste trends than to changes in income. This could also be evidence of Eastern European countries’ converging their consumption patterns with the Western European countries as they integrate economically.
The estimated Eastern European demand system would also suggest a greater tendency to reduce wine purchases in response to a decline in beer prices, but a greater tendency to buy more wine when spirits prices decline. These countries also have increasing wine shares, starting from a very low level (less than 12% on average), whereas alcoholic spirits shares are declining from very high initial levels, which does not suggest that the complementary relationship between wine and spirits is of long standing.
One source of the across-country price differences is the large differences in taxation regimes across countries. Taxing alcohol, either with export taxes, excise taxes, or Value Added Tax (VAT), to raise revenue is a common practice in most nations. There is a great degree of variability in levels of taxation across countries, even within the EU (Anderson, Reference Anderson2014). If we look at levels of standard consumption taxation, VAT is within 15% and 27% for the countries in the sample in 2012 and between 13% and 27% in 2015 (Anderson, Reference Anderson2014; EU, 2015). Excise taxes vary quite a bit more. Levels of excise taxation are not available for every country in the sample in 2012, but for those included in Anderson (Reference Anderson2014), we note that the levels of excise taxation on wine are highest for the Northern European countries and lowest for Central European countries and traditional wine producers. Northern European countries have prices that are on average higher than the other groups, but not necessarily universally higher; that is, some individual countries in other groups have very high wine prices (Table 2).
At first glance, these differences in taxation may seem to explain some of the across-country variation in wine demand, especially the fact that levels of wine consumption are historically higher in the traditional and Central European countries than Northern European ones. However, if we look at relative taxation levels, we generally find that for most countries, the taxation per liter is highest for spirits and lowest for wine, with beer somewhere in between, which makes the variation in relative shares for each type of alcoholic beverage more difficult to explain. If we consider taxation per standard alcoholic drink, we note that Eastern European countries are more likely to have higher spirits taxes, and wine and beer taxes that are close in value, which mirrors the shift out of spirits into wine and beer. Northern European countries are also more likely to have wine and beer taxation levels that are close in value, which may help explain the shift into wine. However, Central European and traditional wine producers both have beer taxes that exceed those of wine, and one group is shifting into beer, whereas the other is shifting out. Differences in taxation cannot explain all of the variation across countries.
One potential problem with considering taxes as an explanatory factor in alcoholic beverage demand is that the taxation levels may be endogenous. Governments sometimes rely on alcoholic beverage taxes as revenue earners or as curbs on excessive consumption. In this case, the government may tax the sector of highest consumption, which obfuscates the effects of the taxes on the demand function.
VI. Conclusions
Overall, we note that the demand for wine in the EU is more price and income elastic than the demand for beer. Alcoholic spirits demand resembles that of a luxury good. A significant positive time trend in the demand for wine indicates an increasing taste for wine consumption.
We also note that the EU cannot be characterized as a monolithic market for wine. Different subgroups of countries have different demand characteristics. In a few high wine-producing countries, we see wine characterized as a staple good, with high, declining consumption shares and low expenditure and price elasticities. In those countries, wine is losing share to beer and spirits. In Northern European countries, wine share is growing relative to beer, but income elasticities are not as high, and time trends appear to be more significantly correlated with demand. The same is true for Eastern European countries, although there, wine is gaining share from spirits. In other Central European countries, wine still has high expenditure and price elasticities of demand, despite having fairly high average consumption shares within the alcoholic beverage category. In those countries, rising incomes would lead to increased consumption.
The fact that income effects seem small for some groups is a result of the fact that many countries had increases in wine share, despite falling or stagnating incomes due to a global recession from 2008 to 2012. This suggests that for some of these groups, rising wine share may be the result of changes in tastes or convergence in some countries, trends that take place over time. For Northern and Eastern European countries, we see a shift toward wine and away from beer and spirits, respectively, with time having a greater effect. These effects are not consistent across groups, however. For the traditional wine producers, we see a shift away from wine, but here, the indication of convergence is found largely in the greater income sensitivity of beer. Central European countries evidence some kind of time trend toward wine, even though wine shares were quite high compared with the rest of the sample for some countries in this group.
We also note that alcoholic beverages may be complements for one another rather than substitutes. This may be due to the fact that many countries may regard alcoholic beverages as a whole as a good, so that countries that consume wine may also consume beer and spirits.
These results have only considered a one-stage model of alcoholic beverage consumption—that is, alcoholic beverage demand in relation to other alcoholic beverages. A more complete picture of demand would be achieved by considering a two-stage model, in which alcoholic beverages compete for shares of household income against all other goods.
Appendix
We use the elasticities for first-differences linear almost ideal demand system models outlined by Neves (Reference Neves1994).
Expenditure elasticity:
$$1 + \displaystyle{{b_i} \over {w_i}} $$
Own-price elasticity:
$$\displaystyle{{c_{ii} - \left( {b_i w_i} \right)} \over {w_i}} - 1$$
Cross-price elasticity:
$$\displaystyle{{c_{ij} - b_i w_i} \over {w_i}} $$
