Tobacco consumption and national rates of personal violence (suicide and homicide)

There is evidence at the individual level that smoking is associated with depression (1), and that nicotine dependence is associated with anhedonia (2). Lynn (3) explored the meaning of tobacco consumption at the aggregate level in a sample of 18 industrialized nations. He found that cigarette consumption was positively associated with a number of other social indicators (such as divorce rates and accidental death rates) and was more strongly associated with murder rates (for which the association was positive) than with suicide rates (for which the association was negative). Thus, it is of interest to explore the association of tobacco consumption at the aggregate level using time-series data (rather than cross-sectional data as used by Lynn), since time-series studies often give results di¡erent from cross-sectional studies. Research on individuals suggests the hypothesis that tobacco consumption should be positively associated with suicide rates, while the aggregate level research by Lynn suggests that tobacco consumption should be positively associated with homicide rates. Data on tobacco consumption to test these predictions is available for Norway for the period 1973 ^ 1995, as reported by Kraft and Svendsen (4), who noted that the percentage of daily smokers in Norway from 1973 ^ 1995 declined in males but not in females. To explore the association of tobacco consumption and rates of personal violence (suicide and homicide) over time in Norway, the data used by Kraft and Svendsen were obtained from the National Council on Tobacco and Health, with the aide of Rita Lill Lindbak (Senior Executive O¤cer), who provided the percentages of men and women smoking for the years 1973 to 1993. Suicide and homicide rates by gender for this period were available from the World Health Organization (5). The association of tobacco consumption with suicide and homicide rates was examined using a statistical package from Doan (6), with the Cochrane ^ Orcutt technique to correct for the serial autocorrelation in the data set. The simple Pearson correlation coe¤cients indicated that the percentage of smokers was negatively associated with male suicide and homicide rates (Pearson r~ ^ 0.84 and ^ 0.75, respectively, twotailed pv0.001) but not with female suicide and homicide rates (r~0.39 and 0.10, respectively). In the multiple regressions including divorce, marriage and birth rates as predictor variables, the percentage of smokers was positively associated with the male suicide rate (b coe¤cient~0.30, pv0.10) and negatively associated with the male homicide rate (b coe¤cient~ ^ 0.08, pv0.05). The b coe¤cients for female suicide and homicide rates were not statistically signi¢cant (0.07 and 0.02, respectively). The association of tobacco consumption with timeseries suicide rates (positive in the regression equations) are consistent with the predictions made from the individual-level research and con£ict with those reported by Lynn (3) in cross-sectional aggregate studies. The present results suggest the possibility that national levels of tobacco consumption over time might be an indicator of national levels of depression.