Eastern European transition and suicide mortality

Social Science & Medicine 51 (2000) 1405±1420

www.elsevier.com/locate/socscimed

Eastern European transition and suicide mortality Ilkka Henrik MaÈkinen* Stockholm Centre on Health of Societies in Transition (SCOHOST), University College of South Stockholm, P.O. Box 4101, S-141 04 Huddinge, Sweden

Abstract The current paper seeks to systematize the discussion on the causes of the changes in Eastern European countries' suicide mortality during the last 15 years by analyzing the changes in relation to some common causes: alcohol consumption, economic changes, ``general pathogenic social stress'', political changes, and social disorganization. It is found that the developments in suicide have been very di€erent in di€erent countries, and that the same causes cannot apply to all of them. However, the relation between suicide mortality and social processes is obvious. A model consisting of the hypothetical general stress (as indicated by mortality/life expectancy), democratization, alcohol consumption, and social disorganization (with a period-dependent e€ect) predicted the percentual changes in the suicide rates in 16 out of the 28 Eastern Bloc countries in 1984±89 and 1989±94 fairly accurately, while it failed to do this for Albania, Poland, Romania, Slovakia, and the Caucasian and Central Asian newly independent states. Most interesting were the strong roles played by changes in life expectancy, the causes of which are discussed, and the fact that economic change seemed to lack explanatory power in multiple analyses. The data are subject to many potential sources of error, the small number of units and the large multicollinearity between the independent variables may distort the results. Nevertheless, the results indicate that the changes in Eastern European suicide mortality, both decreases and increases, may be explained with the same set of variables. However, more than one factor is needed, and the multicollinearity will continue to pose problems. 7 2000 Elsevier Science Ltd. All rights reserved. Keywords: Alcohol; Eastern Europe; Homicide; Mortality; Russia; Suicide

Introduction and aim Only ten years ago the ``Eastern Bloc'' was still an existing, albeit crumbling, giant in European political geography. At a time when former reluctant members of this Soviet-led constellation of countries are joining NATO, it is sometimes hard to remember that they have shared at least 40 years of common history.

* Tel.: +46-8-5858-8149; fax +46-8-5858-8450. E-mail address: [email protected] (I.H. MaÈkinen).

Moreover, 1989 did not mean the end of the commonalities: the dismantling of socialism has led to similar problems in many of these countries. Privatization of production has been accompanied by all kinds of adaptation problems, from new work ethics to the redirection of former military production. Unemployment has increased, and incomes are generally more unequal than before. In many countries, an increase in alcohol use (and abuse) is reported (Harkin, Anderson & Goos, 1997). The rapid transformation from state-controlled media to liberal media politics has introduced Western values (including their

0277-9536/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 0 0 ) 0 0 1 0 5 - 2

1406

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

extremes) with a rapidity that has led to concerns of ``anomic'' development. At the same time, the very integrity of the state has been at issue in many cases, ranging from peaceful separations to civil Ð and external Ð wars. One should perhaps not forget that the intended goal of the reforms is to give these societies a better future than socialism could deliver. Despite the problems, the new, international economy and the civil rights and freedoms may o€er a varying measure of hope and, in some cases, lead to a ``rich man's crisis'' as well. The development has been of interest from the perspective of public health, too. The most cited example is probably the dramatic rise in mortality (especially that of males) in the countries of the former USSR after 1990, after the marked fall of 1984±87 (Shapiro, 1995; Leon et al., 1997). This has, quite naturally, led to discussions of the causes of this speci®c development, as well as an interest in the general relation between societal change and public health. Although not a disease, suicide mortality has been no exception. Its changes have often been very abrupt (see Table 1), and their size almost unprecedented as far as we know1. Despite the gravity of the problems per se (Sartorius, 1995; Bille-Brahe, 1998), and their apparent importance for the understanding of suicide in general, they have until now not been described or analyzed in the literature to the extent that one would perhaps expect. The most visible (and most frequently described) developments here are the large falls in the suicide rates in Russia and some other former Soviet republics in 1984±87, (Ambrumova & Postovalova, 1991; VaÈrnik & Wasserman, 1992; Wasserman et al., 1997), followed in some cases by even larger rises, starting some time 1986±88 (Wasserman & VaÈrnik, 1994; Sartorius, 1995; Mokhovikov & Donets, 1996; Lester, 1998; Orlova, 1998). Decreases in the 1980s have also been noted in other Eastern and Central European countries (Ho€meister, Wiesner, Junge & Kant, 1990; Kolozsi, 1990; HorazdovskyÂ, 1993), and the situation in the 1990s has raised concerns in these, too (Biro & Selakovic-Bursic, 1996; Gilinskii & Rumyantseva, 1998; Jarosz, 1998). The developments in suicide mortality have been more or less simultaneous with the great societal transformation. Thus, the tendency to ascribe changes 1 In order to ®nd a Swedish counterpart to the 50-70% increase in suicide in Russia and the Baltic countries between 1990 and 1994 one has to go back to 1815±18 (+2/3), or to 1918±21 (+55%). Moreover, the increase in the East started from an already high level.

in suicide mortality to simple psychological causes has this time not dominated the explanations. Even personality-based causal explanations like the ``PostSoviet syndrome'' (Mokhovikov & Donets, 1996; Mokhovikov, 1998) are linked to actual social themes. The most frequently proposed cause of the changes in suicide rates has probably been alcohol consumption (Ambrumova & Postovalova, 1991; Wasserman, VaÈrnik & Eklund, 1994; Wasserman & VaÈrnik, 1998b), with or without a speci®cation of the role it is sup-

Table 1 Percentual change in the suicide rate (E950±959 per 100,000 inhabitants) between 1984±89 and 1989±94 in 28 former Eastern Bloc countries Country

Change 1984±89

Change 1989±94

Albaniaa Armeniab Azerbaijan Bosnia and Hercegovina Belarusc Bulgaria Croatiad Czech Republic East Germanye Estonia Georgiaf Hungary Kazakhstan Kyrgyzstan Latvia Lithuania FYR Macedoniag Moldova Poland Romania Russia Slovakia Slovenia Tajikistanh Turkmenia Ukraine Uzbekistan Yugoslaviai

n.a. 20.0 ÿ31.1 n.a. ÿ26.7 ÿ3.6 ÿ1.8 ÿ9.3 ÿ18.4 ÿ22.5 ÿ4.2 ÿ9.4 ÿ25.5 ÿ17.4 ÿ24.6 ÿ25.1 n.a. ÿ27.0 ÿ19.3 n.a. ÿ32.1 +10.9 ÿ5.3 ÿ25.8 ÿ17.4 ÿ20.0 ÿ15.7 ÿ3.4

+9.5 +29.2 ÿ80.6 n.a. +40.6 +6.1 +4.1 ÿ2.2 ÿ27.1 +60.4 ÿ43.5 ÿ15.1 +21.9 +6.5 +58.4 +69.0 ÿ18.3 +5.9 +26.5 +16.5 +62.0 ÿ20.9 ÿ8.0 ÿ19.6 ÿ18.3 +26.4 ÿ18.7 +13.6

a

Albania 1989±93. Armenia 1994: estimate based on the standardized rate. c Belarus 1994: estimate based on the standardized rate. d Croatia 1985±89. e East Germany 1990±94: ``New LaÈnder and East Berlin''. f Georgia 1994: estimate based on the standardized rate. g Macedonia 1991±94; estimates based on the standardized rate. h Tajikistan 1994: estimate based on the standardized rate. i Serbia and Montenegro. b

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

posed to play in the process2. Other explanations include the political developments and the optimistic or frustrated feelings they have supposedly led to (Kolozsi, 1990; HorazdovskyÂ, 1993; Sartorius, 1995; Watson, 1995; VaÈrnik, 1997b; Jarosz, 1998), general socio-economic circumstances, especially unemployment (GailieneÂ, Domanskiene & Keturakis, 1995; Sartorius, 1995; Biro & Selakovic-Bursic, 1996; Jarosz, 1998; Orlova, 1998), the quality of medical services (Belau, 1991; Sartorius, 1995; Orlova, 1997), changes in norms and values (Wasserman & VaÈrnik, 1994; Orlova, 1998), broken social relations (Orlova, 1997), as well as the change in itself in the sense of its being anomic (Jarosz, 1998) according to the model ®rst proposed by Durkheim (1992). A more general version of the latter is the (social) stress that has been postulated to in¯uence all kinds of mortality (Shapiro, 1995; Shkolnikov et al., 1998; Leon & Shkolnikov, 1998). The existence of so-called ``social correlates of suicide'' (Sainsbury, Jenkins & Levey, 1980), thought to accompany suicide rates at all times, has previously been questioned (MaÈkinen, 1997a). Even as regards Eastern Europe, Dinkel and GoÈrtler (1994, cf. Schmidtke & Weinacker, 1994) have shown that the di€erence between East and West Germany in the 1980s was merely a result of large di€erences in some cohorts (notably those born around 1930). However, even these are supposed to be based on their (more or less politically determined) life experiences. Even if one subscribes to the theory of the cultural origin of the frequency of suicidal behavior and dismisses the idea of constant socio-structural correlates to suicide, the situation in Eastern Europe, where a number of countries are going through more or less similar processes of societal change, tempts one to try to identify some common denominator(s) for these developments in suicide mortality. So far, the analyses have been fewer than the hypotheses Ð one exception being those of alcohol consumption and suicide carried out by Wasserman, VaÈrnik, and Eklund, (1994, 1998). Even these have been criticized by Lester (1998). The distribution of suicide mortality demonstrates large and persistent di€erences between nations. There 2 Voytsekhovich and Redko: ``The increase in alcoholism, in itself a ``facilitating condition for the execution of suicide'' [ . . .], now becomes a catalyzator in the ``sociopsychological disadaptation of personality'' [ . . .] as a result of the economic problems, the worsening quality of life, and the spread of micro-social con¯icts'' (1994, p. 57, my translation). 3 For the di€erence between the level and the shape of a mortality curve see, e.g. Vassin & Costello, 1997. 4 For the sake of brevity, ``East Germany'' will below be used for that area even after 1990.

1407

have been previous attempts to classify the Eastern European countries according to their di€erent suicide mortality rate pro®les, such as Smidovich's (1990) division of the Soviet republics into ``European'' and ``Asian'' types. Dinkel and GoÈrtler (1994), among others, have noted that the Eastern German suicide pro®le (and its development) largely resembles that of West Germany. It has also been observed that the Eastern Slavic countries have a tendency towards a suicide mortality peak in middle age instead of, or alongside, the more usual old-age one (VaÈrnik, Wasserman, Dankowicz & Eklund, 1998a). In this study, suicide is seen as a mainly cultural phenomenon (Kral, 1994). Not only the suicide rates, but also the distribution of suicides in the population vary greatly between countries, re¯ecting perhaps the di€erent cultural contexts of the act. Because the mortality patterns themselves are di€erent, we did not expect the eventual social factors to in¯uence them all in the same way3. In the analysis, this was acknowledged by ®rst grouping the countries according to their suicide pro®les. The aim of this study is to try to assist in systematizing the hypothesis-building by describing the development of suicide mortality in Eastern Europe, thereby investigating the covariation between changes in some of the hypothesized causes (alcohol consumption, economic situation, general social stress, political situation, societal (dis)organization; see above) and the changes in the suicide rates. In order to deliver a more substantial picture of the development, we not only attempt to describe the changes in suicide mortality, but also those in the national suicide mortality pro®les, consisting of both suicide rates and their distribution in the population (MaÈkinen 1997b).

Data Due to considerable diculties in data collection caused by earlier censorship of suicide and homicide data in many countries (Bardehle & Casper, 1990; VaÈrnik, 1997b), current lack of reports from others, wars, and not least the changing political situation which has deleted some units while creating others, the research period was limited to 1984±94. These years were also expected to coincide best with those e€ects on suicide mortality which we aimed to identify. Changes in both independent and dependent variables during the subperiods of 1984±89 and 1989±94 were registered in all the 27 states existing in the territory of the former Eastern Bloc in 1994 and in the new LaÈnder and East Berlin of Germany4. For economic change and homicide rates, we were compelled to use 1985±89 for data availability reasons.

1408

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

The main source of all suicide data was the World Health Organization, its yearbooks (World Health Statistics Annual) and its data base Health For All. These were completed by data from the pioneering works of VaÈrnik, Wasserman and others (VaÈrnik & Wasserman, 1992; Wasserman, VaÈrnik & Eklund, 1994; VaÈrnik, 1997a,b) and from national publications, especially regarding East Germany and the Yugoslav republics. Despite occasional discrepancies caused by uncertain population ®gures and data reported at di€erent stages of processing, the ®gures seem to be consistent to a satisfactory degree. The overall reliability of older Soviet data has been found satisfactory by VaÈrnik and Wasserman (VaÈrnik, 1997b; Wasserman & VaÈrnik, 1998a,b), with some uncertainties concerning the Caucasian and Central Asian republics. The main dependent variable was the rate of ``certain'' suicide (E950±959 in ICD-9, class 173 in the 1988 Soviet classi®cation) per 100,000 inhabitants per year. Only six data values out of 84 were missing completely. In three cases, the data year di€ered from the ideal one, and in six the crude rate was estimated from the age-standardized rate5. The general validity and comparability of suicide rates have been subject to much discussion (O'Carroll, 1989), which will not be taken up here. However, the main validity problem in this study was without doubt the exclusion, due to uncertainty of their real nature, of deaths with uncertain intention (E980±989 in ICD-9, class 175 in the 1988 Soviet classi®cation), which in Western Europe are judged mostly to consist of suicides (Barraclough & Hughes, 1987). Their number has dramatically risen in some countries (Belarus, Kazakhstan, Latvia, Russia) during the period, but so far no satisfactory explanations of this phenomenon have been proposed. However, the suicide increase is accompanied by a rise in ``certain'' suicide as well. A parallelism in trends between the two classes has been noted earlier for Soviet Union (Wasserman & VaÈrnik, 1998a). The independent variables tested were the changes in alcohol consumption, economic situation, the level of general pathogenic social stress, political situation, and societal (dis)organization. These were collected and/or constructed as follows: Change in alcohol consumption The main source here were the ®gures of per 5

The age-standardized rate could not be used because it was unavailable in even more cases. It was not necessary either, since the changes in the general age structure hardly in¯uence those in suicide rates during a 10-year period. The actual errors are likely to be small.

capita consumption reported to the WHO. These were controlled with the help of data on changes in consumer expenditure on alcoholic drinks, liver cirrhosis mortality ®gures (where appropriate), and other sources, and the reported trend was in some cases slightly revised. In four cases the estimates were made without ocial consumption data. In cases where the reported ®gures were contradictory (e.g. Lithuania and Ukraine, 1989±94), the value was set at missing. The change was divided into 5 classes ranging from ÿ2 (large fall) via 0 (even trend) to 2 (large rise). In the ®nal analysis, data for 1984±89 were missing for two countries, and those for 1989± 94 for nine. Change in the economic situation Estimating the development of Eastern Bloc economies is traditionally very dicult due to di€erent measurement methods. In particular, the Gross National Product of the former socialist countries has been shown to have been exaggerated. Unemployment, previously politically ``non-existent'', later hard to estimate, may not be a reliable indicator. The task was facilitated by dividing the research period into subperiods, whereby the development during the ®rst period was estimated with the help of the gross industry output (SNA) according to World Bank data (Marer, Arvay, O'Connor, Schrenk & Swanson, 1992). The estimates were di€erentiated within the Soviet Union and Yugoslavia by use of national statistics on industry development. They were also checked against the available ®gures on GNP and the development of the output of the service sector, in the light of which the reported trends could be slightly revised. The basic measure for the development during the second period was the annual trend in industry output in 1987±97 as reported to the World Bank. These trends were checked against the available ®gures on GNP development and unemployment, and against other reports (e.g. Batalden & Batalden, 1997). In most cases a consistent estimate could be given, but one case for 1989±94 was set at missing. The values given for the change were numerically identical with those for change in alcohol consumption. Change in the level of general pathogenic social stress Since the stress hypotheses present the stress as related to several kinds of mortality, it was estimated with the help of relative changes in the average life expectancy at birth. Data were missing in three cases for 1984, and in one case for 1989 also. In 14 cases the data year fell short of the ideal one; in eight, the total

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

life expectancy was calculated from the sum of male and female ®gures (which in most cases should not result in too much error). Some problems with this indicator are a) its connection, per de®nitio, with suicide mortality, and b) its multifactorial nature (see Discussion). Political situation (democratization) The estimate for democratic change was constructed with regard to four basic variables: national and cultural rights of the majority populations6, the possibility of free elections resulting in a democratic government, freedom of speech in mass media, and the independence of the judicial system as indicated by both formal and informal criteria. Each unit was given a score between 0±4 on this scale, so the values for its change could theoretically vary between ÿ4 and 4. Change in societal (dis)organization We wanted to capture the amount of ``lawlessness'' in the Eastern European countries. This was estimated by using the homicide rate per 100,000 and year (i.e. deaths as result of homicide reported to WHO), and the relative changes in it, as an indicator. Here, the ®gures for some countries had to be discarded7, and others modi®ed8. For most countries, however, an acceptable ®gure could be obtained: data for 1985 were missing for four countries, those for 1989 for one, and those for 1994 for three. In ®ve cases, the data year di€ered from the ideal one; in nine, the crude homicide rate was estimated from the age-standardized rate. The multiple problems associated with the data collection and the naturally uncertain nature of the judgements may well have in¯uenced the results. The judgements of the values of the independent variables, even if far from perfect, are in no way unsystematic. 6

Incl. the absence of serious internal or international strife. Georgia reported only 0.4 cases per 100,000 in 1994, a very low ®gure which must be checked before being accepted. 8 Azerbaijan's ®gures until 1994 seemed to include some victims of the war over Nagorno-Karabach (or of the internal con¯icts) as well: otherwise the ®gure of over 100/100,000 for males is hard to understand, even if the victims of war were classi®ed in their own category (E990±999 in ICD-9). Moreover, since the ®gures fell by 4/5 in 1995 (the year after the armistice in Nagorno±Karabach), that year was used instead. 9 Due to uncertainties about the absolute level of alcohol consumption and economic development, they were omitted from the static analyses.

1409

Nevertheless, there is a possibility of any results being in¯uenced by weaknesses in data quality.

Method The countries were ®rst analyzed with regard to the chosen suicide-mortality indicators by dividing them into groups according to the most important variables, and testing several possibilities. The goal of this was to locate clusters of countries with more or less similar suicide mortality pro®les at the beginning of the research period. In order to see whether these clusters persisted during the changing developments, similar analyses were performed for the middle and end points as well. Separate forms of analysis were then employed to describe and analyze the development of the clusters more closely during the subperiods of 1984±89 and 1989±94 with regard to key characteristics. The analysis of covariation between the independent variables and the main dependent variable, the suicide rate, started with simple correlation analyses (Pearson's r and Spearman's r(s) were used according to the construction of the variables involved) between the variable values9 in a pooled material, which was subsequently divided into periods and country-groups (clusters) in order to locate partial correlations. The same kind of analysis was performed for the changes in the independent variable values and those in the suicide rates. In multiple analysis, best linear regression models were developed in a similar order. Interactions between the variables as well as period and cluster e€ects were also tried out where appropriate. The models were evaluated on the basis of the signi®cance levels of the included variables, the number of cases they covered, and the percentages of variance explained by them.

Results of the analyses Description of the suicide mortality pro®les and changes in the groups of countries

7

The preliminary analysis of the countries according to the values of the main suicide-mortality variables in 1985, 1989, and 1993 (or nearest years) yielded the following clusters of countries: The high-suicide countries: 1. The ``high-suicide, unequal sex distribution'' group consisted of Belarus, Estonia, Kazakhstan, Latvia, Lithuania, Russia, Slovenia, and Ukraine. Characteristic for this group of countries was a

1410

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

higher-than-average suicide rate, high sex quota10, and a low age quota11. 2. The ``high-suicide, unequal age distribution'' group consisted of Croatia, East Germany, and Hungary. This group was distinguished from the ®rst one by its relatively low sex quota and a high age quota. In 1994, East Germany demonstrated strongly declining suicide rates.

The low-suicide countries: 3. The ``low-suicide, unequal sex distribution'' group was made up of Kyrgyzstan, Poland, Romania (for which data were missing for 1984), and Slovakia. These countries showed a lower-than-average suicide rate together with a high sex quota and a low age quota. Moldova seems to have experienced a real suicide mortality pro®le change during the period; its suicide level fell, while the sex quota rose and the age quota fell. As a result, it changed group from (2) above in 1984 to this one in 1994. 4. The ``low suicide, unequal age distribution'' group included Bulgaria, Czech Republic, FYR Macedonia, and Yugoslavia (Serbia and Montenegro). In these countries, a lower-than-average suicide rate was accompanied by a low sex quota but also by a higher-than-average age quota. 5. The ``low suicide, equal distribution'' group demonstrated lower-than-average values of both suicide rate, its sex quota, and the age quota. It consisted of Albania, Armenia, Azerbaijan, Georgia, Tajikistan, Turkmenistan, and Uzbekistan, i.e. the former Soviet republics with less Russian in¯uence, together with the only predominantly Muslim country of Europe. Due to the very low suicide rates in this group the quota values could vary considerably. This made the classi®cation more dicult here, but the patterns were consistent. Albania, Bosnia, FYR Macedonia, and Romania were missing from the ®rst analysis; Bosnia and Macedonia from the second one, but only Bosnia from the last one. For 1985, group-speci®c suicide data were 10 De®ned as the quota between the male suicide rate and the female suicide rate. 11 De®ned as the quota between the age group with the highest rate and that with the lowest rate, using the WHO 10-year age-group classi®cation (15±24, 25±34, 35±44, 45±54, 55±64, 65±74, 75-). High values are often accompanied by a high rate for the oldest age group.

missing partly for one further country and completely for two; for 1989, this was the case for one country (Serbia and Montenegro) partly; for 1993, data were partly missing for four further countries and completely for two. The groups of countries di€ered in several ways. A general comparative analysis of the 1985, 1989, and 1993 data is presented in Table 2. The group-wise developments between the chosen time points can be summarized as follows: 1. The ``high-suicide, unequal sex distribution'' group experienced a large drop in suicide rates 1985±89, especially for middle-aged males, followed by a large general increase in 1989±93, which also included women and the young. The oldest group was the most stable part of the picture. 2. In the ``high-suicide, unequal age distribution'' group the rates, stable in 1985±89, generally fell somewhat after 1989, especially among the middleaged and the elderly. 3. In the ``low-suicide, unequal sex distribution'' group the rates for men and those for older women tended to increase somewhat over the whole period. The overall trend was a slight increase 1989±93. 4. The ``low-suicide, unequal age distribution'' group showed a continuous fall in the suicide rate of the old, while that of the middle-aged increased. The overall trend was, however, a fall in the suicide rates. 5. The ``low suicide, equal distribution'' group, ®nally, demonstrated falling rates for the entire period in all groups except that of the youngest.

The concomitants of the suicide rate: single correlates Static analyses Usable data on the absolute levels of the independent variables existed for only three of them: homicide, life expectancy, and the estimated degree of political rights and freedoms. None of the correlates showed a stable, signi®cant relation with suicide rates. The correlation between homicide and suicide, signi®cantly positive in the total material (r = 0.43; p < 0.0001), declined between 1984 (r = 0.50) and 1989 (0.22), to reappear in 1994 (0.60). The positive correlation between the degree of democracy and the suicide rate seemed to emerge over time (from 0.27 in 1984 up to 0.64 ( p < 0.0003) in 1994) as the democratization process continued and the di€erences between countries became more marked. The static correlations between life expectancy and suicide rates were insigni®cant throughout.

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

Dynamic analyses The analysis was then extended to the changes in the values of the variables, whereby all ®ve independent variables could be used. The correlates were counted for both subperiods (1984±89 and 1989±94), as well as for pooled data (see Table 3). The analysis of the changes con®rmed the earlier view, according to which the correlates of the levels of suicide and those of the changes in suicide rate need not be the same (MaÈkinen, 1997a). None of the corre-

1411

lates demonstrated a constant relation with both the suicide rate level and its changes. Increases in alcohol consumption, as well as decreases in life expectancy, seemed to be connected with simultaneous increases in suicide rates. In order to ®nd out whether these relations were constant across the groups of countries, the next analysis was performed on the groups presented earlier (see Table 4). It became obvious that the correlations were largely, if not entirely, steered by the ``high-suicide, unequal sex distribution'' group, in

Table 2 Averages of the values of some suicide±mortality related variables by country group in 1985, 1989, and 1993 (For the country groups, see above) Variable

Group 1

Group 2

Group 3

Group 4

Group 5

Total suicide rate 1985 Total suicide rate 1989 Total suicide rate 1993

28.3 25.1 33.4

29.7 29.8 26.1

13.6 13.6 14.1

16.7 16.3 15.6

5.4 4.5 3.6

Male suicide rate 1985 Male suicide rate 1989 Male suicide rate 1993

47.3 40.3 59.0

43.9 42.8 39.1

20.3 21.5 23.5

20.9 23.2 22.0

7.7 6.3 5.2

Female suicide rate 1985 Female suicide rate 1989 Female suicide rate 1993

11.5 11.5 12.9

16.6 17.6 14.0

5.1 6.0 5.3

9.0 9.5 9.4

3.2 2.8 2.0

Suicide rate/15±24 yrs 1985 Suicide rate/15±24 yrs 1989 Suicide rate/15±24 yrs 1993

15.9 15.6 22.5

13.1 12.2 13.5

9.2 9.4 9.4

9.2 9.1 10.6

5.1 4.6 6.2

Suicide rate/45±54 yrs 1985 Suicide rate/45±54 yrs 1989 Suicide rate/45±54 yrs 1993

52.9 44.7 63.5

48.1 43.8 41.0

27.3 26.4 27.8

16.7 21.7 23.3

12.1 10.2 7.1

Suicide rate/75- yrs 1985 Suicide rate/75- yrs 1989 Suicide rate/75- yrs 1993

44.0 45.7 46.6

76.9 87.9 71.3

16.7 27.2 23.4

80.2 64.5 59.1

11.4 11.0 8.1

Sex quota 1985 Sex quota 1989 Sex quota 1993

4.17 3.65 4.57

2.76 2.40 2.75

4.08 3.72 4.53

2.33 2.44 2.32

2.44 2.48 2.50

Age quota 1985 Age quota 1989 Age quota 1993

3.46 3.50 3.03

6.02 7.27 5.83

3.13 3.26 3.04

8.72 7.18 5.60

4.35 2.98 2.36

Age quota/men 1985 Age quota/men 1989 Age quota/men 1993

3.97 4.43 3.48

7.04 8.90 6.63

3.29 3.54 3.30

10.25 9.30 6.05

5.34 3.81 2.68

Age quota/women 1985 Age quota/women 1989 Age quota/women 1993

5.00 4.55 3.62

7.44 9.66 8.71

3.28 3.63 4.79

7.81 7.92 8.43

6.71 5.05 6.74

1412

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

Table 3 Correlations between the changes in the levels of alcohol consumption, economic development, homicide, life expectancy, political rights and freedoms, and changes in suicide mortality in former Eastern Bloc countries 1984±94 Correlation with the suicide rate: Change in

p(r/r(s))a

r/r(s)

Alcohol consumption, both periods Alcohol consumption, 1984±89 Alcohol consumption, 1989±94

0.54 0.43 0.40

Economic development, both periods Economic development, 1984±89 Economic development, 1989±94

0.0002 0.0418 (<0.10)

ÿ0.39 ÿ0.45 0.08

Homicide rate, both periods Homicide rate 1984±89 Homicide rate 1989±94

0.0051 0.0293 n.s.

0.23 0.16 0.02

n.s. n.s. n.s.

Life expectancy, both periods Life expectancy 1984±89 Life expectancy 1989±94

ÿ0.56 ÿ0.55 ÿ0.44

0.0001 0.0063 0.0216

Political rights and freedoms, both periods Political rights and freedoms 1984±89 Political rights and freedoms 1989±94

0.41 ÿ0.23 0.26

0.0026 n.s. n.s.

a

Spearman's rank correlation coecient was used for correlations with the estimated changes in alcohol consumption, economic development, and the political rights and freedoms.

which changes in all ®ve variables were highly signi®cantly correlated with those in the suicide rates. Although the signi®cances in other country groups

were possibly hampered by the fact that there were fewer observations in them, no other variable except the economic development in Group 2 (high suicide,

Table 4 Correlations between the changes in the levels of alcohol consumption, economic development, homicide, life expectancy, political rights and freedoms, and changes in suicide mortality in former Eastern Bloc countries 1984±89 and 1989±94, by country group (for the country groups, see above) Correlationsa with the changes in suicide rates and their p-values in Changes in Alcohol consumption p(r(s)) Economic development p(r(s)) Homicide p(r ) Life expectancy p(r ) Political rights and freedoms p(r(s))

Group 1

Group 2

Group 3

Group 4

Group 5

0.87 0.0001

ÿ0.03 n.s.

0.42 n.s.

ÿ0.45 n.s.

ÿ0.16 n.s.

ÿ0.83 0.0001

ÿ0.80 0.03

0.26 n.s.

ÿ0.33 n.s.

0.10 n.s.

0.39 n.s.

0.04 n.s.

0.63 n.s.

ÿ0.31 n.s.

ÿ0.94 0.0001

ÿ0.44 n.s.

ÿ0.36 n.s.

ÿ0.38 n.s.

0.32 n.s.

0.85 0.0001

ÿ0.18 n.s.

0.08 n.s.

0.27 n.s.

0.07 n.s.

0.82 0.0001

a Spearman's rank correlation coecient was used for correlations with the estimated changes in alcohol consumption, economic development, and the political rights and freedoms.

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

unequal age distribution) reached statistical signi®cance in these. In order to test whether the missing signi®cant relations could depend entirely on the lesser number of observations, the data were re-analyzed using various combinations of the groups of countries whose correlations went into the same main direction. Thus, changes in the economy reached signi®cance as a correlate also when Group 4 was added to Group 2 (r(s)=ÿ0.66; p < 0.02), while changes in the life expectancy did so in the combined Groups 2, 3, and 4 (r=ÿ0.47; p < 0.04). Summing up, changes in life expectancy at birth seemed to correlate signi®cantly negatively with changes in suicide mortality in all groups of countries with the exception of the one de®ned by low suicide rates and equal distribution, where most of the countries were either Caucasian or Central Asian newly-independent states. However, life expectancy as such did not seem to be connected with the basic level of national suicide rates. At the same time, changes in the economy also correlated negatively with changes in suicide rates in most of the countries with the exception of two groups: the ``low suicide, equal distribution'' group mentioned above, and the one de®ned by low suicide rates together with unequal sex distribution, which consisted chie¯y of Eastern Central European countries plus Kyrgyzstan. This correlation seemed, however, to have been valid overall only during the last years of the socialist period, 1984±89. After the economic change, the relation disappeared. As for alcohol consumption, changes therein seem to have accompanied those in suicide mortality only in Group 1, demonstrating high suicide rates and

1413

Table 5 The best linear regression model for all countries, both periods (R2=0.497) Independent variable

b estimate

se (b )

p

Intercept Change in Change in Change in Change in

ÿ7.64 6.31 ÿ3.06 ÿ2.08

3.73 2.57 1.50 0.80

0.048 0.019 0.048 0.013

alcohol consumption life expectancy alcohol consumption life expectancy

unequal sex distribution. In other groups, the relation can even be the reverse. In the ®rst-mentioned group, democratization was also strongly linked to an increase in suicide, just as its level correlated positively with that of suicide in all countries in 1989 and 1993. Finally, the level of homicide correlated highly positively with that of suicide on two occasions, but the correlation disappeared in the intervening period. Changes in homicide were signi®cantly correlated with those in suicide in the ®rst group of countries only. A closer investigation of the relations between the variables in the ®rst group of countries con®rmed that the strong relations there were not caused by outliers or other statistical artefacts. However, they were all con®ned to their own periods: during 1984± 89, the development of alcohol consumption, together with economic development and democratization, seemed to di€erentiate best between the changes in suicide rates within that group. In 1989±94, these were replaced by the development of homicide mortality and life expectancy. While not being directly spurious, the correlations seem limited.

Table 6 The best linear regression models for 1984±89 and 1989±94 Beta estimates, their standard errors, and p-values in Independent variable Change in alcohol consumption se p Economic development se p

1984±89

1989±94

6.17 1.51 0.0005

16.09 5.34 0.0087

ÿ9.12 2.46 0.0013

20.06 6.29 0.0061

Change in life expectancy se p Adjusted R2

ÿ5.47 1.90 0.0116 0.705

0.531

1414

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

The concomitants of the suicide rate: multiple approaches The next stage was a simultaneous analysis of all the independent variables. Linear regression was used in order to ®nd a model which would best describe the relations between changes in them and those in suicide mortality. The best model for all countries and both periods is presented in Table 5 (only the relevant variables are shown for each model). The best combination of independent variables was that of alcohol consumption, life expectancy, and their interaction. Dividing the material into the two periods, the two models presented in Table 6 were obtained. Here, the parameter for economic development changed its sign between the periods. Alcohol consumption could be included in the best model in both periods, while life expectancy was signi®cant in 1989±94 only. The relatively simple ``alcohol-andpoverty'' model could explain 70% of the variation in suicide rate change in 1984±89. The correlation analysis indicated that the relations between variables varied greatly between the country groups. The analysis was accordingly performed on the groupings presented above. The number of cases was so small that the analysis could only be carried out for both periods together. Table 7 summarizes the results. These analyses con®rmed the split nature of the picture. For Group 1, the model was very powerful but not entirely satisfactory since it included a period term. In no other group could alcohol consumption be used as a signi®cant independent variable. Economic development was included in the model in two 12 Here we also tested the e€ects of the assumption that the Romanian suicide rate would have been decreasing 1984±89. According to Cozman (1995), the rates were rapidly declining in Cluj-Napoca, the capital of the country's high-suicide area during this period. Group 3 was re-analyzed assuming a 30% decrease. The correlations were essentially unchanged; in the regression analysis democratization dropped out as signi®cant variable. Changes in life expectancy and economy attained an adjusted R2 of 0.571. 13 However, the signi®cance of single variables varied greatly in period-wise analyses. 14 Prediction from a de®cient model was made for cases in which some of the independent variables were missing: for Kazakhstan, Lithuania, Ukraine, and Yugoslavia in 1989±94, it was the alcohol consumption. The model was applied to them without that variable. For East Germany 1984±89 the missing variable was homicide, for East Germany 1989±94, both alcohol consumption and homicide, and for Yugoslavia 1984±89, alcohol consumption and life expectancy.

groups of mainly Central European countries, but with a di€erent sign. Changes in life expectancy and democratization both appeared as signi®cant predictors in the model for Group 312. For the last two groups, no satisfactory model at all could be achieved using the independent variables at hand. To control for the possibility that the split had been caused by the very small number of cases in many groups, some combinations of groups were reanalyzed. This revealed that Groups 2±4, consisting almost entirely of Central European countries, could not be combined, since no common model with signi®cant predictors could be assembled for them. However, Groups 2 and 4 could be combined, and the new model obtained was more powerful than the previous ones for either of them. For a combination of these two groups and Group 1, a common model with an R2 value (0.926) in between those of its two components could be produced. The changes in life expectancy and homicide were the most signi®cant predictors. This was the model that best covered the largest number of cases (see Table 8). It was, moreover, highly signi®cant in both periods (R2=0.916 and 0.944, respectively, without period interaction in homicide13). The correlation (r ) between the predicted and observed values (including those predicted from a de®cient14 model) was 0.95 ( p < 0.0001). The only anomaly was the changing sign of the homicide change (negative during the ®rst period, positive during the second). The residuals were not signi®cantly correlated with the dependent variable. The model, and the variables included, would have been signi®cant (R2=0.812) even if Group 3 had been added: however, the values predicted for Moldova in the later period, for Poland, Romania, and partly also for Slovakia, would have been highly erroneous. The model was also tested for the possibility that the greater variation in life expectancy and homicide change data (which were on interval level as compared to other, ordinal-level independent variables) could have been the cause of their prominence in the model. The results were clearly negative. The models are merely attempts at combining the ``best'' of the variables. However, the problem of multicollinearity is overwhelming: among all countries, changes in nearly all the independent variables are signi®cantly related to each other as shown in Table 9. Discussion As mentioned above, there are uncertainties related to the ®gures used here. The greatest of them pertain to suicides possibly ``hidden'' in the category of ``unde-

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

1415

Table 7 The best linear regression models for the country groups, both periods (for the country groups, see above) Beta estimates, their standard errors, and p-values in Changes in

Group 1

Intercept se p

22.51 5.45 0.0026

Alcohol consumption se p

24.12 2.81 0.0001

Group 2

Group 3

Group 4

Group 5

ÿ24.79 6.74 0.0213 ÿ7.76 2.96 0.0587

Economic development se p

ÿ13.76 3.60 0.0088

9.28 5.07 0.1047

41.44 7.54 0.0054

Homicide se p Life expectancy se p

ÿ10.47 1.97 0.0061

Political rights and freedoms se p Perioda se p

10.86 3.40 0.0331 ÿ48.29 10.29 0.0011

PeriodHomicideb se p

0.24 0.02 0.0001

R2 Adjusted R2

0.991

a b

ÿ35.26 14.07 0.0418

0.660

0.824

0.540

0.351

Coded as a dummy variable; 0 for 1984±89 and 1 for 1989±94. Homicide was not signi®cant in its own right when its interaction with the period was added to the model.

termined intention'', the use of which has increased very strongly in some countries since 1989. However, Table 8 The best linear regression model for the combined country groups 1, 2 and 4, both periods (R2=0.926) Independent Variable

b estimate se (b ) p

Change in alcohol consumption 4.41 Change in life expectancy ÿ5.06 Democratization ÿ4.96 Change in homicide rate Perioda 0.17

1.71 1.10 1.71 0.05

0.0186 0.0002 0.0092 0.0015

a Period was coded as a dummy variable; 0 for 1984±89 and 1 for 1989±94.

in many of the countries studied here hanging, less likely to be misclassi®ed, is the method used in the majority of cases (Voytsekhovich & Redko, 1994; Gailiene et al., 1995; VaÈrnik, 1997a; Jarosz, 1998). The results of this study should be seen as provisional until better analyses can be carried out. They are based on a small amount of data, some of it perhaps not fully reliable, some of it hard to estimate, which is also beset by problems of multicollinearity. Nevertheless, the basic results are trustworthy. In order to check the graded estimates of changes in alcohol consumption, economy, and political situation, reanalyses were made with fewer data value classes. These did not change the results much.

1416

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

Table 9 Correlations between the changes in the levels of alcohol consumption, economic development, homicide, life expectancy, and political rights and freedoms in Eastern Bloc countries, both periods Correlations (and their p-valuesa) with changes in: Changes in

Alcohol consumption

Economic development

Alcohol consumption p(r(s))

1.00 0.0000

ÿ0.49 0.0007

Economic development p(r(s))

1.00 0.0000

Homicide p(r/r(s)))

Homicide 0.20 n.s.

Life expectancy

Political rights

ÿ0.50 0.0005

0.56 0.0001

ÿ0.48 0.0006

0.58 0.0001

ÿ0.32 0.0158

1.00 0.0000

ÿ0.57 0.0001

0.32 0.0263

1.00 0.0000

ÿ0.35 0.0114

Life expectancy p(r/r(s)) Political rights and freedoms p(r(s))

1.00 0.0000

a Spearman's rank correlation coecient was used for all correlations involving the estimated changes in alcohol consumption, economic development, and the political rights and freedoms.

Transition and suicide mortality

The division of Eastern European countries

Strictly speaking, we cannot identify any general trend of suicide mortality in Eastern Europe, or in the ``newly independent states'' of the former Soviet Union. The countries of the former Eastern Bloc could be divided into ®ve groups on the basis of their suicide mortality pro®les (the level of suicide and its age/sex distribution). In three of these groups, suicide rates tended to decline between 1984 and 1994; in one, they increased in 1989±94, and in one they both decreased (1984±89) and increased (1989±94) in a way that has attracted much attention and led to subsequent discussions of the relation between suicide and societal transformation. However, all the countries studied have been going through similar transformations during the research period. From our results it is obvious that rapid transformations of society do not per se necessarily produce more suicide, something which con¯icts with the classical Durkheimian theory (Durkheim, 1992). An explanation to this could be the intermediate role of culture (re¯ected in the di€erent suicide mortality pro®les): similar social factors can lead to di€erent outcomes in di€erent environments. Another obvious conclusion (in line with Durkheim) is that there do indeed exist forces, originating from outside the individual consciousness, which in¯uence in a most concrete manner the individual's propensity to commit suicide. Whether they exercise their in¯uence through serious psychopathology or some less grave forms of mental unhealth, is an interesting issue for further research.

The general mortality patterns of Eastern European countries are highly diverse (Kinkgade & Arriaga, 1997). The present division of countries was based on the form of suicide mortality, which we believed was re¯ecting cultural di€erences in relation to suicidal death. It also resulted in a more or less cultural-geographical division. Smidovich's notion of ``European'' and ``Asian'' types was supported inasmuch as most of the former Soviet republics genuinely did fall into two main groups: one of ``high suicide, unequal sex distribution'', including all the ``European'' republics (except for Moldova) plus Kazakhstan (with a sizeable Russian minority), and one of ``low suicide, equal distribution'' group, consisting of six Caucasian and Central Asian republics. The division of the pro®le groups was not essentially changed during the sometimes stormy period under study. Thus, the primacy of the enduring national patterns of suicide mortality (Diekstra, 1993; MaÈkinen & Wasserman, 1997) seems to have been once more con®rmed. Moreover, it seems that these patterns themselves have not been transformed in the transition process. Interpreting the model The level of suicide mortality and its changes are not necessarily explained by the same phenomena. The level may be determined by those aspects of society

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

which usually change very slowly (MaÈkinen, 1997a). On the other hand, there have been international trends such as those at the times of World War I (Chesnais, 1977/78) and the Great Depression (Sainsbury, Jenkins & Levey, 1980), which have in¯uenced most of the European countries very strongly, if perhaps not in a lasting manner. In our analysis, too, there seemed to be little consistency between the ``static'' and the ``dynamic'' correlations. No possible universal determinants of suicide were included among the present independent variables. It was, however, possible to construct a model in which the changes in life expectancy, alcohol consumption, democratization, and homicide (with a period-dependent e€ect) explained more than 92% of the variation in the changes in suicide rates in more than half of the countries during both periods, retaining its power across time. Overall, the predictive value of the model was surprisingly great, as was its persistency over the periods, taking the very dramatic developments in Group 1 especially into account. In this respect it succeeded much better than the attempts to ®nd corresponding ``social correlates of suicide'' for 18 Western European societies in the 1960s±1980s (Sainsbury et al., 1980; MaÈkinen, 1997a). It seems that di€erences in suicide rate development can be related to the above-mentioned variables in a multiple analysis. Why the variables do not seem to predict the rates equally well in all countries will be an interesting research theme for the future. Certainly, cultural di€erences may mean that the factors in¯uencing changes in suicide rates vary as well (as do individual motives (Bhatia et al., 1987)), but this hypothesis should perhaps not be uncritically extended to the geographically and culturally close Group 3 (Moldova 1989±94, Poland, Romania, Slovakia), even if it might constitute a correct interpretation of the absence of any good models for Group 5 (Albania along with the Caucasian and most Central Asian countries). The concept of ``general pathogenic social stress'' was here intended to refer to the unspeci®c, non-medical, non-economic factor which many authors, in slightly di€erent terms, assume greatly to in¯uence the mortality ®gures in Eastern Europe (Shapiro, 1995; Watson, 1995; Bobak & Marmot, 1996, even Bobak, Pikhart, Hertzman, Rose & Marmot, 1998). It is perhaps a little surprising that life expectancy at birth, thought to indicate the level of that hypothetical construct, appeared as the most signi®cant predictor in the model. The hypothesis, whatever its exact form, could perhaps link life expectancy and suicide mortality. However, the sources of the ``general stress'' would then lie outside the model variables, which would in itself be problematic. Other concepts, like that of general adaptation problems (Wasserman & VaÈrnik,

1417

1994), or ``neo-traditional forms of adaptation'' (Watson, 1995), not necessarily con®ned to those ®elds, could be useful here. One possible explanation could lie in the de®nitional relation between suicide mortality and life expectancy: changes in suicide mortality do of course in¯uence those in general mortality and in life expectancy as well. However, the role of suicide in general mortality is not great. Notzon, Komarov, Ermakov, Sempos, Marks and Sempos (1998) estimate the share of suicide in the change in Russian mortality between 1990 and 1994 to be 5.2%, which is too little to cause covariation on its own right. Life expectancy is also generally assumed to be in¯uenced by a multitude of factors (Notzon et al., 1998), and factors other than stress do in¯uence it, among them the level of health services. Judging from earlier comments on the former level of Eastern European psychiatric services (VaÈrnik & Wasserman, 1992) however, it seems contraintuitive to assume that the deterioration of these alone might cause a wave of suicide. Leon et al. (1997) are of the opinion that this deterioration of services cannot explain the changes in the general health situation either. Whatever the exact causal pathways between the changes in life expectancy and suicide mortality, their general levels do not seem to be associated at all. Any possible e€ect may be limited to a period of (perhaps speci®c) change. The relation is interesting, but needs closer investigation. The change in the homicide rate, the second most signi®cant variable, was a positive predictor of suicide rate change overall. It did, however, change its sign between the ®rst (negative) and second (positive) periods in the model. One tentative interpretation of this phenomenon could be that a certain amount of disorder in a totalitarian society Ð expressed in the rise of homicide Ð would be a sign of vitality (as already noted by Durkheim (1992)), while the larger increases during the second period would be indicative of a vitality which has become disorganization (Jarosz, 1998). The connection between homicide and suicide rates would be the general atmosphere, hopeful activity or chaos, both of which would have repercussions at individual level. Thus, the two periods seem to require somewhat di€erent explanations. The small number of observations is, however, a de®nite obstacle for closer investigation at our level. Democratization, a correlate to suicide rates in its own right (see Tables 3±5), became a negative predictor when the in¯uences of the three previous factors had been taken into account. Some support was thus given to the hypothesis that the hope of democratization has a bene®cial in¯uence on people. The role of alcohol consumption, earlier investigated by, among others, Wasserman, VaÈrnik, and Eklund

1418

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

(VaÈrnik, 1997b; Wasserman & VaÈrnik, 1998b; Wasserman, VaÈrnik & Eklund, 1994, 1998) was here also con®rmed for a larger group of countries. However, it did not seem to be the main factor among the highly intertwined predictors. The most interesting ®nding in causal terms was that alcohol consumption retained its position even as the economic and political changes were included in the model Ð it did not seem to be a simple re¯ection of these. Economic development, ®nally, seemed to exert less in¯uence on suicide rates than has been expected. Its behavior in the models was varying. Interpreted in Durkheimian terms, this could also mean that all kinds of change may lead to more suicide (Durkheim, 1992). However, such an interpretation is not in accordance with the most general results of this study, which show highly di€ering outcomes of the economic transformation. Moreover, the absolute economic changes do not correlate signi®cantly with changes in suicide rates. The large R2 value of the model is, of course, also explained by a small number of cases, heuristic choice of measurement periods, multicollinearity (simultaneous developments), and the aggregate-level units. Future tasks The most immediate research tasks should be to extend this rudimentary analysis into more units and, possibly, other time periods. Regional developments are of utmost interest here. Moreover, with the aid of better data the inequalities in economy (the size of the disadvantaged groups, the developments in groups and sectors Ð such as agriculture Ð experiencing rapid change) could perhaps be investigated as well. Research into cultural and attitudinal changes in the context of political and economic transformation is no less important, and might clarify some of the relations found here. The main problem will probably lie in their de®nition, operationalization, and, not least, their separation from the more concrete social factors. The cohort structure of suicide mortality naturally also needs an analysis. Unfortunately, a number of Eastern European countries may also provide the opportunity to study hypotheses on wartime suicide mortality. Most urgent, however, is the issue of the cases of ``undetermined'' intention. If the proportion of suicides among these is the same as would normally be expected in many Western countries, then the extent of the problem is even greater than appears to be the case. Acknowledgements I would like especially to thank PaÈr SpareÂn, Dr Med

Sc, Therese Reitan, PhD, Erik Hansen, MA, and Prof. Denny VaÊgeroÈ for their comments on my manuscript, as well as Ms Dominique Paturot, MA, for her help with data collection, and Mr Andrew Stickley, MA, and Dr Onja Grad for helpful information. This paper has been made ®nancially possible by the grant from the Swedish Baltic Sea Foundation to the project ``Social Development and Public Health in Eastern Europe'' at the University College of South Stockholm.

References Ambrumova, A. G., Postovalova, L. I. (1991). Sotsialnye i kliniko-psikhologicheskiye aspekty samoubiistv v sovremennom obshchestve, Obozreniye psikhiatrii i med.psikhologii im. V.M.Bekhtereva 26±38. Social and clinicalpsychological aspects of suicides in contemporary society. Bardehle, D. & Casper, W. (1990, eds.). Das Gesundheitswesen. Jahresgusundheitsbericht 1989 fuer das Gebiet der ehemaligen DDR. [The health care system. The annual health report 1989 for the territory of the former GDR]. Berlin: The Federal Health Ministry. Barraclough, B., & Hughes, J. (1987). Suicide. Clinical and epidemiological studies. London: Croom Helm. Batalden, S. K., & Batalden, S. L. (1997). The newly independent states of eurasia. Handbook of former Soviet Republics. Phoenix: ORYX. Belau, D. (1991). Zur Interpretation suizidalen Verhaltens im Kontext sozialer und politischer Ursachen - auf dem Hintergrund sozialer Erfarungen in der DDR (Toward the interpretation of suicidal behavior in the context of social and political factors Ð against the background of social experiences in the GDR). MMG, 16, 276±284. Bhatia, S. C., Khan, M. H., Mediratta, R. P., & Sharma, A. (1987). High risk suicide factors across cultures. The International Journal of Social Psychiatry, 33, 226±236. Bille-Brahe, U. (1998). Suicidal behaviour in Europe. The situation in the 1990s. Copenhagen: WHO European Regional Oce. Biro, M., & Selakovic-Bursic, S. (1996). Suicide, aggression, and war. Archives of Suicide Research, 2, 75±79. Bobak, M., & Marmot, M. (1996). East-West mortality divide and its potential explanations: proposed research agenda. BMJ, 312, 421±425. Bobak, M., Pikhart, H., Hertzman, C., Rose, R., & Marmot, M. (1998). Socioeconomic factors, perceived control and self-reported health in Russia. A cross-sectional survey. Social Science and Medicine, 47, 269±279. Chesnais, J.-C. (1977/78). As mortes violentas. Primeira parte: o homicõ dios e as execuc° oÄes capitais e o suicõ dios (Violent deaths. Homicides, death penalties, and suicides). Revista do Centro de Estudos Demogra®cos, 23. Cozman, D. (1995). Epidemiologia comportamentului suicidar in municipiul Cluj-Napoca, intre anii 1985±1991 (The epidemiology of suicidal behavior in Cluj-Napoca 1985± 1991). Romanian Journal of Legal Medicine, 3, 59±68. Diekstra, R. F. W. (1993). The epidemiology of suicide and

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420 parasuicide. Acta Psychiatrica Scandinavica, 371(Supplement), 9±20. Dinkel, H. D., & GoÈrtler, E. (1994). Die Suizidsterblichkeit der GeburtsjahrgaÈnge in beiden Teilen Deutschlands (Cohort suicide mortality in both parts of Germany). Sozial- und PraÈventivmedizin, 39, 198±208. Durkheim, E. (1992). Suicide. London: Routledge. GailieneÂ, D., DomanskieneÂ, V., & Keturakis, V. (1995). Suicide in Lithuania. Archives of Suicide Research, 1, 149± 158. Gilinskii, J., & Rumyantseva, G. (1998). Samoubiistva v Rossii (Suicide in Russia). Mir Rossii, 6. Harkin, A. M., Anderson, P., & Goos, C. (1997). Smoking, drinking and drug taking in the European region. Copenhagen: WHO Regional Oce for Europe. Ho€meister, H., Wiesner, G., Junge, B., & Kant, H. (1990). Selbstmordsterblichkeit in der DDR und in der Bundesrepublik Deutschland (Suicide mortality in the GDR and in the Federal Republic of Germany). MMW, 132, 603±609. HorazdovskyÂ, V. (1993). Sebevrazednost v Ceskoslovensku v letech 1975±1990 (Suicide rate in Czechoslovakia in 1975± 1990). Cesko-Slovenska Psychiatrie, 1993, 233±238. Jarosz, M. (1998). Suicide. Warsaw: Institute of Political Studies at Polish Academy of Sciences. Kinkgade, W. W., & Arriaga, A. A. (1997). Mortality in the new independent states: patterns and impacts. In J. A. Bobadilla, C. A. Costello, & F. Mitchell, Premature death in the new independent states (pp. 156±183). Washington: National Academy Press. Kolozsi, B. (1990). Erleichterung in der ungarischen Selbstmordepidemie (1988) und Verlauf ihrer Komponenten (A relief in the Hungarian suicide epidemic (1988) and its component factors). OÈsterreichische Zeitschrift fuer Soziologie, 15, 95±98. Kral, M. J. (1994). Suicide as social logic. Suicide and LifeThreatening Behavior, 24, 245±255. Leon, D. A., Chenet, L., Shkolnikov, V. M., Zakharov, S., Shapiro, J., Rakhmanova, G., Vassin, S., & McKee, M. (1997). Huge variation in Russian mortality rates 1984±94: artefact, alcohol, or what? The Lancet, 350, 383±388. Leon, D. A., & Shkolnikov, V. M. (1998). Social stress and the Russian mortality crisis. JAMA, 279, 790±791. Lester, D. (1998). Suicide and homicide after the fall of communist regimes. European Psychiatry, 13, 98±100. Marer, P., Arvay, J., O'Connor, J., Schrenk, M., & Swanson, D. (1992). Historically planned economies. A guide to the data. Washington: The World Bank. Mokhovikov, A. (1998). The posttotalitarian pattern of suicidal behavior: impact on youth. Crisis, 19, 8±11;14. Mokhovikov, A., & Donets, O. (1996). Suicide in the Ukraine: epidemiology, knowledge, and attitudes of the population. Crisis, 17, 128±134. MaÈkinen, I. H. (1997a). Are there social correlates to suicide? Social Science and Medicine, 44, 1919±1929. MaÈkinen, I. H. (1997b). On suicide in European countries. Some theoretical, legal, and historical views on suicide mortality and its concomitants. Stockholm: Almqvist & Wiksell. MaÈkinen, I. H., & Wasserman, D. (1997). Suicide prevention and cultural resistance: stability in European countries'

1419

suicide ranking, 1970±1988. Italian Journal of Suicidology, 7, 73±85. Notzon, F. C., Komarov, Y. M., Ermakov, S. P., Sempos, C. T., Marks, J. S., & Sempos, E. V. (1998). Causes of declining life expectancy in Russia. JAMA, 279, 793±800. O'Carroll, P. W. (1989). A consideration of the validity and reliability of suicide mortality data. Suicide and LifeThreatening Behavior, 19, 1±16. Orlova, I. B. (1997). Samoubiistvo kak indikator obshchestvennogo regressa. In G. Osipov, V. Levashov, & V. Lokosov, Rossiya: novyi etap neoliberalnykh reform. Sotsialnaya i sotsialno-politicheskaya situatsiya v Rossii v pervoi polovine 1997 goda (pp. 123±129). Moscow: Respublika. Orlova, I. B. (1998). Samoubiistvo - yavleniye sotsialnoye (Suicide Ð a social phenomenon). Sotsiologicheskiye issledovaniya, 8/1998, 69±73. Sainsbury, P., Jenkins, J., & Levey, A. (1980). The social correlates of suicide in Europe. In R. Farmer, & S. Hirsch, The suicide syndrome (pp. 38±53). London: Croom Helm. Sartorius, N. (1995). Recent changes in suicide rates in selected Eastern European and other European countries. International Psychogeriatrics, 7, 301±308. Schmidtke, A., & Weinacker, B. (1994). SuizidalitaÈt in der Bundesrepublik und den einzelnen BundeslaÈndern: Situation und Trends (Suicidality in the Federal Republic and in LaÈnder). Suizidprophylaxe, 1994, 4±16. Shapiro, J. (1995). The Russian mortality crisis and its causes. In A. AÊslund, Russian economic reform at risk (pp. 149± 178). London: Pinter. Shkolnikov, V. M., Cornia, G. A., Leon, D. A., & MesleÂ, F. (1998). Causes of the Russian mortality crisis: evidence and interpretations. World Development, 26, 1995±2011. Smidovich, S. G. (1990). Samoubiistva v zerkale statistiki (Suicides in the mirror of statistics). Sotsiologicheskiye issledovaniya, 4/1990, 74±79. VaÈrnik, A. (1997a). Enesetapud Eestis 1965±1995 (Suicide in Estonia 1965±1995). Tallinn: Estonian-Swedish Suicidology Institute. VaÈrnik, A. (1997b). Suicide in the Baltic countries and in the former republics of the USSR. Stockholm: National Swedish and Stockholm County Council of Suicide Research and Prevention & Dept. of Clinical Neuroscience at Karolinska Institute. VaÈrnik, A., & Wasserman, D. (1992). Acta Psychiatrica Scandinavica, 86, 86±88. VaÈrnik, A., Wasserman, D., Dankowicz, M., & Eklund, G. (1998a). Age-speci®c suicide rates in the Slavic and Baltic regions of the former USSR during Perestroika, in comparison with 22 European countries. Acta Psychiatrica Scandinavica, 98(Supplement), 20±25. Vassin, S. A., & Costello, C. A. (1997). Spatial, age, and cause-of-death patterns of mortality in Russia, 1988±1989. In J. A. Bobadilla, C. A. Costello, & F. Mitchell, Premature death in the new independent states (pp. 66±119). Washington: National Academy Press. Voytsekhovich, B. A., Redko, A. N. (1994). Pust umirayet tot, kto nye khochet zhit? (Should one who doesn't want to live die?). Kuban Scienti®c Medical Journal Edition, Krasnodar. Wasserman, D., Dankowicz, M., VaÈrnik, A., & Olsson, L.

1420

I.H. MaÈkinen / Social Science & Medicine 51 (2000) 1405±1420

(1997). Suicide trends in Europe, 1984±1990. In A. J. Botsis, C. R. Soldatos, & C. N. Stefanis, Suicide: biopsychosocial approaches (pp. 3±10). Amsterdam: Elsevier. Wasserman, D., & VaÈrnik, A. (1994). Increase in suicide among men in the Baltic countries. Lancet, 343, 1504± 1505. Wasserman, D., & VaÈrnik, A. (1998a). Reliability of statistics on violent death and suicide in the former USSR, 1970± 1990. Acta Psychiatrica Scandinavica, 98(Supplement), 34± 41. Wasserman, D., & VaÈrnik, A. (1998b). Suicide-preventive e€ects of Perestroika in the former USSR: the role of alco-

hol restriction. Acta Psychiatrica Scandinavica, 98(Supplement), 1±4. Wasserman, D., VaÈrnik, A., & Eklund, G. (1994). Male suicides and alcohol consumption in the former USSR. Acta Psychiatrica Scandinavica, 89, 306±313. Wasserman, D., VaÈrnik, A., & Eklund, G. (1998). Female suicides and alcohol consumption during Perestroika in the former USSR. Acta Psychiatrica Scandinavica, 98(Supplement), 26±33. Watson, P. (1995). Explaining rising mortality among men in Eastern Europe. Social Science and Medicine, 41, 923± 934.