Rozmiary i dystrybucja spożycia alkoholu w Polsce. Oceny oparte na badaniu na próbie udziałowej i losowej
Authors: Jerzy Jasiński
Number of views: 21
The Problem Statement
The main goals of the alcohol consumption surveys, repeatedly carried out Poland, are to describe the changes in the drinking subcultures of different segments of the population, to monitor the emergence of new drinking practices and the discontinuance of old ones, and to record the evolution of attitudes toward alcohol. In order to facilitate the analysis of these and related phenomena, similar instruments were used and applied in a consistent manner in several of the consecutive surveys. Any substantial change in the formulation of the questions asked or in the manner they are presented to the respondents (e.g. a mailed questionnaire versus a personal interview), raises a number of problems regarding the interpretation of the results obtained. Should the changes observed be treated as reflecting the transformations of the processes studied, or should they be attributed to alterations in the way the respondents were approached? On the other hand, repeated use of the same questionnaire applied in exactly the same manner does not seem to be an attractive proposition, since new problems and interests emerge and new survey research methods are being developed. In the Polish alcohol consumption surveys an attempt has been made to find a balance between securing continuity and introducing change.
However, sometimes changes have to be made irrespectively of the researcher’s wishes. They may stem from resources constraints forcing the researcher to adopt a cheaper or less time consuming method of conducting a survey, or from some other reasons. In the case of the Polish alcohol consumption surveys such an external change occurred recently in relation to the sample design.
Nine surveys have been conducted in Poland since 1961. Five of these surveys have utilized quota samples, while in four, random samples were employed. However, the studies based on random samples, for the most part, are not representative of the majority of Polish research on drinking practices. Instead, three of the four surveys based on random samples are unique in that they were devoted to an examination of the following issues: meaning attached by the general public to the notion of abstinence; estimation of the content and influence exercised by the antialcohol propaganda; and the effectiveness of the appeals by the Roman Catholic church to abstain from drinking alcoholic beverages during the month of August in the last 3 years.
The studies based on quota samples are much more representative of Polish surveys on drinking practices. In these studies, the primary focus has been on the description and nature of drinking occasions. For example, this was the focus in the pioneering surveys by A. Swiecicki carried out in 1961 and 1962, in J. K. Falewicz’s survey of 1968, and in the Polish Alcohol Consumption Surveys of 1980 and 1985. In all of these surveys the same way of measuring alcohol consumption was adopted and their results were analyzed according to the same methodology. The data collection was done by the Center for Public Opinion Survey and Program Studies of the State Committee for Radio and TV in Warsaw for all the surveys. However, in recent years, this Center has been increasingly using random samples of the population in its research, first restricting and more recently abandoning completely the use of quota samples. From now on, the new Polish alcohol consumption surveys will be conducted on random samples only. While the shift to the use of random samples of the population represents a methodological step forward, it raises some important questions. Perhaps most important is whether results from these new surveys will be comparable with results from former surveys? The study reported in this paper is aimed at providing an answer to this question.
In the case of a random sample, at least initially, every object in the population has to have an equal probability of being drawn, and at least the first step in the sampling procedure has to be taken randomly. In a quota sample, on the other hand, the usual procedure is that the interviewer has to interview a specified number of subjects with predetermined characteristics (e.g. men; 20‒29 years old; having primary educaiion only; etc.) How the interviewer finds these persons is left to his ingenuity, and even if he does not select them by applying any krown criteria, he is not drawing his part of the sample randomly. If all the interviewers follow exactly the instructions in selecting interviewees, the distribution of the characteristics of the quota sample should be the same as in the general population. It is hoped, but only hoped, that it will secure the representativeness of the sample. To what extent is this hope fulfilled cannot be determined.
In September 1985 the data for the second Polish Alcohol Consumption Survey were collected on a quota sample (QS). Four weeks later, in October 1985, some items from the questionnaire used in that survey were added to a routinely conducted public opinion poll carried out on a random sample (RS). The sizes of the samples were 1808 and 882, respectively. In both cases the field work was done by the Center for Public Opinion Survey.
The samples were comparable to one another in terms of the respondents’ occupation, sex, level of education, and proportion of people living in urban and rural areas. However, both samples differed from the general population with respect to age. In the QS, persons aged 60 and older were underrepresented (by about 12%), while those aged 40‒59 were overrepresented (again by about 12%). Within the RS, those aged l6‒19 were underrepresented (by about 6%). The former discrepancy must have originated from the interviewers’ not foliowing closely enough the instructions given to them as to the age of persons they had to interview. The latter discrepancy stemmed from the sampling procedure; the RS was derived from the 1984 election lists, which covered the population of 18 years old and over. One year later, in 1985, these lists included only persons aged 19 and older.
The differences in the age structure in the QS, RS, and in the general population are of some significance for estimations of tne coverage rate and for comparability of the samples, and thus will be further discussed later on in the paper.
Frequenct and Quantity
In the Polish Alcohol Consumption Surveys, as in previous Polish research, the level of consumption has been estimated by means of the last occasion approach. In order to make such estimates a number of assumptions have to be accepted. They are related to the typicality of the occasion described by the respondent in terms of kind of occasion, amount of alcohol consumed, etc. Additionally, some assumptions regarding the frequency of drinking also have to be accepted. These requirements may seem unrealistically far going. However, two points have to be considered. Firstly, the more typical an occasion of drinking is ‒ the more likely, it is to be reported to the interviewer. Secondly, in epidemiological research the interest is on categories or types of respondents and not on a single individual; in groups of respondents, unconventional occasions tend, at least in part, to annul their impact on the means and other measures describing the drinking modalities.
The variables used in the Polish Alcohol Consumption Surveys to establish the drinking patterns were: kind of beverage, quantity consumed, and frequency of drinking. There were no differences between the QS and the RS in respect to the frequency of drinking wine purchased in stores as well as home made fruit wine. However, there were some differences in the case of spirit and beer. In the RS, two categories of spirit drinkers could be recognized more distinctly than in the QS; infrequent (i.e. drinking spirits less often than once a month), and very frequent (i.e. several times a week). The former category consisted of 45,5 per cent (QS) and 48,1 per cent (RS), respectively, and the latter of 21,3 per cent (QS) and 30,0 per cent (RS). In the case of beer consumers, there were significantly more infrequent drinkers in the RS than in the QS (48,6 and 34,9 per cent, respectively), and fewer very frequent drinkers (38,5 and 32,5 per cent, respictively).
Significantly higher quantities consumed on one occasion were reported in the QS than in the RS, irrespectively of beverage. This would suggest that the interviewers collecting data for the QS tended, possibly without even realizing it, to select prospective respondents from among persons consuming more alcohol than average.
One of the consequences of the above differences in regard to the frequency of drinking and quantities of alcohol consumed is the overestimation of the concentration of consumption in research based on QSs. The concentration of consumption is traditionally used to refer to the size of proportion of consumers who drink about half of all alcohol consumed. According to an estimate derived from the RS in 1985 in Poland, half of spirits were consumed by about 10 per cent of drinkers (rather than by 7 per cent as indicated by the QS), and of wine purchased ‒ by 8 per cent of drinkers (rather than by 7 per cent); in the case of drinkers of home made wine (5 per cent), and of beer (10 per cent) the estimates of concentration of consumption proved to be the same in the QS and in the RS.
The shrinking of the coverage rate was another consequence of the overestimation of the quantities of alcohol consumed of the frequency of drinking of some alcoholic beverages. The comparison of the amount of alcoholic beverages sold during 1985 in the country with the amount accounted for by the surveys, reveals the following picture:
spirits 50,6 per cent 47,8 per cent
wine purchased 49,3 per cent 25,8 per cent
beer 81,5 per cent 45,2 per cent
Total 55,9 per cent 48,1 per cent
The substantial differences in the coverage rate are noticed only in relations to wine and beer, e.g. to the beverages of lesser importance considering the Polish drinking patterns. The absence of substantial differences between the RS and the QS in relation to spirits stemmed from the fact that the overestimation of the amount drunk on one occasion was partly compensated for by the underestimation of the size of the category of the very frequent drinkers. As far as the better and wine drinkers were concerned no such compensation was present.
The data on the sale of alcoholic beverages indicate that in that in 1985 about 68 per cent of all alcohol purchased in Poland was spirits. Of the total amount of alcohol consumed, the QS indicated that spirits accounted for 62 per cent, while the RS indicated that spirits accounted for 74 per cent. The first estimate was too low, the other too high, each by 6 per cent.
As could be expected from the low coverage rate of wine, in the research based on the RS, the place of wine in the structure of beverages consumed was underestimated (9 per cent instead of 15 per cent of all alcohol consumed). The place occupied by beer was estimated correctly by the RS (18 per cent). The modest contribution of beer and wine to the total amount of alcohol consumed should be attributed mainly to the drinking habits prevailing in Poland, but also partly to the unmet demand for beer and for wine, in particular for imported grape wine.
The analysis of the data from the QS showed that drinking one kind of alcoholic beverage raised the likelihood of drinking another kind of alcoholic beverage. This was true with regard to every beverage type and all these relationships were statistically significant. The data from the RS corroborated fully the above results. In the case of persons who consumed two kinds of alcoholic beverages, a further question could be asked, viz. do persons who drink greater quantity of one beverage also tend to drink more of the other one. This was the case in respect to all of beverages both in the QS in the RS.
In order to estimate the respondents’ level of alcohol consumption all beverages consumed had to be recounted according to their content of alcohol, and added up. The distribution of the respondents according to the level of their consumption revealed some important differences between the estimates derived from the QS and RS. First of all, according to the RS, there were significantly more teetolers in the general population (23 per cent) than according to the QS (16, 1 per cent). This is partly due to the underrepresentation persons aged 60 and older in the QS. Relatively more persons in this age group are non-drinkers than in the lower age brackets. The other reason for this discrepancy in the estimate of the size of the fraction of non-drinkers is the already mentioned tendency of the interviewers to select as prospective respondents drinkers rather than abstainers while collecting data for the QS.
The other difference between the RS and QS relates to the proportion of persons drinking substantial amounts of alcohol. On the whole, the RS indicates that it is smaller than the QS does. The extrapolation of the results obtained in the RS on the adult population in the country leads to the following conclusion. In 1985 there were about 2,1 million persons who annually consumed more than 16 liters of pure alcohol in Poland. They were consuming at least one half a litter bottle of vodka in every ten days. Among them nearly 1,5 million consumed at least 24 liters of pure alcohol annually, i.e. at least one bottle of vodka per six days. This group of drinkers represented only 5 per cent of the adult population, but consumed nearly 40 per cent of all alcohol in the country.
It was stated before that only about half of the total amount of alcohol consumed was accounted for by the survey based on the RS. The most important reason for the underreporting seems to be, apart from forgetting and concealing, the underrepresentation of heavy drinkers in the sample. Because of this, it is likely that the group of drinkers who consume at least 24 liters of pure alcohol annually is considerably more numerous, and that they probably drink even a greater proportion of all alcohol consumed in the country.
Drinkers, Abstainers, Teetolaters
Who is a teetotaler seems self-evident and not requiring special consideration only as long, however, as the responds from the general population are not asked what their understanding of the term is. In one of the recent Polish general pop lation surveys this question was addressed and some surprising results were obtained. Although 69 per cent of the respondents considered a teetotaler somebody who has not been drinking any beverages containing alcohol, there were also persons having other opinions in this matter. As many as l5 per cent described a teetotaler as a person who drinks small quantities of alcohol, infrequently, a few times a year only, and further 2 per cent described a teetotaler as a person who does not drink vodka.
In order to avoid this issue, in the questionnaire applied both the QS and to the RS, a question was asked whether the respondent had been drinking any alcoholic beverages during the last 12 months preceding the interview. Those who responded in the negative are considered to be teetotalers. Those who have not been drinking a particular beverage are called abstainers, i.e. abstaining refers only to not drinking a particular beverage. This means, that according to this definition abstainers include persons who drink as well as those who do not. Consequently, all teetotalers are abstainers, but among the latter there also drinkers (of some beverages). According to the terminology adopted here only drinkers who consume all kinds of alcoholic beverages listed in the questionnaire can not be called abstainers.
The beverage consumed by most Poles is vodka. According to the RS only 28 per cent of the respondents abstained from drinking vodka (12 per cent of men and 42 per cent of women). In this respect the differences between the QS and the RS were not significant. In the case of all other beverages the RS produced significantly higher estimates of the fraction of abstainers than the QS. According to the RS, 65 per cent of respondents do not drink beer (46 per cent of men and 82 per cent of women), 68 per cent do not drink wine purchased (46 per cent of men and 71 per cent of women), and 78 per cent do not drink home made wine (75 per cent of men and 81 per cent of women).
The QS and the RS are in concordance on a number of points. Firstly, on the degree of popularity of the various alcoholic beverages. Judging from the size of the fraction of abstainers, spirits is the most widely consumed beverage, followed by beer, wine purchased, and home made wine. Secondly, there are consistently more drinkers among men than among women; only in the case of home made wine is the difference in the size of the fraction of drinkers among men and women not significant. Thirdly, more than half of women drink spirits, i. e. many more than any other alcoholic beverage. It contradicts the belief, widely held in Poland, that vodka is a man’s and wine a woman’s beverage. Fourthly, abstaining from any beverage is equally spread in rural as in urban areas. It illustrates the process of unification of behaviour patterns in Poland, or as it is sometimes formulated – of the urbanization of rural culture, and ruralization of the urban life styles.
The size of the fraction of abstainers is, irrespective of the beverage, closely
related to the age of the respondents. The older the respondent – starting from the 20–29 age group – the larger the fraction of abstainers.
Drinking is also related to the level of education attained. In the case of spirits, wine purchased, and home made wine, the higher the respondents educational attainment the greater the fraction of drinkers. It corresponds to the observation made in a number of countries, according to which the higher the social status – the smaller the fraction of abstainers, the grater the frequency of drinking, and the smaller the amount consumed on one occasion. The above relationship do not apply to beer, which in Poland seems to be a beverage consumed primarily by blue–collar workers.
The QS and the RS show the same pattern of results between drinking and living in rural versus urban communities as found for age and level of education of the respondent.
Finally, the teetotalers differ markedly from the drinkers. Here again the results of the QS and RS are in harmony. There are significantly more teetotalers among women than among men, and in the age group 40 years old and over (in particular – among those 60 years old and over). There are also more teetotalers among less educated than more educated, and among unqualified blue–collar workers and peasants than among qualified blue–collar workers and white–collar employees.
There has been a prospective change in the sampling method used in Polish alcohol consumption surveys. This called into question the feasibility of long term trend analysis of the drinking habits prevailing in Poland. In order to dispel the arising doubts, the present inquiry has been carried out comparing the results of the new sampling method (primarily relying on the use of random samples) with the prior method which utilized quota samples. In this project, the same questions relating to drinking practices were put to two groups of respondents. One consisted of a quota sample (QS; n =1808), and the other of a random sample (RS; n = 882) of the adult population.
The comparison of results from the two samples leads to the following two observations. Firstly, the numerical values of various parameters characterizing drinking practices, such as the overall consumption level and that of particular beverages, or that of quantities imbibed on one occasion, etc., differ in the QS and the RS. In a number of instances these differences are statistically significant. The results based on the RS suggest a somewhat lower level of alcohol consumption than the ones based on the QS, and shift the coverage rate slightly below 50 per cent.
Secondly, the QS and the RS generated very similar pictures of drinking patterns prevailing in Poland, in terms of the kinds of beverages consumed, interrelationship between the various beverages consumed, concentration of consumption, and social traits of people drinking more and drinking less alcohol.
The general conclusion emerging from the research project described here can be formulated as follows. The QS provided a fairly true picture of the drinking practices in Poland, however, in order to establish the long term trends in these practices, it is necessary to take into account the corrections stemming from the estimates derived from the RS.