Abstract: | Studies of sentencing in jurisdictions with sentencing guidelines have generally failed to specify adequately the effects of offense seriousness and criminal history—the principal factors that, by law, should determine sentencing decisions. As a result, the explanatory power of those models is seriously limited, and regression coefficients representing both legal and extralegal factors may be biased. We present an alternative approach to specify more precisely the effects of legally relevant factors on sentencing outcomes and test the approach using felony sentencing data from Washington State. We find that controlling for the presumptive sentence substantially improves the fit and explanatory power of models predicting sentencing decisions, and that the estimated effects of extralegal factors, specifically sex and race, reduce considerably. The findings have both substantive and methodological implications. |