A comparative analysis of nations with low and high levels of violent crime

Abstract

The great majority of cross-national crime research has used quantitative analysis of large-n samples or qualitative case studies of one or two nations. Data quality problems make the results of the large-n studies problematic, while lack of generality limits the usefulness of case studies. This study included enough nations to allow results that have some generality of findings and yet few enough nations to allow for better data verification and greater in-depth analysis. Six low violent crime (LVC) nations were compared and contrasted with six high violent crime (HVC) nations. Analysis resulted in five categories of factors that effectively distinguish LVC from HVC nations: social integration versus disorganization, economic stress versus support, care versus abuse of children, official/approved violence, and the degrees of corruption and efficiency of the criminal justice system. Using structural and cultural contexts relevant to violent crime, these factors were integrated into a model of cross-national violent crime variation.

Introduction

The majority of cross-national crime research has been of the large-n type, using a quantitative methodology, usually multiple regression analysis (e.g., Avison and Loring 1986, Bennett 1991, Kick and LaFree 1985, Krahn, Hartnagel, and Gartrell 1986, LaFree and Kick 1986, Messner 1986, Messner 1989, Messner and Rosenfeld 1997, Neapolitan 1994, Neapolitan 1995, Neapolitan 1998, Ortega et al. 1992, Unnithan and Whitt 1992). In most studies, homicide rates have been the dependent variable, as homicides are considered to be less subject than other crimes to national variations in definition, reporting, and recording Ali 1986, Gartner 1995, Huang and Wellford 1989, Lynch 1995, Neapolitan 1997a, Wilkins 1980. To date, large-n, cross-national crime research has yielded results, which are often contradictory, in part due to currently unavoidable problems.

There are three problems making large-n, cross-national crime research difficult: the quality of the indicators of crime rates, the quality of the indicators of explanatory concepts, and the difficulty in considering the context of social facts and the full complexity of nations. Homicide rates are considered the most valid and reliable of cross-national crime indicators, but even they are extremely problematic, particularly for many developing nations.

The most frequently used source of cross-national homicide rates has been the International Crime Statistics, published by the International Criminal Police Organization (Interpol). There are two major problems with the use of these data. First, some nations include attempted murders in their reported rates while other nations do not, making comparisons invalid. Comparing samples of nations with and without attempts, Neapolitan (1996) found significant differences in the associations of relevant explanatory variables to homicide rates. Most past research has not adjusted Interpol rates for attempts, and such adjustments are not even possible for all nations, as some nations do not include the proportion that are attempts in their reports.

The second major problem was that many developing nations report crime and murder rates much lower than they actually are. These nations do not have adequate mechanisms for collecting and archiving data, are often unconcerned with accurate data collection, and have many offenses that are not reported to police as they are handled informally (Ali 1986, Gartner 1991, Gartner 1995, He and Marshall 1995, Marenin 1993, Souryal 1990, UNAFRI 1993, 1995). By all news and personal accounts, for example, Algeria and the Central African Republic have extremely high rates of violent crime, and yet they report very low rates to Interpol. On the other hand, developing nations sometimes report absurdly high rates by defining deaths from internal conflict as murders, such as the homicide rate of 12,500 reported by Rwanda in 1994 (Interpol, 1994).

Homicide rates included in the World Health Organization (WHO) World Health Statistics annuals are considered to be more accurate than Interpol data, as they are based on an actual count of deceased persons Kalish 1988, Neapolitan 1997a. Yet even WHO data have a definitional problem in that the “cause of death” categories include both “homicide” and “death by other violence,” and nations may vary in how they distinguish between these.

The major problem with WHO data, however, is that the great majority of nations reporting to WHO are developed or Latin American nations, with few being African, Arab, or Asian. Past research comparing Interpol and WHO homicide data has concluded that they are of sufficient quality for analytic and explanatory purposes in large-n data analysis Bennett and Lynch 1990, Huang 1993, Huang and Wellford 1989, Vigderhous 1978. These studies: (1) have only compared nations reporting rates to both sources and thus have not considered sample differences; and (2) were based on a relatively small number of nations, most of which are developed. In the Bennett and Lynch (1990) study—generally considered the best of its type—the number of nations in Interpol and WHO homicide comparisons was twenty-one, only six of which are developing nations. Neapolitan (1996) found that when homicide rate and sample differences are considered, the associations of relevant explanatory variables to WHO and Interpol homicide rates differ significantly.

Key explanatory variables are also of questionable quality. For example, the most consistent finding in cross-national research on homicides has been that of a positive association between income inequality and homicides Avison and Loring 1986, Braithwaite and Braithwaite 1980, Kick and LaFree 1985, Krahn, Hartnagel, and Gartrell 1986, LaFree and Kick 1986, MacDonald 1976, Messner 1982, Messner 1986, Messner 1989, Neapolitan 1994, Savage and Vila 1995. The World Bank (1995), the main source for income inequality data, noted there are numerous problems with these data—due to national differences in collection, measurement, and scope—making their use in comparisons across nations highly questionable. Other investigators of the quality of these data have also questioned their value in making cross-national comparisons Beisner 1996, Blackwood and Lynch 1994, Danziger and Smolensky 1975, Fields 1994, Jenkins 1991. Many indicators of other explanatory concepts have similar problems.¹

Finally, large-n studies reduce nations to scores on isolated variables, ignoring the complexity of nations and the historical and social contexts of the variables. It is increasingly recognized that crime research, and particularly cross-national crime research, must consider social facts as located in contexts, requiring differing levels of analysis and explanation Arthur and Marenin 1995, Cooney 1997, Gartner 1995, Groves and Newman 1989, Short 1998.

Case and comparative case studies avoid many of the problems of large-n studies by paying closer attention to data quality and by examining nations as complex, coherent wholes with consideration of context and situation (e.g., Adeyemi 1990, Birkbeck 1992, Ellis 1992, Hagan 1989, Helal and Coston 1991, Souryal 1990, Wikström 1992). These one- or two-nation studies, however, suffer from a lack of generality of findings and difficulty in testing theories.

Research is needed that focuses on a sufficient number of nations to make for some generality of findings and yet few enough for better data quality verification and in-depth analysis. Nations should be strategically selected to maximize the likelihood that results are accurate, informative, and general to other nations. Nations Not Obsessed with Crime (Adler, 1983)—which examined nine low-crime nations—is the best study of this type. By focusing exclusively on low-crime nations, however, it is not very informative on what distinguishes low and high crime nations.

This research focused exclusively on violent crime, since the same nations that are low or high in violent crime are not necessarily low or high in other crimes. Nations with low violent crime rates (LVC) were compared to those with high violent crime rates (HVC). The goal was to study a sufficient number of nations so as to have results with good generality, while also allowing for a good degree of in-depth analysis and verification of data quality and amounts of crime.