The Short‐term Effects of Executions on Homicides: Deterrence, Displacement, or Both?*

Abstract

Does the death penalty save lives? In recent years, a new round of research has been using annual time‐series panel data from the 50 U.S. states for 25 or so years from the 1970s to the late 1990s that claims to find many lives saved through reductions in subsequent homicide rates after executions. This research, in turn, has produced a round of critiques, which concludes that these findings are not robust enough to model even small changes in specifications that yield dramatically different results. A principal reason for this sensitivity of the findings is that few state‐years exist (about 1 percent of all state‐years) in which six or more executions have occurred. To provide a different perspective, we focus on Texas, a state that has used the death penalty with sufficient frequency to make possible relatively stable estimates of the homicide response to executions. In addition, we narrow the observation intervals for recording executions and homicides from the annual calendar year to monthly intervals. Based on time‐series analyses and independent‐validation tests, our best‐fitting model shows that, from January 1994 through December 2005, evidence exists of modest, short‐term reductions in homicides in Texas in the first and fourth months that follow an execution—about 2.5 fewer homicides total. Another model suggests, however, that in addition to homicide reductions, some displacement of homicides may be possible from one month to another in the months after an execution, which reduces the total reduction in homicides after an execution to about .5 during a 12‐month period. Implications for additional research and the need for future analysis and replication are discussed.

Criminal Deterrence: A Review of the Literature

“…Time-series studies typically use vector autoregression to assess whether murder rates appear to decline in the immediate aftermath of an execution. Prominent examples include Stolzenberg and D'Alessio (2004), which finds no evidence of deterrence, and Land, Teske, and Zheng (2009), which finds evidence of short-run deterrence. Event studies such as those of Grogger (1991) and Hjalmarsson (2009a) examine the daily incidence of homicides before and after executions.…”

Envisioning Criminology

“…The nonautoregressive model lag-1-month deterrent estimate is −1.4 and that of the autoregressive model is approximately −1.96 homicides in the month after an execution. Both of these estimates are larger than the corresponding nonautoregressive model lag-1-month estimate (−1.3) and the autoregressive model lag-1-month estimate (−1.6) for all homicides combined of Land et al ( 2009 ). In addition, the cumulative deterrent effect of an execution on non-felony-level homicides estimated in the autoregressive model is 1.4 during a 12-month period.…”

“…We applied dynamic regression (DR)-linear transfer function (LTF) model building strategies with seasonal autoregressive integrated moving average (seasonal ARIMA) time-series models to these monthly data. Based on these time-series analyses and independent validation tests, the best-fi tting model in Land, Teske, and Zheng ( 2009 ) implied that, from January 1994 through December 2005, evidence existed of modest, short-term reductions in homicides in Texas in the fi rst and fourth months that follow an execution-about 2.5 fewer homicides in total. Another model suggested, however, that in addition to homicide reductions, some displacement of homicides may be possible from 1 month to another in the months after an execution, which reduces the total reduction in homicides after an execution to about .5 during a 12-month period.…”

Is life without parole an effective way to reduce violent crime? An empirical assessment

“…Second, executions and formal death sentences are statistically rare events that are not imposed in a time-stable manner, which means that even slight variations in measurement practices or model specifications can yield dramatically different results and outlier cases can have an undue influence on the outcome of statistical models (Berk, 2005;Cohen-Cole, Durlauf, Fagan, & Nagin, 2009;Donohue & Wolfers, 2005;Land, Teske, & Zheng, 2009;Nagin & Pepper, 2012;Shepherd, 2005). It also means that, because executions happen so infrequently, any deterrent effect that they have, no matter how large, might simply be too small detect (Katz, Levitt, & Shustorovich, 2003).…”

“…A third major problem for death penalty studies is that death sentences and executions are discrete events that occur at a single moment in history, but it is not clear how long their deterrent effects can be expected to last (Zimmerman, 2004), and the time period one chooses to investigate can have an enormous impact on the outcome of a study (Donohue & Wolfers, 2005;Land et al, 2009). In contrast, LWOP is a continuous event that lasts indefinitely, which means that its deterrent or incapacitating power-if it exists-should also be continuous, thus, eliminating this problem.…”