Document Type

Working Paper

Publication Date



Both legal scholars and social scientists have leveraged new research evidence on the deterrent effects of the death penalty into calls for more executions that they claim will save lives and new rules to remove procedural roadblocks and hasten executions. However, the use of total intentional homicides to estimate deterrence is a recurring aggregation error in the death penalty debate in the U.S.: by studying whether executions and death sentences affect all homicides, these studies fail to identify a more plausible target of deterrence - namely, those homicides that are punishable by death. By broadening the targets of deterrence, these studies overestimate the number of murders that are averted by the threat of execution. We address this error by focusing on the subset of homicides that have been defined statutorily as capital-eligible to provide a more sensitive indicator of the deterrent effects of the death penalty. We use a public-use data archive based on police descriptions of homicides from 1976-2003 to construct rates of potentially death-eligible killings. We estimate that less than 25% of total criminal homicides are eligible for the capital sanction under the range of current state statutes. We find no changes over time in the rate of these capital-eligible homicides in death penalty states, despite fluctuations in capital punishment over time. Nor are there differences capital-eligible homicides between death penalty and non-death penalty states. We find similar flat trends in Texas, and also in Harris County, the county that supplies the most death cases in Texas. Using hierarchical regression models to fit growth curve trajectories over time and with a rich set of covariates that account for competing influences on homicide rates, we find no deterrent effects either from the presence of the death penalty or from variation over time in the dosage of any of its components in the states. Similar models for Texas counties produced identical results. The results show that none of the distinctive patterns one might expect from marginal death penalty deterrence can be found in the three decades since Gregg. Where the risk of execution goes up in a death penalty state, the death-eligible cases where that risk should make a difference decline no more than the non-death-eligible cases, nor is the proportion of all homicides that risk a capital sanction in death states any smaller in those states than it is in states without any death penalty. The rate of capital-eligible homicides is insensitive over time to variations in the incidence of executions or to the large swings from one decade to the next in the number or rate of non-death-eligible killings. Our search for death penalty deterrence where it should be a strong influence on homicide rates has produced consistent results: the marginal deterrent effect of the threat or example of execution on those cases at risk for such punishment is invisible.