In social sciences, all studies have limitations and all make assumptions. To the extent that policy decisions are based on them, those decisions are based on knowledge that is less than perfect. What degree of confidence should we require of a study before we consider it in policy-making? Setting the threshold too high can result in an assumption that is even less likely to be true than the one we consider unworthy.
In a comment to Doug Berman's post on my study on the effect of plea bargaining effect on the net cost of capital punishment, commenter "Overburdened Statistics" says that he once thought that we should go with the best information we have available, but now he would apply a more selective filter. At least in the context where a decision has to be made, I think he was right the first time.
Legislatures in several states are being asked to make a decision of monumental importance, to repeal the death penalty. One of the primary arguments is that they will save large amounts of money in trial costs by avoiding the more expensive capital trials. The most widely publicized studies on cost do not include the plea bargaining effect in their calculations. Yet there is no way to avoid making some assumption or estimate of this effect. A study that simply ignores it, as the Urban Institute study of costs in Maryland does, implicitly assumes the effect is zero. Whatever the limitations of my estimate are (and I note some of them in the paper), can anyone seriously claim that zero is a better estimate?
In a comment to Doug Berman's post on my study on the effect of plea bargaining effect on the net cost of capital punishment, commenter "Overburdened Statistics" says that he once thought that we should go with the best information we have available, but now he would apply a more selective filter. At least in the context where a decision has to be made, I think he was right the first time.
Legislatures in several states are being asked to make a decision of monumental importance, to repeal the death penalty. One of the primary arguments is that they will save large amounts of money in trial costs by avoiding the more expensive capital trials. The most widely publicized studies on cost do not include the plea bargaining effect in their calculations. Yet there is no way to avoid making some assumption or estimate of this effect. A study that simply ignores it, as the Urban Institute study of costs in Maryland does, implicitly assumes the effect is zero. Whatever the limitations of my estimate are (and I note some of them in the paper), can anyone seriously claim that zero is a better estimate?
The Urban Institute study contains a similar assumption on deterrence,
but this time it is expressly stated. They briefly review some of the
papers on deterrence, noting the difficulty of making an estimate and
the spread among the estimates. Predictably leaning the most heavily on
the one paper that best suits their agenda, they say that because of
the difficulties they will not consider deterrence at all. That is an
assumption that the deterrence factor is zero.
A third, unstated, assumption in the Urban Institute study is that Maryland will never fix its overly extended review process. They count as a cost of the death penalty the additional cost of keeping a murderer on death row for his entire natural life span. That is not a cost of the death penalty; that is a cost of the obstruction of the death penalty. If Maryland could conclude its capital cases in five years, as Virginia typically does, the extra cost would be incurred only for five years, after which it would drop to zero, whereas the expense would continue in life imprisonment cases. The study notes that its assumption is "counterfactual." That is putting it mildly.
The best information we have available is that the death penalty has offsetting savings in trial costs through plea bargaining, that it saves a substantial number of lives (and, incidentally, the cost of trying those murders) through deterrence, and that mending not ending the death penalty will bring down the incarceration cost dramatically. If a decision must be made now, then it should be made with the best estimates, albeit less than perfect, that we have, not on unjustified and often unstated assumptions of zero effect.
A third, unstated, assumption in the Urban Institute study is that Maryland will never fix its overly extended review process. They count as a cost of the death penalty the additional cost of keeping a murderer on death row for his entire natural life span. That is not a cost of the death penalty; that is a cost of the obstruction of the death penalty. If Maryland could conclude its capital cases in five years, as Virginia typically does, the extra cost would be incurred only for five years, after which it would drop to zero, whereas the expense would continue in life imprisonment cases. The study notes that its assumption is "counterfactual." That is putting it mildly.
The best information we have available is that the death penalty has offsetting savings in trial costs through plea bargaining, that it saves a substantial number of lives (and, incidentally, the cost of trying those murders) through deterrence, and that mending not ending the death penalty will bring down the incarceration cost dramatically. If a decision must be made now, then it should be made with the best estimates, albeit less than perfect, that we have, not on unjustified and often unstated assumptions of zero effect.
Leave a comment