Miguel F. P. de Figueiredo, a professor at U.C. Berkeley and Yale Law School, conducted a study (here is the abstract) in Arkansas asking whether harsher sentencing for a higher blood alcohol content (BAC) affects the recidivism rate. A common practice today is to increase the time period of incarceration or the amount of time a license is suspended depending on the level of a defendant's BAC at the time of arrest in order to deter the defendant from committing the offense again. For example, in Tennessee if your blood alcohol level is over .20 the law requires a mandatory 7 days in jail for a standard first offence--5 days more than some whose blood level is over the legal threshold of .08. The focus of the study was to determine whether those increased sentences have the desired deterrence effect.
The study observed 15,973 defendants in two jurisdictions, making this study the most comprehensive of its kind. Because of the size, the numerical estimates and the causal inferences are more reliable. Specifically, the research design examined defendants whose BAC was just above or below .15 to see if the effect of an increased license suspension by an additional two months has an effect on recidivism.
Without getting into the math which involves multiple variables, the study found that the increased sentences had no real effect of the rate of recidivism. This finding directly contradicts the common theory that defendants will be deterred from committing the same offense if the Court imposes a harsher sentence.