Democratization and Political Change as Threats to Collective Sentiments: Testing Durkheim in Russia

Abstract

Durkheim argued that acute political crises result in increased homicide rates because they pose a threat to sentiments about the collective. Though crucial to Durkheim’s work on homicide, this idea remains untested. The authors took advantage of the natural experiment of the collapse of the Soviet Union to examine this hypothesis. Using data from Russian regions (N = 78) and controlling for measures of anomie and other covariates, the authors estimated the association between political change and change in homicide rates between 1991 and 2000. Results indicated that regions exhibiting less support for the Communist Party in 2000 (and thus greater change in political ideals because the Party had previously exercised complete control) were regions with greater increases in homicide rates. Thus, while democratization may be a positive development relative to the Communist juggernaut of the past, it appears that the swift political change in Russia is partially responsible for the higher rates of violence there following the collapse of the Soviet Union.

In this study, we drew upon Durkheim’s (1897/1979) ideas on threats against collective sentiments to examine the association between political change and homicide rates in Russia. Democratization and marketization¹ in Russia during the 1990s represented a significant shift away from the group-oriented norms of the past, thus presenting a clear threat not only to the deeply ingrained socialist ideology and norms of the Soviet era but also to long-standing Russian cultural traditions and values that had privileged the collective over the individual. The political changes were accompanied by a sharp increase in violence, and Russia’s homicide victimization rate of about thirty per one hundred thousand residents is nearly five times higher than in the United States and is among the highest in the world (Pridemore 2003a). Russia is a vast nation, however, and homicide levels and trends vary widely. Initial research suggested that several structural variables—including poverty, single-parent households, and aggregate alcohol consumption—are associated with the cross-sectional variation of Russian homicide rates (Pridemore 2005). Similar variables were shown also to have influenced the change in regional homicide rates during the 1990s (Andrienko 2001). Furthermore, Chamlin, Pridemore, and Cochran (2005) and Kim and Pridemore (2005) used interrupted time-series methods and change models, respectively, to test explicitly Durkheim’s hypothesis that rapid social change leads to societal deregulation and anomie and, in turn, to higher homicide rates. Both studies found consistent support for these ideas.

[I]t appears that the swift political change in Russia is partially responsible for the higher rates of violence there following the collapse of the Soviet Union.

Each of the studies mentioned above focused on the effects of socioeconomic factors and did not consider political characteristics. Yet both theory and empirical evidence suggest it is unwise to ignore such issues. For example, recent studies have shown the various ways in which the political features of a society can influence criminal justice and rates of interpersonal violence (Stucky, Heimer, and Lang 2005; Stucky 2003; Villarreal 2002). Furthermore, distinct from the effects on violence of rapid social change and anomie, Durkheim argued that during periods of acute political crisis, interpersonal violence will increase due to the threat to collective sentiments posed by the crisis.

[B]oth theory and empirical evidence suggest it is unwise to ignore [political characteristics].

DiCristina (2004, 81) noted that the “sentiments about collective things” hypothesis, although crucial to Durkheim’s work on homicide, has only rarely been acknowledged in the theoretical literature and has gone unaddressed empirically.² We believe the collapse of the Soviet Union and the subsequent transition toward democracy provided a unique opportunity to test this hypothesis, and our study was an attempt to begin to fill the void, especially since Durkheim himself argued that the strength of these sentiments are a key determinant of societal homicide rates (see chap. 10 in Professional Ethics and Civic Morals [1900/1957], titled “Duties in General, Independent of Any Social Grouping—Homicide,” and Suicide [1897/1979, 356-58, 368-69]).

Theory

Durkheim’s (1897/1979, 356-57; 1900/1957) two main theses about homicide were that homicide rates would decrease as (1) the “religion of humanity” became stronger over time as the result of gradual societal development and the shift to social bonds based on “complementary differences” or “exchange relations” (see DiCristina 2004, 85) and (2) collective sentiments weakened and became fewer in number over time (Durkheim 1897/1979, 356-57; 1900/1957). Although these ideas were based on the level of development in a society and assumed relatively slow evolutionary change, Durkheim also argued that the rate of change was important. Contrary to slow steady development, according to Durkheim, rapid change would lead to an anomic division of labor, social deregulation (Durkheim 1893/1984), and a greater number of anomic homicides (e.g., see Durkheim 1897/1979, 356-57), while threats to the collective would lead to a greater number of altruistic homicides (Durkheim 1897/1979, 357-58, 368-69). Durkheim (1897/1979) outlined four basic types of suicide: egoistic, altruistic, anomic, and fatalistic. He stated that extreme individuation (which he believed leads to egoistic suicide) did not cause homicide, and he did not discuss fatalism in relation to homicide, thus leaving anomic and altruistic homicides. In general, the former can be expected to occur usually as a result of social conditions accompanying increased egoism or individuation in the absence of well-defined norms and values, while the latter are expected to be associated with strong sentiments about the collective and its protection.

The hypotheses related to development and to anomie have received considerable attention in the criminological literature, including the empirical literature on social structure and homicide, but the hypotheses related to sentiments about collective things, and threats to them, remain largely unaddressed. This absence represents a key gap in the literature. For example, although Durkheim believed that the development of the religion of humanity would help lead to a reduction in homicide rates over time, he argued that a long-term decrease in homicide would be due less to this respect for individual dignity and more to the weakening of collective sentiments (Durkheim 1900/1957, chap. 10). Specifically, he stated that “if homicide is diminishing, it is rather that the mystic cult of the State is losing ground than that the cult of the human being is gaining” (Durkheim 1900/1957, 115). In one of the few tests of this hypothesis, Karstedt (2001) found that nations that exhibited stronger collectivistic relative to individualistic norms had significantly higher homicide rates. Durkheim’s views on this issue likely came from his stress more generally throughout his work (see especially Division of Labor in Society [1893/1994]) on the importance of the collective conscience. Subtle erosion of sentiments about collective things over long periods of time may not represent a concrete challenge to a society, especially if a new form of solidarity and new types of social bonds arise, but an acute and immediate threat to the solidarity based on these sentiments might well present a crisis, especially if the threat is against a collective belief that is emotionally charged. Such “deeply written” or “cherished” (also translated as “deeply engraven”) beliefs can include those related to religion or other “sacred” collective objects or social symbols such as culture, the state, the family, and work.

Durkheim (1893/1984) explicitly afforded things such as tradition and the state a religious character and discussed how he expected the religiosity associated with them to decline as the division of labor increased. Such cherished collective sentiments were abundant in Russia, where we find (1) ancient cultural traditions, including those that privileged the group over the individual (see Kharkhordin 1999); and (2) three-quarters of a century of top-down socialism rife with ideological symbols and icons reminiscent of religion. The omnipresent Soviet state was not only a manifestation of the latter, but was also citizens’ sole provider of nearly all the tangible things required for well-being, including health care, education, employment, and goods and services.

Durkheim (1900/1957) was clear in his belief that sentiments inspired by the collective (including the “cult of the State”) are “stimulants to murder” (p. 115), which “is why political beliefs . . . often in themselves carry the seeds of homicide” (p. 116). Mentioning strong sentiments about politics, religion, and family together in the same paragraph in Suicide, Durkheim (1897/1979, 356) argued that “where family spirit has retained its ancient strength, offences against the family are regarded as sacrilege . . . [and w]here religious faith is very intense it often inspires murders and this is also true of political faith.” Thus during times of political crisis (Durkheim 1897/1979, 353)—and during times of rapid political change that present an acute threat to the state and to long-standing traditions, and thus to collective sentiments—we should expect an increase in homicide rates. It follows that where political change and threats to collective sentiments are generally present, those areas experiencing greater change in political faith should be those areas where increases in homicide rates are greater.

It is important to note that while Durkheim spoke of threats to collective things, the type of violence he expected was not group-related or organized violence to protect the collective (i.e., calculated murders directed at a specific threat). On the contrary, he clearly believed that “the evolution of the trend of homicide . . . is better brought out by the curve of unpremeditated murder” (Durkheim 1897/1979, 349). This is probably because he more generally restricted his ideas about homicide to unpremeditated murders since to him “homicide is inseparable from passion” (Durkheim 1897/1979, 356). Thus, while our data did not allow us to discern between premeditated and unpremeditated homicide, it was more appropriate to test this aspect of his theory with data on interpersonal violence, as we did, and not group-related or organized crime and violence.

In this study, we tested the short-term aspect of Durkheim’s hypothesis related to threats against collective sentiments. The large-scale natural experiment of the sudden collapse of the Soviet Union provided a rare opportunity to test this hypothesis, and the 1990s Russian political landscape presented unique measurement possibilities in relation to political change (we elaborate on both, especially the latter, in the Discussion section below). We argue that swift political change in Russia represented a political crisis. That is, the transition toward democratization and marketization—which stress individual freedoms, goals, rights, and responsibilities— represented a clear departure from and threat to (1) the Soviet state and all it represented and (2) deeply seated Russian cultural traditions and a lifestyle that privileged the group over the individual long before the Soviets arrived (Kharkhordin 1999).³ Although these changes were occurring generally in Russia, their pace varied substantially throughout the vast nation. Thus, specifically, we tested the hypothesis that regions facing greater political change would be regions that experienced the greatest increases in homicide rates.

Data and Method

The unit of analysis in this study was the Russian region. There are eighty-nine of these regions, which are administrative units roughly equivalent to states or provinces. Data from the contiguous Ingush and Chechen Republics were considered unreliable for several reasons, including the ongoing war in the area, so they were excluded from this analysis. Data from nine of the smaller regions (each called an autonomous okrug, or district) were covered by the larger regions in which they are embedded. Note that data were not lost or missing from these regions but instead were simply included in the data from the larger region of which they are a part. This left seventy-eight cases for analysis (i.e., the entire sample of regions minus the Chechen and Ingush Republics).

The use of regions as units of analysis is important. Nearly all comparative political and social research within Russia employs these regions. In their study of electoral competition in Russia, for example, Moraski and Reisinger (2003, 280) stated that “Russia’s regions are important polities with much influence over their residents’ lives.” The authors went on to state that “regional politics may well promote progress toward democracy for Russia as a whole or serve as a brake on such progress” (p. 280), which is a key theoretical issue at stake in this study. Finally, an important methodological issue is variation, and Moraski and Reisinger showed “that the regions vary significantly in electoral democratization . . . unlike during the Soviet era, when political institutions, policies, and mass behavior were kept relatively uniform” (p. 281).

The measure of violence in this study was the change in regional homicide victimization rates between 1991 (the final year of the Soviet Union’s existence) and 2000.⁴ We used the residual change score instead of the raw change score because it reflects the amount of change in a region’s homicide rate unexplained by its initial levels (Bohrnstedt 1969). Thus, in this study, ΔHomicide2000 = Homicide2000 - (α+β× Homicide1991). Furthermore, because all of the regions were used to estimate the regression from which the residuals were drawn, the residual change scores reflect the developments of the entire ecological system under study (Morenoff and Sampson 1997). In other words, the score on the dependent variable for a specific region not only controls for that region’s homicide rate in 1991 but also takes into account the changes in homicide rates between 1991 and 2000 in all of the other regions. This is important since we know from prior research that for the past several decades the spatial patterning of homicide in Russia has shown consistently lower rates in the Northern Caucasus and higher rates east of the Ural Mountains (Pridemore 2003a; Shelley 1980; Shkolnikov 1987; see also Stickley and Mäkinen 2005). Pridemore (2003b) described and compared Russian homicide estimates provided by the vital statistics and police reporting systems, concluding that the former provided significantly more reliable estimates of the overall number of homicides than the latter. We therefore used the age-standardized death rate (per one hundred thousand residents) due to homicide as our measure of the main dependent variable. The data employed here were prepared for the first author from vital statistics data from the Russian State Committee on Statistics.⁵

We used “political change” as a proxy for Durkheim’s more abstract collective sentiments. Notice here a major assumption: we did not have a measure of collective sentiments, and thus we assumed rapid political change creates threats to these sentiments, which in turn heighten passions and lead to higher homicide rates. While we recognize that the statement that political change threatens sentiments about the collective is a proposition to be tested in itself, assumptions such as this are common in nearly every study of social structure and homicide, and we wish to make ours explicit here. Below, we present arguments about why we believe this to be a logical assumption, especially in Russia.

In our measurement of this concept, we began by recognizing Communist Party hegemony during the Soviet era, which resulted in essentially 100 percent of the vote being cast for Party candidates. We then obtained measures of the proportion of all voters who cast their ballots for Communist Party of the Russian Federation (CPRF) or Liberal Democratic Party of Russia (LDPR) candidates in the 1999 Duma and 2000 presidential elections.⁶ As detailed in the Discussion section below, these were the two main opposition parties in Russia that called for a return to something resembling Communist or Soviet rule. An index was created that summed the z-scores for these four measures (i.e., CPRF vote in 1999 and 2000 and LDPR vote in 1999 and 2000).⁷ Thus, the smaller a region’s score on this index, the smaller its vote for these opposition parties, and therefore the greater the amount of political change in the region since the collapse of the Soviet Union (because Communist support was assumed to be nearly universal during the Soviet era, at least in terms of voting). This measure was reverse-coded so it could be interpreted in terms of the main concept under examination here, political change (i.e., dissipation of Communist Party electoral strength). We note that a unique set of political circumstances existed during the 1990s (until Vladimir Putin was elected president in 2000) that made this measure a strong indicator of political change. Specific political shifts since that time, however, mean that similar measures from subsequent elections probably would not tap the same sentiments we measured here. We elaborate on these unique circumstances (and their disappearance following the 2000 election) in the Discussion section below.

Several controls were employed in an attempt to isolate the effects of political change on changes in homicide rates. These controls come from prior structural-level research on homicide in Russia, the more general literature on the structural covariates of homicide, and from what we know about those most likely to have voted for the Communist Party in the Russian parliamentary and presidential elections. First, recent research testing another Durkheimian theory about rapid social change, social deregulation, anomie, and homicide showed that Russian regions facing the strongest negative effects of socioeconomic change during the 1990s were regions where homicide rates increased the most during this time (Kim and Pridemore 2005). Thus, we included as a control Kim and Pridemore’s (2005) negative socioeconomic change index, which gauged regional changes over the 1990s in population, privatization, unemployment, poverty, and foreign capital investment (in their creation of the index, the authors coded these measures in such a way that privatization and foreign capital investment were positive).

Second, in the one cross-sectional model included in this analysis, we included the proportion of the regional population living below the poverty line in 1999 as a measure of poverty. This was done not only because area measures of poverty are the most consistent covariate of homicide rates in the social structure and homicide literature (see Messner and Rosenfeld 1999; Pridemore 2002b) but also because the Communist Party was popular among the poor during the election cycle under examination (March 2001). Similarly, measures for educational level of the population and the proportion of the population living in urban areas were included as controls because they were likely confounded with CPRF voting. That is, not only might these variables have an influence on homicide rates themselves, but those with a lower education and those living in rural areas were more likely to vote for the Communist Party and “against reform and reformers” (Slider, Gimpelson, and Chugrov 1994, 718; March 2001; also see chapters in Hesli and Reisinger [2003] and in Wyman, White, and Oates [1998]).⁸ It was also important to control for the proportion living in urban areas because Durkheim (1897/1979, 353) argued that homicide rates in rural areas were higher due to stronger sentiments about the collective in rural areas and greater respect for the body and the individual in urban areas (see Durkheim [1897/1979, 293], where he argued that “gross and violent” acts conflict “with the gentleness of urban manners and the regard of the cultivated classes for the human body”).⁹ Education was measured as the number of people enrolled in college per one thousand residents, and urbanization as the proportion of the regional population living in cities with more than one hundred thousand residents (data for all three of these measures were obtained from Goskomstat [2001]).

Andrienko (2001) and Pridemore (2002a) have shown alcohol consumption to be associated with the growth of crime and violence in Russia and with the spatial distribution of regional homicide rates in the country, respectively. Thus, we used Pridemore’s (2002a) proxy for heavy drinking, which was the regional age-standardized rate (per one hundred thousand) of deaths due to alcohol poisoning (examples of and reasons for using this proxy in Russia are explained elsewhere: Chenet et al. [2001]; Shkolnikov, McKee, and Leon [2001]; Shkolnikov and Meslé [1996]). Finally, prior research has shown that homicide rates in the regions east of the Ural Mountains are currently (Pridemore 2003b) and have been historically (Shelley 1980; Shkolnikov 1987) significantly higher than in the rest of the nation. Therefore, we included a dummy variable for those regions located in this area.

We note that the measures of alcohol consumption, education, and urbanization did not represent change scores between 1991 and 2000, although the dependent variable did. While this presents a problem with model estimation, we did not have access to all data for all variables in the final years before the breakup of the Soviet Union. We recognize this as a limitation but thought it better to employ these controls using available data rather than ignoring them completely, especially because some may be confounded with voting for the CPRF and LDPR. We discuss our attempts to minimize this limitation in the discussion of model sensitivity below.

[W]hat disappeared with the Soviet Union was not simply an abstract ideology or government, but a provider and an icon, a way of life fused with the longer history and traditions of Russian culture.

Finally, the Chukot and Jewish Autonomous Okrugs were missing observations on 1991 homicide rates, so 1992 rates were substituted when creating the residual change scores. Two of the variables—education and percentage urban—had skew statistics greater than twice their standard errors, and thus we logarithmically transformed their values to more closely approximate a normal distribution. Four of the regions had no cities with more than one hundred thousand people, so 1 was added to each score before taking the log since the natural logarithm of 0 is undefined. With one exception (discussed below), all models were estimated using ordinary least squares regression. Common exploratory data analysis techniques were employed and regression diagnostics and tests of model sensitivity were carried out.

Results

Brief definitions and descriptive statistics for each variable are shown in Table 1. Means and standard deviations in this table are for each variable before log transformations were carried out. The table shows that the mean regional homicide victimization rate in 2000 was about thirty per one hundred thousand residents. The table also shows that, on average, regional homicide rates increased by nearly fourteen per one hundred thousand between 1991 and 2000, which represented a greater than 80 percent increase. Looking at absolute changes, every region except one (Kursk Oblast, in which the homicide rate decreased by less than one per one hundred thousand residents) experienced an increase in its homicide rate during this period. The mean regional proportion of the vote going to the CPRF and LDPR (as a percentage of those who voted) in the presidential elections in 2000 was 32.6 percent, and in the parliamentary elections in 1999 was 32.1 percent (not shown in the table). Note that these are regional means, not the actual proportion of the vote for these parties in the country as a whole.¹⁰

TABLE 1.

BRIEF DEFINITIONS AND DESCRIPTIVE STATISTICS FOR DEPENDENT AND INDEPENDENT VARIABLES (N = 78)

Variable	Description	Mean	SD
ΔHomicide	1991 homicide victimization rate subtracted from 2000 homicide rate	13.6	8.6
Homicide 2000	2000 homicide victimization rate	30.1	17.5
Political change	Z-scored index of political change (see text for description)	0.0	2.5
SE change	Index of negative socioeconomic change (see text)	1.4	1.1
Alcohol	Proxy (see text): Number of deaths per one hundred thousand residents due to alcohol poisoning in 2000	28.7	17.5
Education	Number of students enrolled in higher education per one thousand residents in 2000	27.0	13.8
Urban	Proportion of population living in cities with more than one hundred thousand residents in 2000	39.0	16.5
Poverty	Proportion of population living below subsistence minimum in 1999	42.7	16.2
East	Dummy variable for those regions located east of the Ural Mountains	—	—

NOTE: The change score for homicide in this table is the raw change score. In model estimation, the change score was the residual when 2000 rates were regressed on 1991 rates (see text). SE = socioeconomic

Results of model estimation are shown in Table 2. Models 1 through 3 employed the residual change score as the dependent variable, while model 4 used the 2000 homicide rate as the dependent variable. Model 1 was a baseline model estimated without the inclusion of political change. As expected, the negative socioeconomic change index, the proxy for heavy drinking, and regional location east of the Urals were all positively and significantly associated with the change in homicide rates between 1991 and 2000. Model 2 was the same as model 1, but included the political change index. The results indicated support for Durkheim’s prediction. That is, holding constant the other variables in the model, regions that experienced greater political change (and thus, we assume, greater threats to sentiments about the collective) between 1991 and 2000 were regions where homicide rates increased the most. The inferences drawn for the other variables remained unchanged with the addition of the political change index, with the possible exception of education, which had a p-value of .065 in model 2.¹¹

TABLE 2.

RESULTS FOR ΔHOMICIDE RATES AND 2000 HOMICIDE RATES REGRESSED ON POLITICAL CHANGE AND CONTROL VARIABLES (N = 78)

	Model 1: DV = ΔHomicide			Model 2: DV = ΔHomicide			Model 3: DV = ΔHomicide			Model 4: DV = 2000 Homicide
Variable	b	SE	p	b	SE	p	b	SE	p	b	SE	p
Constant	8.25	6.33	.196	9.24	6.13	.137	8.53	1.38	<.001	13.39	11.78	.260
Political change	—	—	—	0.79	0.32	.017	0.67	0.37	.075	1.71	0.59	.005
SE change	2.13	0.80	.009	2.06	0.77	.009	2.07	0.84	.015	—	—	—
Alcohol	0.19	0.05	<.001	0.18	0.04	<.001	—	—	—	0.43	0.08	<.001
Log education	-3.17	2.11	.139	-3.87	2.07	.065	—	—	—	-3.58	3.77	.346
Log urban	1.52	1.16	.195	1.78	1.13	.119	—	—	—	1.23	2.05	.549
Poverty	—	—	—	—	—	—	—	—	—	0.11	0.09	.227
East	6.59	1.81	<.001	8.48	1.91	<.001	7.85	2.16	.001	24.58	3.36	<.001
Adjusted R2		.41			.45			.25			.56

NOTE: DV = dependence variable; SE = socioeconomic

Model sensitivity

Aside from the normal regression diagnostics routinely employed, further precautions were taken to test the sensitivity of these findings to the peculiarities of the model resulting from data limitations. First, since the controls for heavy drinking, education, and urban did not represent change scores, a reduced model was estimated in which they were excluded. The results are shown as model 3 in Table 2. The p-value in this model for the political change index was .075. This does not allow for strong conclusions, but given the sample size, the absence of potential confounders from the model, and the unidirectional hypothesis, this is in line with expectations. The inferences drawn for the other two variables in model 3 remained the same as in previous models. Second, technically the political change index was not a change score because it represented outcomes from the 1999 and 2000 elections (and thus has only measures from time 2 not time 1). The initial response to this is that it does reflect change since in the past the Communist Party exerted hegemonic control over the political realm and received virtually 100 percent of the vote. Nevertheless, to address this limitation, an alternative model was estimated in which the dependent variable was the regional homicide rate in 2000. In this model, the negative socioeconomic change index was dropped because it represented a change score, though the other controls from earlier models were included since they were measured in 2000. Furthermore, a measure of poverty was included since it is often found to be associated with area homicide rates and since it is likely confounded with the vote for the opposition parties. The findings are presented as model 4 in Table 2. The results revealed that the effect size for the political change index in this model was substantially higher than in previous models and that the inferences drawn remained the same as in previous models.¹²

Discussion

Several prior structural-level studies have employed voter turnout to represent various concepts potentially related to crime and violence, including civic engagement (Chamlin and Cochran 1995; Rosenfeld, Messner, and Baumer 2001), political disaffection (Callahan 1998), and level of conformity with norms (Coleman 2002). For whom an electorate votes, however, may be an even more important indicator of larger social forces than voter turnout, especially in the midst of a political transition. In the case of Russia, the shift was from a totalitarian past toward democracy and a free market. The former state had not only provided cradle to grave benefits for its citizens, some of whom were dependent upon this largesse and thus more vulnerable to the forces of economic transition, but also instilled (or at least forced) among many an almost religious deference. Thus, what disappeared with the Soviet Union was not simply an abstract ideology or government, but a provider and an icon, a way of life fused with the longer history and traditions of Russian culture.

In his review and critique of the contemporary theoretical and empirical literature on Durkheim’s theories of homicide, DiCristina (2004, 81) stated clearly that “no empirical study of Durkheim’s theory has directly tested his proposition concerning the effect of collective sentiments on homicide. In fact . . . few descriptions of Durkheim’s theory even acknowledge this core proposition. This is a serious oversight.”¹³ In this context, it appears as if DiCristina was speaking mainly of the expected decrease in homicide over the long term as a result of the decline in the number and strength of such sentiments. In our study, we tested for the first time a related version of Durkheim’s hypothesis that DiCristina (pp. 69-70) also discussed: political crisis (in our case, rapid political change) represents a threat to collective things, which is expected to incite passions and thus increase homicide (see Durkheim 1897/1979, 352-55).

Taking advantage of the unique research and measurement opportunities presented by the large-scale natural experiment of the dissolution of the Soviet Union and the subsequent democratization and marketization in Russia, we found support for this hypothesis. Specifically, regions that experienced the greatest amount of political change during the 1990s were regions where homicide rates increased the most, even after controlling for the effects of rapid socioeconomic change. The results of prior research (Kim and Pridemore 2005) showing an association between socioeconomic change (i.e., the “anomie” hypothesis) and the change in regional homicide rates in Russia are strengthened by our study, given their replication in a model that includes a control for the “collective things” hypothesis. Furthermore, while we do not take a strong stance on the issue, and while political and socioeconomic change are not mutually exclusive, in Durkheimian terms the homicides due to the deregulation resulting from rapid socioeconomic change would be labeled anomic and those resulting from political change or crisis would be labeled altruistic since the latter result from threats to sentiments about the collective (see chap. 5 in book 2 of Durkheim’s Suicide [1897/1979] as well as pp. 357-58 and 368-69).

[W]hile there may be alternatives to Durkheim’s hypothesis when explaining our results, an idealized vision of Soviet communism and its vestiges is not one of them.

Although they did not directly test Durkheim’s hypothesis, other scholars recently have found similar results. For example, Villarreal (2002) studied the connection between political characteristics and homicide in Mexico, another nation experiencing an uneven transition to democracy. He found a positive association between electoral competition at the local level and homicide rates. The author concluded that in societies that possess strong patronage networks (which Russia certainly did, given the history and entrenchment of the Communist Party), political changes that threaten the former political order result in increased rates of violence. Though Villarreal did not interpret his findings in Durkheimian terms, his results were consistent with those presented here, and in places his wording closely parallels Durkheim’s. In another study, Kondrichin’s (2000) research on Belarus, reported in a Russian-language journal, found an inverse correlation between a region’s vote for a Communist candidate and its homicide rate. The results of these studies, as well as our own, suggest that where there is less (rapid) movement away from the past there is less of a political crisis and less threat to sentiments about collective things (DiCristina 2004); and where there is less threat to these sentiments, homicide rates are lower. All of this coincides with Karstedt’s (2001) findings of higher homicide rates in nations with stronger collectivistic values.

The radical political change experienced by Russian citizens cannot be overstated. Socialist ideology was infused into every aspect of daily life, and the Communist Party was deeply entrenched not just nationally but at the local level and in multiple aspects of daily life such as education and employment. Furthermore, by design Soviet institutions created narrow and homogenized interests and desires, not to mention dependence on the state. Yet while vestiges of state socialism and the Communist Party certainly remain and in some aspects are still quite strong, the political break from the past could have hardly been swifter or sharper. The “shock therapy” that instituted market reforms was radically at odds with the former centrally planned economy in its most fundamental assumptions. The same is true for group protection and the dependence engendered by the Soviets in their relation to the new individual rights, responsibilities, and diversity of democracy. These profound changes, together with the dramatic and publicly visible events of the time, by their very nature represented a direct assault on and break from the past (Bunce 1998, 192-93, 200-201). Slow evolutionary change allows strong sentiments to dissipate, erode, and be replaced, thereby leaving the original sentiments (and the “things” themselves) to fade into the past. On the other hand, rapid political change occurs in a context where the collective things are still present and the sentiments about them still strong, thus perhaps engendering interpersonal violence via the mechanisms described by Durkheim.

One alternative explanation for our findings is that the conditions in regions with a greater vote for the Communist Party are similar in structure and institutional strength (as opposed to sentiments) to the communist past and thus exert more social control, thereby resulting in smaller increases in homicides. While we do not have a measure of social control in our analysis, and while several of the regions with higher vote totals for the Communist Party in the presidential and parliamentary elections also had governors who were members of or sensitive to the Communist Party, it is important to understand the limits of this explanation. First, the belief that rates of violent crime were significantly lower in Soviet Russia than in Western nations was propaganda and patently false. Pridemore (2003a) has employed newly available mortality data to track Russian homicide victimization rates over time and found that they were as high as or higher than in the United States for at least the past several decades. Karstedt’s (2001) findings about collectivistic value structures and higher violence rates mentioned earlier are also relevant here. Second, the high level of corruption and low morale among the police in post-Soviet Russia are not due solely to present conditions, but are rooted in the (police) culture of the Soviet era (Beck and Robertson 2005). Similarly, regions governed by CPRF politicians likely possess fewer reasons (and demand) for police transparency and responsibility than more democratic regions. Furthermore, even where Communist Party candidates received a large proportion of the vote or won local mayoral or gubernatorial elections, the regions were not truly communist in structure and institutions, at least as thought of during the Soviet era. Finally, several studies have shown a strong link in Russian regions “between democratic success and success in solving key economic and social problems” (Moraski and Reisinger 2003, 297). Those regions that resist democratic and economic reforms and maintain high levels of firm-level barter, subsidies, and social spending have been unable to attract business and thus build a strong tax base to support their policies in the long run (Konitzer-Smirnov 2003), which has led to a reduction in standards of living and an increase in dependency in these regions. All this suggests that while there may be alternatives to Durkheim’s hypothesis when explaining our results, an idealized vision of Soviet communism and its vestiges is not one of them.

Given the topic of this special issue of the journal, we should note that the results of our study are more about the effects on changing homicide rates of rapid political change generally, not necessarily democratization specifically. In discussing democracy, transition, and crime, for example, Karstedt (2003) made a similar point: that it is the failure to democratize rather than democratization that is responsible for rising crime rates in several post-Soviet countries. She noted specifically the failure of these nations to develop a functioning civil society, along with other deficiencies such as continuing distrust of the legal system, as a cause of crime. Similarly Karstedt (1999) focused on the crisis of deregulation (i.e., the lack of regulation, or at least the dissolution of the former mode of regulation) in her discussion of social transformation and crime as opposed to how specific characteristics of democratization and democracy lead to higher crime rates. And as mentioned earlier, Karstedt (2001) found that nations in which respondents scored higher on the individualistic and egalitarian values associated with democracy had lower rates of violence than nations in which respondents scored higher on collectivistic values. Thus, while it is true in the case of Russia that the initial change was from Soviet socialism and a centrally planned economy toward democracy and a free market, our analysis does not allow us to conclude that there is something special about the transition to democracy that is important, but rather it is the rapid change that appears to be the operative mechanism. Future comparative research may find that the conditions accompanying the Russian transition are in fact common to democratizing nations, and thus there truly may be something special about becoming a democracy that in the short term leads to increased levels of crime. On the other hand, other paths to democracy may not result in increased rates of crime; or we may find that Russia is not truly becoming a democracy and that what led to increased crime rates was state failure, social and legal deregulation, and the criminogenic conditions engendered by rapid change in the political economy.

Unique natural experiment and measurement opportunities

Certain specific and unique elements made Russia an appropriate case study to test this previously unexamined Durkheimian theory, and a fleeting political context made this study feasible and allowed us to create a valid measure of political change. First, some of the foundational theories of sociology and structural-level criminology were created in an attempt to understand the causes and effects of social change. However, scholars are usually stuck with slow-moving cases as examples or, even further from theoretical tenets but often necessary, using cross-sectional studies to compare cases possessing a range of scores on a particular variable and assuming these differences represent linear development. In Russia, we have a natural experiment in socioeconomic and political change associated with democratization and marketization that has had a profound influence on Russian society and social institutions, thereby presenting scholars with a large-scale case study of the effects of rapid social change on a society.

Second, a unique set of circumstances made our measure of political change appropriate for the period under study, but following the 2000 presidential election, changes in the political landscape made this specific indicator suspect. During the 1990s, the Communist Party (and to a lesser extent the LDPR) was clearly the place to turn not only for ideological communists but for those dissatisfied with the transition. The Party was most popular among people (e.g., the poor, uneducated, and pensioners) and in places (e.g., rural areas and regions with failing farms and unproductive heavy industry) left behind by the sweeping changes (March 2001). Further distinguishing the Communist Party as the antichange party (and party of the past) was Yeltsin’s virulent anticommunist position, as well as the Party’s own strong stance of anti-Westernism.

The Russian homicide rate more than tripled between 1988 and 1994.

The CPRF took advantage of the popular dissatisfaction with marketization by stressing economic issues in its campaigns and running on a strongly patriotic and nationalist platform, for example, calling for the state to “return to citizens their guaranteed socioeconomic rights” and to again become “the main guarantor of social justice” (as quoted in Cook 2005, 51). Similarly, the Party continued to call for the protection of the “culture, language, beliefs, and customs of all Russian people” while declaring itself the refuge of “patriots” who had experienced “pain” and “humiliation” as a result of the Soviet downfall (Kommunisticheskaya Partiya Rossiskoi Federatsii 2005). Throughout the elections of the 1990s and until 2000, the Communist Party created broad coalitions of socialist and nationalist opposition organizations (McFaul 1997). The Party was the main victor in the 1995 Duma election (garnering about one-third of all seats); its presidential candidate, Gennady Zyuganov, received 32 percent (to Yeltsin’s 35 percent) in the first round of the 1996 presidential election, which Zyuganov lost in the runoff (but still received 40 percent of the vote); and the Communist Party again won the largest faction in the 1999 Duma elections, securing 50 percent more seats than its closest rivals (114 compared to 73 for Unity and 67 for Fatherland-All Russia).

Following the 1999/2000 election cycle, however, several developments converged to decrease the popularity of the Communist Party and to make a measure such as ours a much weaker indicator of political change. Even in the 1999 Duma election, the CPRF ranked at the bottom in attracting new voters, those who had previously voted for other parties, and young Russians (Colton and McFaul 2003; March 2001), yet the Party ran on the same platform in 2003. Furthermore, before the 2003 Duma election President Putin increased electoral competition on the Left by working with communist splinter parties that were pro-Kremlin, and he also orchestrated a widespread media campaign aimed at discrediting the Communist Party. All this resulted in a sharp drop in the number of Duma seats won by Party candidates. President Putin enjoyed enormous popularity during his first term, during which he seized control of the media, continued to build a superpresidential system (see Sakwa 2005a), and co-opted elites across the political system (as well as symbols and ideas—such as patriotism, selfreliance, and Soviet references—that had previously “belonged” to the CPRF).¹⁴ The perennial Communist Party presidential candidate, Gennady Zyuganov, withdrew from the 2003 presidential election because of Putin’s insurmountable lead, and the replacement candidate received only 14 percent of the vote. In the 2003/2004 election cycle, the Communist Party not only failed to gain new voters, but also lost votes from specific regions and demographic groups that had previously provided solid support. Recent surveys also suggest that even citizens who maintain hope for a return to a Soviet-style economy believe Russia should be friendly toward the West, which is in contrast to the Party’s distinctly anti-Western platform. Thus, while the CPRF still has a voice, it is muted, and the Party no longer fills the distinctive ideological niche it did during the 1990s.

Conclusion

The Russian homicide rate more than tripled between 1988 and 1994. Although the rate decreased somewhat from its 1994 peak and stabilized, in 2000 it was still twice as high as it had been a decade earlier. Prior research has established compelling evidence for Durkheim’s anomie/social deregulation hypothesis as an explanation for increases in both national-level (Chamlin, Pridemore, and Cochran 2005) and regional-level (Kim and Pridemore 2005) homicide rates. This study examined another important, though rarely tested, aspect of Durkheim’s ideas about change, finding support for the hypothesis that rapid political change (or political crisis) has separate and distinct effects on homicide rates.

Durkheim’s overarching thesis concerning societal development and solidarity should lead us to expect a decrease in violence in the long run because of the strengthening of the religion of humanity, or moral individualism. Similarly, during gradual societal development, the weakening of the power of the collective, and thus of the strength of passions and collective sentiments, should lead to fewer homicides. According to Durkheim (1897/1979, 353), however, short-term political crises or threats to the collective (more generally, see pp. 352-55) will incite passions that result in increased violence. Soviet Russia was hardly an undeveloped society representative of Durkheim’s mechanical solidarity. To the contrary, it was technologically advanced, possessed a developed (if forced) division of labor, and exhibited certain values that highlighted individuality (Karstedt 2003; Kharkhordin 1999). Nevertheless, it is fair to argue that long-standing Russian cultural traditions included strong bonds based on collective sentiments, and the Soviet era resulted in ideologically rooted and exaggerated sentiments about the collective. Such cherished beliefs, according to Durkheim, are bound to elicit heightened passions when threatened.

We have provided initial support for this hypothesis here, and the results were made even stronger by controlling for the negative effects of rapid socioeconomic change, thereby distinguishing two different mechanisms through which different aspects of change (socioeconomic and political) can influence rates of interpersonal violence. So while we would expect democratization, political competition, and the development of civil society to be healthy (especially relative to the Soviet juggernaut of the past) and, in the long run, to result in lower rates of violence, the process is not without considerable accompanying growing pains. In Russia, at least, it appears that the threats to sentiments about the collective resulting from swift political change was partially responsible for the increase in the Russian homicide rate, which is now among the highest in the world.