regional differences in drinking among finnish adolescents

9
Addiction (1995) 90, 57-64 RESEARCH REPORT Regional differences in drinking among Finnish adolescents SAKARI KARVONEN Department of Public Health, University of Helsinki, Finland Abstract Regional variation in health behaviour can be explained either by differences in the socio-demographic characteristics of population or can be based on some other differences, such as cultural traditions. The purpose of this study was to find out the extent to which provincial differences in adolescents' alcohol use can be reduced to socio~demographic differences between adolescent populations of the provinces. A cross-sectional postal survey data from the Adolescent Health and Lifestyle Survey was used, hi 1991, the sample was weighted according to the size of the province. It represented 16- and 18-year-old Finns in each province (n = 6600). Provincial variation in weekly alcohol use was reduced for about one quarter and the ordering of the provinces in drinking changed when the differences in urbanization level between provinces were adjusted. When socio-demographic factors were adjusted two provinces still differed significantly from the other provinces. Our study shows that variation in alcohol use among Finnish adolescents is related to where they live, and not simply to demographics. Introduction Studies examining health and health-related hchaviours have usually used geographic space only as a way of organizing data, leaving geographic differences usually unanalysed or reduced to social and economic causes. For instance, the Black Report (Townsend & Davidson, 1982) having shown clear regional differences in mortality ratios in Britain—the healthiest region lying in the southern part of the country—states (p. 199): In our view much of the evidence on social inequalities in health can be adequately under- stood in terms of specific features of the socio- Correspondence to: Sakari Karvonen, Department of Public Health, BO Box 52 (Manncrheimintie 96), FIN-00014 Univer- sity of Helsinki, Finlnnd. 0965-2140/95/010057-08 economic environment: features...which are strongly class-related in Britain and also have a clear causal significance. Yet, as it is evident that all social action takes place at a definite location it remains unclear whether geographic space, or more specifically region of residence, holds a genuine analytic power. The lack of analysis of significance of location led Duncan, Jones & Moon (1993, p. 725) in their multi-level analysis of regional variation in health-related behaviour to state that [the numerous "lifestyle" studies conducted recently] "have been little used in explorations of the difference that place makes in the structuration of lifestyle behaviour". This paper aims to analyse the role of regional © 1995, Society for the Study of Addiction to Alcohol and Other Drugs

Upload: sakari-karvonen

Post on 06-Jul-2016

217 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Regional differences in drinking among Finnish adolescents

Addiction (1995) 90, 57-64

RESEARCH REPORT

Regional differences in drinking amongFinnish adolescents

SAKARI KARVONEN

Department of Public Health, University of Helsinki, Finland

AbstractRegional variation in health behaviour can be explained either by differences in the socio-demographiccharacteristics of population or can be based on some other differences, such as cultural traditions. The purposeof this study was to find out the extent to which provincial differences in adolescents' alcohol use can bereduced to socio~demographic differences between adolescent populations of the provinces. A cross-sectionalpostal survey data from the Adolescent Health and Lifestyle Survey was used, hi 1991, the sample wasweighted according to the size of the province. It represented 16- and 18-year-old Finns in each province(n = 6600). Provincial variation in weekly alcohol use was reduced for about one quarter and the orderingof the provinces in drinking changed when the differences in urbanization level between provinces wereadjusted. When socio-demographic factors were adjusted two provinces still differed significantly from the otherprovinces. Our study shows that variation in alcohol use among Finnish adolescents is related to where theylive, and not simply to demographics.

IntroductionStudies examining health and health-relatedhchaviours have usually used geographic spaceonly as a way of organizing data, leavinggeographic differences usually unanalysed orreduced to social and economic causes. Forinstance, the Black Report (Townsend &Davidson, 1982) having shown clear regionaldifferences in mortality ratios in Britain—thehealthiest region lying in the southern part of thecountry—states (p. 199):

In our view much of the evidence on socialinequalities in health can be adequately under-stood in terms of specific features of the socio-

Correspondence to: Sakari Karvonen, Department of PublicHealth, BO Box 52 (Manncrheimintie 96), FIN-00014 Univer-sity of Helsinki, Finlnnd.

0965-2140/95/010057-08

economic environment: features...which arestrongly class-related in Britain and also have aclear causal significance.

Yet, as it is evident that all social actiontakes place at a definite location it remainsunclear whether geographic space, or morespecifically region of residence, holds a genuineanalytic power. The lack of analysis ofsignificance of location led Duncan, Jones& Moon (1993, p. 725) in their multi-levelanalysis of regional variation in health-relatedbehaviour to state that [the numerous "lifestyle"studies conducted recently] "have been littleused in explorations of the difference thatplace makes in the structuration of lifestylebehaviour".

This paper aims to analyse the role of regional

© 1995, Society for the Study of Addiction to Alcoholand Other Drugs

Page 2: Regional differences in drinking among Finnish adolescents

58 Sakari Karvonen

variation in one health-related behaviour, namelyalcohol use among Finnish adolescents. Alcoholuse has been reported to be related to region ofresidence in many countries (e.g. Blaxter, 1990;Williams & Debakey, 1992). In Finland as well,the first studies on adolescents' alcohol usefound that the Northern and Eastern provincesof Finland have been drier than the Western andSouthern ones. A small but consistent differencebetween north and south has also been foundin other more recent studies (see Rimpela et al.y1990).

Karvonen, Ahlstrom & Rabkonen (1991)showed, however, that alcohol use amongadolescents is related to province so that thewetness of the Southern parts of the countryhas, during the 1980s, been largely due tothe province of Uusimaa, where the capital issituated. Also, despite the relative dryness of thenorth as a whole, the province of Lappi appearedwetter than the rest of the country. The driestprovinces were those of Vaasa (in the northwestern coast) and Oulu (in the northern part ofthe country).

Virtually all the central features of the map ofadolescent alcohol use have remained constantduring the 1980s (Karvonen et al., 1991). Theconstancy of the alcohol map and the similarityof the map both among adolescents and adults(see Simpura & Lahti 1988) suggest that provin-cial differences in drinking frequency rest onfactors that are related to geographic location.

Duncan et al. (1993) concluded that in Britainthe major determinants of geographical variationin alcohol use are individual level socio-demo-graphic factors. In Finland provinces differgreatly according to the socio-demographic char-acteristics of the population. Socio-demographicdifferences could therefore explain differences inalcohol use if the proportion of populationgroups with frequent alcohol use is higher inprovinces where alcohol is used more frequentlythan in provinces where alcohol use is less fre-quent.

The strongest socio-demographic predictorsfor alcohol use, which are also related to prov-ince, are the level of urbanization, social status,the level of education and family structure (see,e.g. Ahlstrom, 1983; Rimpela et al, 1990). Al-cohol use increases along with the level of urban-ization. Drinking is related to social status so thatadolescents coming from farmer families drinkless than others. Alcohol use is also most preva-

lent among adolescents having quit school afterthe compulsory education, and in one-parentfamilies.

These findings suggest two alternative hypoth-eses. First, it can be hypothesized that differ-ences in alcohol use between provinces arelargely explained by differences in the socio-de-mographic characteristics of adolescents. Sec-ondly, differences in alcohol use can be based onother differences, such as cultural traditions. Thepurpose of our study is to find out to what extentprovincial differences in drinking can be reducedto variation in socio-demographic factors be-tween adolescent populations.

Material and methodsDataIn the Adolescent Health and Lifestyle Surveythe data were collected by mailed questionnairesin 1991. The sample was weighted according tothe size of the province and represented 16- and18-year-old Finns in each province. The sample,which was obtained from the Central PopulationRegistry, was based on consecutive days of birthso that all Finns bom on the sample days wereincluded (18 consecutive dates in June, July andAugust; in smaller provinces, also certain datesin September). The average ages of respondentswere 16.7 years and 18.7 years in each of theprovinces (see Rimpela, et al., 1990.)

The number of respondents and responserates by province are presented in Table 1. Re-sponse rates were lowest in the provinces ofUusimaa and Vaasa (Table 1).

The frequency of alcohol use was asked with astructured question: "How often do you usealcohol? Also try to remember the situations inwhich you have only used small amounts, suchas half a bottle of medium-beer or a sip of wine."Tliere were nine alternative responses rangingfrom "daily" to "I do not use alcohol". Alcoholuse was dichotomized and adolescents whodrank at least once a week (24.8% of all respon-dents) were separated from the others becauseweekly use of alcohol was considered a levelrepresenting a regular drinking habit. Non-re-spondents to question of alcohol frequency(2.6% of the whole data) were not included inthe analyses.

In the first step of analysis urbanization levelwas operationalized into three groups: on thebasis of the administrative division municipalities

Page 3: Regional differences in drinking among Finnish adolescents

Region, drinking and adolescents in Finland 59

Table 1. The number of respondents, response rates and the percentage of weekly itsers of alcoholby province among 16- and 18-year-old Finnish adolescents in 1991

Province

UusimaaTurku ja PoriHameKymiMikkeliPohjois-KarjalaKuopioKeski-SuoitiiVaasaOuluI-appiTotal

Number

918620625503516545697683446518529

6600

Response rate (%)

69.876.278.678.576.876.980.180.769.875.272.575.7

Weekly use of alcohol' (%)

28.425.024.024.622.219.120.223.615.517.924.522.3'

'Non-respondems for alcohol use question were not included in the total number.Percentage of non-respondents varied between 1 and 9 by province.

were divided into urban and rural, and after thatbig cities (with over 100 000 inhabitants: thecities of Helsinki, Vantaa, Espoo, Tampere,Turku, Oulu, Lahti); other cities (medium-sizedand small cities) were separated according topopulation. In the final step of analysis the othercities were divided into medium-sized cities(30 000 or more inhabitants) and small cities(less than 30 000 inhabitants). These groups ac-counted for 15, 18, 19 and 48% of the respon-dents.

Respondents were classified into three socialstatus groups on the basis of the standardclassification of occupations of the head of thehousehold (Central Statistics Office of Finland,1975): white collar employees (including—present or former—managers, administrativeor clerical employees) workers (present orformer skilled, specialized, or unskilled workersas well as farm and forestry workers) andfanners (present or former farmer employersand own-account farmers). These groupscomprised of 48, 42 and 10% of respondents,respectively. The level of education was dividedinto three groups: high school (51% ofrespondents), other school (32%) and not atschool (17%). Family structure was analysedusing three types of families: (1) nuclear families(both parents are alive and live together;75% of respondents), (2) single-parent familiesand neo-families (families with stepfatheror stepmother; 21%) and (3) other types,mainly including adolescents who are alreadymarried (4%).

Methods of analysisProvince and each socio-demographic variablewere first cross-tabulated with alcohol use inorder to determine whether a relationship be-tween alcohol use and socio-demographic vari-ables exist in this data. As all of the variablesproved to be related to alcohol use they wereincluded into the analysis of multivariate tables.In order to control differences between provincesin the data, gender and age were also taken intothe multivariate analysis since they both affectalcohol use.

The multivariate tables were then analysed bylogistic regression analysis using the statisticalpackage GLIM (Aitkin et al., 1990). Here theelaborative method presented by Valkonen &Martelin (1988) was applied in the analysis oflogistic regression analysis. In this method theoriginal relationship—here between alcohol useand province—is elaborated by adding indepen-dent variables one at a time into a model andchanges of the odds ratios (OR) of the depen-dent variable (alcohol use) are analysed. Oddsratios (ORj) for weekly alcohol use were calcu-lated in the following way:

where p, signifies the probability of using alcoholweekly in category i of the independent variableand p,, signifies the probability of using alcoholweekly in the reference category of the indepen-dent variable (Aitkin et al., 1990). In the case ofprovince, Uusimaa was selected as the referencecategory and, respectively, the first category of

Page 4: Regional differences in drinking among Finnish adolescents

60 Sakari Karvonen

— o o —r- rri O

o t ^ i nh - fN i ^oo — voinOo-ocoaiooocai^>JiO

* *O (M —a •£• t~

* *O -O 00o r- in

O o r̂* *

O O vOO f- ' *

* *

lOO — o o — — o

O0^XO0Nt~-XCTi^»OO

— o o — •

I csi O •, OO q •

lOO — o o d — OO'o o— CO

— o o — o o ——o

* * * MOT)"C0 O(^t~ O*DO>qvor-̂ oa;>D oin —— do '-^dd — — —

*o aO ONo *O tN

*O CM mO ON VD

lOOO'-'OO— —

—OOOOOO— — — —

o o — o o

O CM ^

— d d

: 0 2— y r-

|0Z|

Page 5: Regional differences in drinking among Finnish adolescents

Region, drinking and adolescents in Finland 61

Table 3. Statistical fii of models on the relationship between alcohol use and province;16- w 18-year-old Finnish adolescents in 1991

Model

1. GM2. I +prv3 . 2 + gen4 . 3 + age5. 4 + urb6. 5 + sta7. 6 + fam

8. 7 t- edu9. 8 4 gen.age

10. 9 + age.edu11. 10 +gen.edu

Scaleddeviance

2163211620511749173017161682

1585158015521541

df

1378136813671366136413621360

1358135713551353

DifferencesModels

1-22-33-44-55-66-7

7-88-99-10

10-11

4665

302191434

975

2811

df p

101 '1 *2 '? J

2 '

2 '1 '2 '2 *

Only statistically significant variables and interactions are described. Abbrevia-tions as shown in Table 2.

each socio-demographic variable (see Table 2).Uusimaa was selected as the reference province,as the unadjusted alcohol use frequency washighest there (Table 1).

The elaborative method requires that thecausal links between the variables are defined.In a cross-sectional study the causal order ofvanables cannot be totally determined. In ourstudy, having introduced province, the con-founding effects of age and gender were con-trolled for first. Thereafter, variables describingadolescents' social background (urbanizationlevel, social status and family structure) wereadded, and finally educational level, which wasconsidered to describe more directly the socialstatus of the adolescents themselves. Havingcontrolled the main effects of socio-demographicvariables the interactions of the variables werefitted one at a time with the model holding al!main effects.

Tlie significance of adding a variable and theinteractions in the model of alcohol were testedby the Pearson chi-square test (p<0.05) (Aitkinet ai, 1990). The differences between provinceswere tested by calculating 95% confidence inter-vals (CI) for parameter estimates of each prov-ince and by comparing them pairwise with thereference province of Uusimaa. Similarly,CI were calculated for each category of thesocio-demographic variables and comparedwith the reference categories.

The effect of adding a variable to the originalrelationship was measured, first, by calcu-lating variances of the parameter estimates

of the additive models and by comparingchanges in variances. Secondly, having addeda socio-demographic variable, changes inindividual estimates were also studied.

ResultsWeekly use of alcohol was related to the provinceof residence of adolescents (p< 0.001) (Table1). The unadjusted percentage was highest in theprovince of Uusimaa. The other southernprovinces (Turku ja Pori, Kymi and Hame),Lappi and the province of Keski-Suomi (in themiddle of Finland) held high rates of weeklyalcohol users. In the provinces of Vaasa andOulu the rate of alcohol users was lowest. Theprovinces of Pohjois-Karjata and Kuopio alsohad low rates of alcohol use.

Adjusting for gender and age produced onlyminor changes in the regional differences in al-cohol use (Table 3). Both gender and age wererelated to alcohol use so that the odds for alcoholuse increased with increasing age and werehigher among boys than girls.

Alcohol use was also related to urbanizationlevel, so that drinking was most frequent in bigcities (Table 2). However, alcohol use did notdiffer between other cities and rural areas.Adding urbanization level both decreased prov-incial variance in alcohol use by 23% andchanged the ordering of provinces in alcohol use(Table 2). After adjusting for urbanization levelalcohol use was most frequent in the provinces ofKymi and Lappi while Uusimaa had the highest

Page 6: Regional differences in drinking among Finnish adolescents

62 Sakari Karvonen

OR before inclusion of this variable into themodel.

The decrease of provincial variance and thechange in ordering resulted from the fact that^apart from Uusimaa^big cities are situated inthree provinces only (the provinces of Turku jaPori, Hame and Oulu). As alcohol use was morefrequent in big cities than in less urbanized areasthe adjustment increased the odds ratios for theless urbanized provinces. This is why the two lessurbanized provinces, where drinking was mostfrequent, were placed before Uusimaa in theorder.

Adjusting for social status and family structurenarrowed provincial differences in alcohol useonly slightly (Table 2). Social status wassignificantly related to alcohol use so that drink-ing was less frequent among farmer's childrenthan other social status groups, and to familystructure so that the most frequent drinkerscame from single-parent and neo-families.

Adjusting for educational level did not pro-duce changes in regional differences in alcoholuse even though alcohol use was related to edu-cational level (Table 2). Alcohol use frequencyincreased with decreasing level of education.

These findings show that provincial differ-ences in social status, family structure or averagelevel of education are not so great that theywould explain differences in drinking betweenprovinces. In other words, differences in alcoholuse do not result from the percentage of farmersbeing smaller or single-parent and neo-familieshigher, nor from the average level of educationbeing lower in provinces where alcohol use ismore frequent.

After adjusting for the socio-demographic vari-ables alcohol use differed significantly from thereference category of Uusimaa (OR set to 1.00)in only two provinces: Vaasa and Oulu (Table2). In the first phase of analysis when no vari-ables were adjusted for the estimates of fourother provinces (Mikkeli, Pohjois-Karjala, Kuo-pio and Keski-Suomi) also differed significantly.

None of the interactions between province andsocio-demographic variables proved significant,which showed that the relationship between thesocio-demographic variables and alcohol use issimilar in each province. Three other interac-tions (those between gender and age, age andeducational level, gender and educational level)became significant but none of them had anyeffect on the relationship between province and

alcohol use. The construction of the model ispresented in Table 3.

As the big cities are situated in only fourprovinces, thus creating many empty cells inmultivariate tables, studying interaction betweenprovince and urbanization level appeared prob-lematic in the logistic regression model. There-fore, two additional models were created: in thefirst case urban areas were categorized further bydifferentiating middle-sized (30 000 or more in-habitants) and small towns. In the secondmodel, big and medium-sized cities were com-bined in order to reduce the number of emptycells in the muttivariate table. The significance ofthe different models was tested by comparing thechange produced when two different categoriza-tions of urbanization level were added into amodel containing only the effects of province,age and gender. The original model reached,however, the best fit (SD of the difference for thefirst new model was 19 with df=3 and thesecond 5 for 2 with df- 2, respectively); nor didthe interaction between province and urbaniza-tion level become significant in the new models.This stage of analysis confirmed the relative dry-ness of the provinces of Vaasa and Oulu: drink-ing was infrequent in these provinces regardlessof the categorization of level of urbanization.

The overall significance of province was testedby constructing a separate model holding all theother significant variables and interactionsspecified earlier. Into this model was addedprovince and the significance of adding it tested.Province proved to be significantly related toalcohol use even after all other variables werecontrolled for (SD of the difference between themodel with and without province was 24 withdf- 10). I

DiscussionRegional variation in alcohol use among Finnishadolescents was reduced by about one-quarterby the differences in the urbanization level be-tween provinces. Adjusting for urbanization levelchanged the ordering of alcohol use betweenprovinces as well, so that the provinces of Kymiand Lappi held highest drinking rates. Thechange in the ordering took place because al-cohol use rate in Uusimaa was lowered whenurbanization level was taken into account. Ap-parently, the relatively high alcohol use in Uusi-maa is due to the fact that most of the

Page 7: Regional differences in drinking among Finnish adolescents

Region, drinking and adolescents in Finland 63

adolescents of the province live in the big citiesof the capital region.

Despite earlier findings of increased alcoholuse, especially in the capital region, it was notanalysed separately as province was selected tobe the main level of analysis. The fact that theinteraction between province and urbanizationlevel did not become significant shows, however,that drinking among adolescents is related simi-larly to urbanization level in all provinces.

The relationship between urbanization leveland alcohol use has been interpreted to resultfrom better availability of alcohol in densely pop-ulated areas, where the amount of restaurantsand alcohol shops is bigger (e.g. Ahlstrom,1983). However, alcohol use was more frequentthan elsewhere only in big cities, even thoughone would expect the availability of alcohol todecrease with the decreasing level of urbaniza-tion.

Rahkonen & Ahlstrom (1989) suggest that lowalcohol use frequency might be based on strictersocial control in rural than urban areas (see alsoSimpura & Lahti, 1988). Again, our findings donot support this explanation since it would re-quire drinking rates of rural areas to be loweredin comparison with urban areas. As the relation-ship between urbanization level and alcohol usewas not the primary concern of this study, fur-ther research on the underlying causes for in-creased alcohol use in big cities is required.

When socio-demographic differences betweenprovinces were controlled most of the variationin alcohol use was produced by only twoprovinces of the middle of Finland, namelyVaasa and Oulu. Lower response rate from theprovinces of Vaasa and Uusimaa than the otherprovinces may slightly distort the comparisonsbetween provinces since earlier follow-up studieshave shown that alcohol users are over-repre-sented among the non-respondents (NTTT,1979). As Uusimaa was set as the province withwhich other provinces were compared, the se-lectedness of respondents should be consideredwhen interpreting the results. The effect of low-ered response rate can, however, be expected tobe fairly insignificant. If the response rate ofUusimaa had risen to the average of the wholecountry (75.7%) and if the percentage of weeklyalcohol users among non-respondents is pre-sumed to be double (56.8%) that observedamong respondents, alcohol use rate would havebeen 30.6—only 2.2% higher than the rate in

this study. The similarity of our results withearlier studies also supports the inference thatthe effect of the lower response rate is mild.

In the case of Vaasa the conclusion of select-edness of respondents by alcohol use does not,however, hold, as a large proportion of non-re-spondents is likely to consist of the Swedish-speaking minority of the province. Due tofinancial reasons the study was conducted inFinnish only, which has led to low response ratesamong the Swedish-speaking adolescents. It isunlikely, however, that the provincial differencebetween the province of Vaasa and the otherprovinces would result from the non-representa-tiveness of this group of adolescents, as a similarfinding has been found in other studies includingthe Swedish-speaking minority and official statis-tics (Simpura & Lahti, 1988; Alcohol StatisticalYearbook, 1992). Increasingly, the response rateof the neighbouring province of Oulu, wheredrinking frequency was low, did not differ fromthat of the other provinces.

There are two possible explanations for thesedifferences in drinking frequency. First, it re-mains possible that regional differences reflectthose structural differences that we were not ableto measure and therefore to explain away. Thesignificance of these differences appears to bemild as the structural variables included in ouranalysis are where the relationship according toearlier studies is the strongest with both provinceand drinking.

Secondly, differences in drinking can be basedon characteristics directly related to region (suchas historical cultural differences or contextualcharacteristics of the region). A relationship be-tween regional context and drinking frequencyhas been reported in the US where the historicalwetness of a region was related to higher alcoholconsumption per drinker (Hilton, 1988). Thesefindings, however, have to be contrasted withresults from the British study, where alcohol usedid not seem to be related to regional factors(Duncan et ai, 1993).

Although international comparisons have to bemade with extreme care, structural differencesbetween the two societies are a possible expla-nation to the dissimilarity in the findings. Britishsociety has been described to have clear andpersistent differences between social statusgroups, where region of residence is closely re-lated to social status (e.g. Townsend, Phillimore& Beattie, 1988). It is probable, then, that

Page 8: Regional differences in drinking among Finnish adolescents

64 Sakari Karvonen

adjusting for social status has, at the same time,adjusted for regional variance. The Finnish so-ciety, on the other hand, has been characterizedwith relative lack of differences between socialstatus groups (Makela 1985). In a society such asthis regional differences, if they exist, are likely toremain apparent when the socio-demographicdifferences between regions have been adjusted.

According to our study a clear regional div-ision in alcohol use exists in Finland. It appearsto have been persistent during the 1980s and isfound both among adolescents and adults (Sim-pura & Lahti, 1988; Karvonen ei al., 1991). Thepersistence of regional differences in alcohol con-sumption was also reponed by Hilton (1988)from the US. The constancy of the differencesleads one to search the underlying causes for thedifferences between regions from factors thathave remained constant. In this study it was notpossible to determine whether the regional dif-ferences reflected the regional context of alcoholuse (as in the US) or some other historicalcultural feature. Further comparative studies aretherefore required to decide which are thespecific factors that contribute to the persistenceof regional differences.

ReferencesAiiLSTROM, S. (1983) Alkoholin kaytto [Alcohol use],

in: RlMPEiA M. ei al. (Eds) Nuonen terveystavatSuomessa [Health habits among Finnish Youth], SarjaTutkimukset 4/1983, pp. 131-164 (Helsinki, Fin-land, Publications of the National Board of Health).

AiTKiN, M., ANDERSON, D . , FRANCIS, B . & HINDE, J.

(1990) Statistical Modelling in GLIM, Oxford Statisti-cal Science Series 4 (Oxford, Oxford Science Publica-tions) .

ALCOHOL STATISTICAL YEARBOOK (1992) (Helsinki,

The Finnish State Alcohol Company).BLAXTER, M . (1990) Healih and Lifestyles (London,

Tavistock/Routledge).CENTRAL STATISTICS OFFICE OF FINIJUMD (1975)

Classification of Occupations (Helsinki, Finland, Cen-tral Statistics Office).

DUNCAN, C , JONES, K . & MOON, G . (1993) Do places

matter? A multi-level analysis of regional variationsin health-related behaviour in Britain, Social Scienceand Medicine, 37, pp. 725-733.

HILTON, M . (1988) Regional diversity in United Statesdrinking practices, British Journal of Addiction, 83,pp. 519-532.

KARVONEN, S., AHLSTROM, S. & RAHKONEN, O .

(1991) Nuorten alkoholin kaytto laaneittain vuosina1977-1987 [Adolescents' alcohol consumption hyprovince in 1977-87], Alkoholipolitiikka, 56, pp.297-305.

MAKELA, K . (1985) Kulttuurisen muuntelun yhteisotli-nen rakenne [The societal structure of culturalmodifications], Sosiologia, 22, pp. 247-60.

N T T T (1979) Nuorten terveystapatutkimus.Tutkimusohjelman tausta ja tavoitteet [The JuvenileHealth Habit Study—Study design, methods andmaterial], Kansanterveystieteen julkaisuja M48/79(Tampere, Kansanterveystieteen laitokset)-

RAHKONEN, O . & AH15TR6.M, S. (1989) Trends indrinking habits among Finnish youth from 1973 to1987, British Journal of Addiction, 84, pp. 1075-1083.

RIMPELA, A. , KARVONEN, S., RIMPEIA, M . & SIIVOLA,

M. (1990) Nuorten terveystottumusten vdestdryhmittdi-set erot ja elinolot I977-J987 [Socio-economic and re-gional differences in health habits of young people inFinland tn }977-87f, Sarja Tutkimukset 1/1990(Helsinki, Finland, Publications of the NationalBoard of Health).

SiMPURA, J. & LAHTI, M . L . (1988) Juomatapojenalueellinen vaihtelu Suomessa vuonna 1984 [Theregional variation of drinking habits in Finland in1984], Alkoholipolitiikka, 53, pp. 114-121.

TOWNSEND, P. & DAVIDSON, N . (Eds) (1982) Inequal-ities in Health (Suffolk, Penguin Books).

Tow'NSEND, P., PHILLTMORE, P. & BEArnE, A. (1988)Health and Deprivation: Inequality and the North(Guilford and King's Lynn, Croom Helm).

VAI-KONEN, T . & MARTE1.IN, T. (1988) Occupationalclass and suicide: an example of the elaboration of arelationship. Research reports No. 222 (Helsinki, De-partment of Sociology, University of Helsinki).

WlLlJAMS, G. & DEBAKEY, S. (1992) Changes in levelsof alcohol consumption: United States, 1983-1988,British Journal of Addiction, 87, pp. 643-648.

Page 9: Regional differences in drinking among Finnish adolescents