Punishment Severity and Recidivism – A Study on Driving Under Influence

By Karoliina Suonpää

Photo of a car driving through a rural Scandinavian landscape
Photo by Caleb Whiting on Unsplash

One of the most important questions of research on criminology and criminal policy pertains to the impact of sanctions: are harsher sanctions more effective compared to the lighter options, or do they increase recidivism instead?

Studying the impact of the punishment severity is notoriously hard because of the selection bias:individuals are not assigned different sanctions at random, but the severity of the sanction is determined by the severity of the crime and prior convictions. Therefore, a straightforward comparison between individuals subjected to, for instance, a fine versus imprisonment, does not reveal the causal impact of the sanction.

Recently, researchers have used natural experiments for analyzing the causal impact of sanctions. In Finland, the sentencing principles of individuals convicted of driving under influence (DUI) provide a quasi-causal design for analyzing the impact of sanction severity on recidivism: The severity of the punishment for DUI results from the level of blood alcohol content; according to the Finnish penal code, the offenders with blood alcohol content at 0.5 per mill are convicted of DUI whereas offenders with blood alcohol content at least 1.2 per mill are convicted of aggravated DUI and therefore receive a harsher penalty. We leverage this sharp increase in the likelihood of receiving a (suspended) prison sentence, instead of a fine, and use the regression discontinuity design for teasing out the causal impact of harsher punishment on recidivism among offenders convicted of driving under influence.

By using Finnish administrative registers, we followed 35,000 individuals convicted of DUI or aggravated DUI and analyzed their recidivism after the conviction. Offenders convicted of DUI were typically sentenced to fine whereas aggravated DUI usually led to suspended prison sentence. In the next three years, a quarter of the offenders engaged in general recidivism, and almost a fifth committed a new DUI offense. On average, recidivism was more common among individuals convicted of aggravated DUI compared to individuals convicted of DUI. Yet, it is clear, that this disparity might reflect the underlying differences between the two groups.

When we applied the regression discontinuity design and compared the individuals just below and above the threshold of the blood alcohol content for aggravated DUI, the harsher penalty seemed to marginally decrease both DUI and general recidivism. However, the difference was not statistically significant. Therefore, the results suggest that having a harsher punishment does not have a meaningful impact on recidivism among offenders convicted of driving under influence.

However, this result also raises questions. While the sentence for aggravated DUI is, in principle, harsher, it is unclear whether the convicted perceive the sentence as more serious. When considering the financial consequences of the sanctions, it is the offenders convicted of DUI who face the most severe penalties. The interpretation of a suspended prison sentence to the convicted individual is a qualitative question and requires further research. In addition, it is possible that the impact of punishment severity is dependent on socioeconomic status, gender, or criminal history of the individual. These are the questions we aim to explore next.


Does Punishment Severity Matter for Recidivism and Social Integration? A Quasi-Experimental Study. Suonpää, Karoliina; Aaltonen, Mikko; Savolainen, Jukka; Tyni, Sasu & Aaltonen, Olli-Pekka. Unpublished manuscript.

Preregistration of the study: https://osf.io/byta6

About the author

Portrait of Karoliina Suonpää
Karoliina Suonpää. Photo: Juho Ruskoaho

Karoliina Suonpää is a postdoctoral researcher at the Institute of Criminology and Legal Policy, University of Helsinki. She is currently studying the consequences of criminal sanctions and social disadvantage.