Alcohol and violence: use of possible confounders in a time-series analysis.

AIMS: To assess the aggregate association between alcohol consumption and violence, while controlling for potential confounders. DESIGN AND MEASUREMENTS: The data comprise aggregate time-series for Norway in the period 1880-2003 and 1911-2003 on criminal violence rates and per capita alcohol consumption. Possible confounders comprise annual rates of unemployment, divorce, marriage, total fertility rate, gross national product, public assistance/social care and the proportion of the population aged between 15 and 25. Autoregressive integrated moving average (ARIMA) analyses were performed on differenced data. Both semilogarithmic and linear models were estimated.
FINDINGS: Alcohol consumption was associated significantly with violence, and an increase in alcohol consumption of 1 litre per year per inhabitant predicted a change of approximately 8% in the violence rate. The parameter estimate for the alcohol variable remained unaltered after including the covariates both in the semilogarithmic and the linear models. Of the seven covariates included in the models, only divorce was associated significantly with violence rate.
CONCLUSIONS: The results suggest that alcohol consumption has an independent effect on violence rates when other factors are controlled for. The results support the assumption of a causal effect of alcohol consumption on violence, and it appears that alcohol consumption is an important factor when we wish to explain changes in violence rates over time.