Main findings
Between-agency discrepancy in alcohol availability estimates remained substantial in 2011, mainly because of other beverages (primarily spirits) and wine, although some decrease could be observed during 2001–2011. Actual per capita consumption was 9.5 l of pure alcohol per person-year in 2011, decreasing 2.3 % annually in 2001–2011, mainly due to decreases in wine and other beverages. The coverage of WHO availability, Tax Agency availability, self-reported purchases, and self-reported consumption compared to actual consumption was, respectively, 99.5, 99.5, 66.3, and 28.0 % in 2011, with downward time trends in the last three estimates, especially for self-reported consumption, and an upward trend in WHO availability. Multisource availability involved an overestimation of 12.3 %, mainly due to tourism imbalance.
Considerable between-agency discrepancy in alcohol availability
The study shows that in 2011 the between-agency discrepancy in per capita alcohol availability estimates remained high. Thus, the between-agency range and CV for total alcohol were 1.3 lpa/py, and 6.8 %, respectively. This probably was mainly due to excessively low estimates for wine (Tax Agency and WHO) and spirits (FAO). Important discrepancies may also exist elsewhere [
1]. In fact, in 2010 in France, Italy, Greece, and Portugal, the between-agency CVs of availability estimates considering WHO, FAO, and Eurostat varied between 3.5 and 14.8 % (total alcohol), 3.2–12.4 % (beer), 3.0–19.0 % (wine) and 8.7–45.6 % (other beverages). This situation is worrisome because alcohol availability is often the standard indicator in spatiotemporal comparisons of alcohol consumption [
1,
31].
The absolute discrepancy in total alcohol availability (as measured by the range) showed an average annual decline of 9.5 % during 2001–2011, although this decline is partly explained by the declining availability. Thus, when CV was used as the measure of relative discrepancy, the decrease was not statistically significant for total alcohol, although it was for other beverages.
Clear decrease in per capita alcohol consumption
Using uncorrected figures of alcohol availability from individual agencies to derive temporal trends in per capita alcohol consumption in Spain may be problematic. Thus, during 2001–2011 significant decreases in estimates of total alcohol availability from the Tax Agency and FAO, but not from WHO, were identified. Failure to identify a significant decrease in WHO availability (“recorded alcohol consumption”) was likely due to changes in data sources or data collection procedures. In fact, Tax Agency and WHO figures for total alcohol were very similar from 2005 onwards, suggesting that the WHO data source had changed from that year on. Mixed trends by agency in the availability of beer and other beverages were also identified.
Correcting alcohol availability by triangulating different sources (multisource availability) provides clear trends in per capita alcohol consumption, which decreased by 2.3–2.4 % annually during 2001–2011, prolonging the pattern started around 1975. Declines have been also observed in most European countries [
22,
32], and could be due to demographic, socioeconomic (including recession and unemployment), and cultural changes, as well as the possible effectiveness of some interventions [
13,
33,
34]. Unfortunately, there is little empirical evidence on the quantification of each of these factors. Population aging during the period was negligible. Instead, the increase in immigrants’ share of Spain’s residents (5.3 % in 2001 and 13.5 % in 2011) may have had some relevance, given that most immigrants in Spain drink less than natives [
35], and the country’s progressive urbanization [
4] (Spanish people living in cities >50,000 inhabitants increased from 47 % in 2001 to 52 % in 2011) may also have had an effect.
The role of economic factors is uncertain. Before the crisis starting in 2007–2008, alcohol consumption decreased, while alcohol affordability remained stable or increased [
4,
36]. However, the identification of joinpoints at 2006 for beer and wine (toward stability and decreasing, respectively), suggests that the crisis could have contributed to the decline in consumption, affecting mostly drinking on premises, which continues to represent the most important component of consumption in Spain [
37]. A recent study suggests that during the economic recession in Spain, regular excessive drinking decreased while binge drinking increased [
38]. Finally, some interventions may have been effective, such as advertising regulations, increasing the minimum drinking age, workplace drinking bans, and traffic safety policies [
34]. However, risk perception of drinking did not increase [
39]. Consumption peaks in 2001 and 2004 were probably artifacts due mainly to stockpiling of spirits, triggered by rumors of possible tax increases [
23], which hardly affected retail purchases. A progressive change of the dominant beverage from wine to beer was observed from the beginning of the study period [
4], so that in 2011 beer accounted for 46 % of total alcohol consumption vs. 28 % for wine.
Although homogenizing trends in drinking patterns across Europe have been suggested [
13], probably due to drinking patterns of youth, in other traditionally wine-drinking countries, such as France, Italy, Portugal, or Greece, wine remains the dominant beverage [
22,
40].
Hidden overestimation of per capita consumption by alcohol availability
The multisource availability figures clearly overestimated actual consumption in Spain (by 12.3 % in 2011 under intermediate assumptions). This is mainly due to the consumption/purchases imbalance between foreign visitors in Spain and Spanish visitors abroad (tourism imbalance). Thus, in 2011 in Spain there was an excess of 56.9 million international visits and 412.4 million international overnights, and most foreign visitors came from European countries with higher alcohol prices than Spain [
22,
41]. In fact, without tourism imbalance, overestimation would have been only 2.2 %. Such an imbalance could also have an important effect on actual alcohol consumption in other countries, although the direction of the effect depends on the predominance of inbound or outbound international visits [
1]. In Spain, correction of the Tax Agency availability indicator to obtain the multisource availability was a methodological need due to underestimation of wine and cider sales/supplies by the tax agency, an underestimation that also transferred to the WHO availability figures starting in 2005 (which is generally referred to as “recorded per capita alcohol consumption”). In the absence of this correction, the similarity between coverage of the Tax Agency and WHO availability (≈100 % in 2011) gives an apparent coherence to these indicators and spuriously supports their validity.
Poor and decreasing coverage of per capita consumption by self-reported behaviors
Self-reported consumption coverage in Spain (28.0 % in 2011; range: 23.8–32.0 %) seems lower than elsewhere (30–65 %) [
1,
7,
9,
14,
16‐
19], although figures outside this range have been reported [
42‐
44].
Consumption by people aged <15 does not explain the low coverage because, as reported elsewhere [
10], analysis of Spanish school surveys [
39] suggested it was negligible.
Consequently, the low coverage must be attributed to underestimation of consumption by population surveys, affecting mainly heavy and binge drinking [
45,
46]. This could be due to various factors [
1,
14,
25], including social desirability bias [
47], difficulty in quantifying and averaging consumption (especially if questions on frequency of drinking are not sufficiently disaggregated by beverage) [
17], high probability of sampling exclusion of heavier drinkers due to homelessness/housing instability or living in communal establishments (student’s group quarters, prisons, hostels, etc.) or high consumption periods like Christmas or summer holidays [
48,
49], and assumption of standard drink volumes lower than commonly used, especially for free-poured beverages, etc. [
50‐
54]. Moreover, the varying alcohol content in emerging products [
55,
56] complicates data collection.
Given the low coverage of self-reported consumption, using uncorrected figures to estimate alcohol-attributable disease burden or prevalence of heavy drinkers [
57] will greatly underestimate alcohol-related problems and intervention needs. Considering the highest range of estimated coverage, in Spain self-reported consumption figures in 2011 should be shifted upwards by as much as 91 % of any estimates of WHO or Tax Agency availability.
Self-reported purchases coverage was 66.3 % in 2011. Higher coverage (87–90 %) has been found in Sweden and UK [
9,
10]. Underestimation of purchases can be due to incomplete recording by panelists (mostly affecting small and irregular purchases), failure to consider certain beverages (i.e., alcopops), sampling bias (e.g., failure to consider off-premise alcohol purchases, made outside the household frame usually by youth) or even alcohol losses between wholesale and retail markets. The observed lower underestimation compared to self-reported consumption could be explained by less reluctance to report purchases than consumption, and less difficulty in remembering and summarizing quantities of bottles/cans purchased than multiple drinking events, especially if purchases are recorded by automatic barcode reading [
9].
Downward trends in coverage of self-reported indicators were observed in 2001–2011. The decline in self-reported consumption coverage (5.7 % annually) was especially intense, and is explained by the faster decline in self-reported consumption (8.4 % annually) compared with actual consumption (2.4 % annually). Large declines in age-adjusted prevalence of risk drinkers based on ENS self-reported consumption have also been found [
58]. However, the decline during 2001–2011 in self-reported consumption from the EDADES survey [
39] was much smaller (1.5 %).
A good spatiotemporal correlation between self-reported alcohol consumption and availability has been found elsewhere [
1,
43]. However, declines in self-reported consumption coverage have been also observed; it has been suggested these may be partly due to the growing contribution to total alcohol consumption of binge drinking and irregular heavy drinking patterns, which are poorly captured by the usual basic QF approach [
25,
49]. This is also plausible in Spain where the prevalence of these patterns is increasing [
59‐
61]. However, the decline in ENS coverage is probably due mainly to the fact that the instruments used in the most recent surveys do not capture consumption as well as those previously used. The main reasons are that the decline is very strong, and that coverage in the EDADES Survey, which has had more consistent data collection over time [
61], has not fallen (31.1 % in 2001 and 36.2 % in 2011).
Finally, a probable increase in the proportion of off-premise alcohol purchases outside the household frame (not collected) may have contributed to decreasing coverage of self-reported purchases. The joinpoint at 2004 toward a decline in beer purchases coverage is probably an artifact caused by the exclusion of alcohol-free beer starting in that year. By beverage, the highest coverage was found for wine and the lowest for other beverages, both in self-reported purchases and consumption, with heterogeneous findings elsewhere [
9,
10].
Strengths and limitations
Data from multiple sources were triangulated to obtain actual consumption for Spanish residents. This revealed the weakness of official national estimates, which is often transferred to international statistics. Calculation algorithms incorporate data on various factors affecting the estimates, especially regarding consumption/purchases during international visits (Spanish residents abroad and foreigners in Spain). However, empirical national data on various aspects were lacking. Thus, assumptions were made based essentially on international studies [
4,
10,
13,
18,
21,
22,
62]. The algorithms and methodology used have been clearly explained, and the sensitivity analyses allowed calculation of uncertainty ranges for estimates. The main uncertainties arise in estimating unrecorded alcohol consumption and purchases during international visits. Estimates of unrecorded alcohol have little empirical basis; in Spain, its main component is probably derived from informal winemaking, although this should have declined with progressive urbanization. Smuggled or surrogate alcohol is likely to be negligible due to low alcohol prices. As for purchases during international visits, the algorithms combine empirical data (i.e., number of visits, average stay, the ratio between Spain and foreign countries of alcohol prices, and of per capita alcohol consumption) and assumptions (price elasticity of alcohol demand [
36], or average quantity of alcohol purchased in a visit involving countries with alcohol prices and per capita consumption similar to Spain) (Additional files
3,
4, and
5). Moreover, the number of entries and overnight stays of foreigners in Spain might be somewhat higher than estimated because some illegal immigrants cannot be included in the population register. However, the phenomenon probably has low relevance, because in 2001–2011 very few conditions were required to be included in this register [
35]. Finally, identical alcoholic beverage strength (ABV percentage) was applied throughout the entire period, but as elsewhere, changes over time cannot be ruled out, especially affecting wine strength, which might have increased [
63‐
65].