|
Earned Time and Recidivism CRIME VICTIMS UNITED |
On May 9, 2002, the Director of the Oregon Department of Corrections, Dr. Ben de Haan, made statements at a Senate Budget Committee hearing and again to the Salem Statesman Journal implying that DOC research showed that the availability of earned time causes lower recidivism.
After examining DOC's numbers, Crime Victims United determined that DOC did not have solid research but did have numbers showing a negative correlation between amount of earned time received and recidivism (the more earned time, the less recidivism). The leap from a correlation to a conclusion of cause-and-effect is unjustified and to characterize unscientific tabulations as research is a stretch.
While one DOC chart showed a negative correlation between amount of earned time received and recidivism, another chart presented at the same time showed a negative correlation between sentence length and recidivism (the longer the sentence, the lower the recidivism). DOC implied that cause-and-effect could be inferred from the first chart but not from the second. But neither chart demonstrates cause-and-effect; they both merely show correlations which can have a myriad of explanations.
On 6/14/2002, CVU policy analyst Howard Rodstein spoke with Dr. Paul Bellatty, the DOC analyst who produced the study. Dr. Bellatty acknowledged that the study was incapable of demonstrating cause and effect.
Crime Victims United sent a letter to Dr. de Haan pointing out the fallacy of DOC's claims in detail. We brought this error to Dr. de Haan's attention because we have seen how unfounded claims can turn into tenacious, widespread misconceptions that recyle indefinitely.
On 6/21/2002, CVU President Steve Doell and Howard Rodstein met with Dr. de Haan to explain our position that DOC's claims regarding recidivism and earned time were unfounded. Dr. de Haan acknowledged that DOC did not have data that demonstrated a cause-and-effect relationship. We thought this was the last we would hear of this study.
However, on 4/25/2003, the Salem Statesman Journal paraphrased Dr. de Haan again making claims about recidivism and earned time to support a new proposal to increase earned time. The Statesman Journal wrote:
But de Haan, the head of the state prison system, endorses earned time as a powerful motivational tool that gives inmates incentive to obey prison rules and to participate in rehabilitative programs.
It can produce significant public-safety dividends, de Haan said. He cited studies indicating that recidivism, or new crimes committed by people released from prison, drops markedly for those who earned more than 20 percent reductions in their sentences.
The article clearly left the impression that DOC had research showing that increasing earned time causes lower recidivism. As the our letter of June 20, 2002 details, DOC has no such research, a fact which DOC had acknowledged to us at least twice.
On April 30, 2003, Howard Rodstein of Crime Victims United gave testimony on this question before the Legislature's Joint Ways and Means Subcommittee on Public Safety. He explained the history behind this claim and explained why it is unsupported. The two witnesses who spoke immediately prior had cited the DOC study as evidence that increasing earned time causes lower recidivism and used this to support their agenda to increase earned time.
On May 27, 2003, an Associated Press article appeared in the Eugene Register Guard in which the suspect DOC claim was recycled.
Howard Rodstein emailed the Associated Press reporter, telling him that his article had recycled an invalid claim, and sent copies of this email to Dr. Paul Bellatty and Dr. Ben de Haan.
On 6/2/2003, Dr. Bellatty emailed Howard Rodstein and once again acknowledged that the DOC study did not show cause and effect. This was at least the third time that DOC had acknowledged this to Crime Victims United.
On 6/6/2003, a Salem Statesman Journal editorial again recycled the false claim and used it to support a call for legislation to increase earned time.
There is now a distinct possibility that legislation will be passed based on recycled misleading statements from DOC. Even people who support increasing earned time should blanch at the idea of legislation and criminal justice policy being formulated on misleading claims. Crime Victims United has called on DOC to correct this false impression by making it clear to legislators and lobbyists that it has no study showing that increasing earned time causes lower recidivism.
On 6/12/2003, to try to stop the false conclusion from recirculating, Crime Victims United sent a letter to legislators and media along with supporting material.
On April 12, 2005, the study was again cited as showing that increasing earned time causes lower recidivism by advocates for criminals testifying at a legislative hearing.
On June 10, 2005, Steve Doell and Howard Rodstein met with Dr. Paul Bellatty who once again said that the study does not show cause-and effect.
This is the letter sent by Crime Victims United to DOC Director Ben de Haan after his initial statements regarding earned time and recidivism.
Dr. Ben de Haan, Director
Oregon Department of Corrections
Dear Dr. de Haan:
I'm writing to respond to recent claims regarding the relationship between the availability of earned time and rates of recidivism.
On May 9, 2002, Crime Victims United President Steve Doell attended a Senate budget committee hearing at which the Department of Corrections proposed a modification to Measure 11 which would allow 20% earned time for certain Measure 11 offenders. At the hearing DOC officials claimed that research showed that the availability of earned time causes reduced recidivism. The DOC distributed charts to support this claim.
On May 24, 2002, an article appeared in the Salem Statesman Journal which contained the following passages relating to earned time and recidivism:
It's a solid proposal, corrections leaders said, because research shows that inmates who earn time off their sentences are less likely to commit new crimes after leaving prison.
. . .
Despite the controversy, corrections officials deemed the proposal worthy of consideration. "What we're doing is surfacing policy options, backed up by research," de Haan said.
A Corrections Department study that tracked former inmates who earned sentence reductions of at least 20 percent found that 10 percent subsequently committed new crimes. That's three times lower than the overall 30 percent recidivism rate for released prisoners in Oregon.
The implication is that DOC research shows that the availability of earned time causes dramatically lower rates of recidivism.
Crime Victims United President Steve Doell asked the DOC for a copy of the research that led to these claims. Eventually his efforts to look into this matter led to a discussion between me and Dr. Paul Bellatty of the DOC research unit. Dr. Bellatty sent copies of the charts to me and explained what the data represents. I appreciate the cooperation of the DOC in explaining this.
After reviewing the data, it is very clear that this study in no way supports the conclusion that the availability of earned time causes lower recidivism. The study suffers from several flaws which render it of no validity whatsoever in making a judgment on this relationship.
The problems with the study fall into two categories. First, the raw data mixes disparate quantities. Second, even if the data were pure, the design of the study lacks the kind of controls that would permit a conclusion on cause-and-effect.
I will first address the problems with the raw data.
Consider the chart labeled "Amount of Earned Time and Recidivism".
Dr. Bellatty of the DOC research unit told me that the quantity shown as earned time is actually a mix of earned time and time spent in county jail, as opposed to state prison. My understanding is that it was computed by taking the total sentence and subtracting the time served at DOC since these are the quantities that are recorded in the DOC database. The data presented leaves no way to tell how much of what is reported as earned time is actually time served in county jail.
Furthermore, there is no way to tell whether a relatively large amount of earned time (more than 4.3 months in the chart) represents a high percentage of earned time from a short sentence or a low percentage of earned time from a long sentence.
Furthermore, the data mixes information from behavioral, property and person offenders. It is well known that behavioral and property offenders have much higher rates of recidivism, whether earned time is available or not. It is very plausible that the offenders in the "more than 4.3 months" group fall there because they are person offenders serving relatively long terms and that the correlation is due merely to the fact that person offenders have lower recidivism rates without regard to earned time.
I will now address the problem with the design of the study, assuming that the study is intended to shed light on a possible cause-and-effect relationship between earned time and recidivism.
In order to a make a reliable conclusion about cause-and-effect, a study must compare two groups which are highly similar except in the factor that is hypothesized to be causative. In this study, all of the offenders were eligible for earned time, so there is no control group. Consequently this study is inherently incapable of demonstrating a cause-and-effect relationship between earned time and recidivism.
The chart "Amount of Earned Time and Recidivism" shows a negative correlation between absolute earned time and recidivism. There is no justification for concluding that this correlation shows that earned time causes lower recidivism. As with any correlation, there can be, and usually are, other factors driving the correlated quantities. In this case, one such factor is the simple fact that those people who can comply with rules on the inside tend to do better on the outside. Another factor is the fact that person offenders serve longer terms, therefore receive more good time measured in absolute time, and exhibit lower recidivism rates whether earned time is available or not. There may be many other driving factors. This study does not have the power to elucidate these factors.
I think you will appreciate the flaw in mistaking correlation for cause-and-effect by examining the chart labeled "Time in Prison and Recidivism".
This chart shows a negative correlation between time spent in prison and recidivism. From this correlation one could jump to the conclusion that we can reduce recidivism merely by increasing sentence length, for example, by eliminating all earned time. Such a conclusion is not supported by the data. The correlation shown in the chart in no way demonstrates a cause-and-effect relationship.
I have a suggestion for another study which, though not perfect, would have greater analytic power. I suggest that DOC compare the 1-year, 2-year and 3-year recidivism rates of Measure 11 offenders to a similar group of pre-Measure 11 offenders. To eliminate complicating variables, both groups should be limited to offenders convicted of the same offense. I suggest starting with robbery in the second degree. Over time, this could be extended to cover other offenses. Data for offenders convicted of different offenses should be reported separately so that compared groups are comparable. Although this would be an interesting study, it would not have the power to prove anything since the two groups would have been released from custody during different eras and would consequently be subject to a myriad of different driving factors.
We have taken the time to scrutinize and respond to DOC's claims because experience tells us that, once promulgated, such claims become part of people's beliefs even if they are totally unfounded, as in this case. We have learned that misconceptions are passed from person to person until they become entrenched as fact in people's minds. Once entrenched, it is very difficult to dislodge these misconceptions.
Consequently, we ask that you take the earliest opportunity to make clear to legislators and to the media that there is no research proving that the availability of earned time causes lower rates of recidivism.
Sincerely,
Howard Rodstein
Crime Victims United
After the 4/25/2003 Salem Statesman Journal article in which Dr. de Haan was again quoted making claims about earned time and recidivism, Howard Rodstein of Crime Victims United testified on this matter at the legislature. Here is his testimony.
Mr. Chair, Members of the Committee, my name is Howard Rodstein. I'm a director of Crime Victims United and do policy analysis for it. I'm also a software developer who has spent the last 25 years working in the area of scientific data analysis, among others. I'm here to testify about claims regarding earned time credits for prisoners.
Last Friday's Statesman Journal quoted DOC director Dr. Ben de Haan as saying studies indicate that recidivism drops markedly for those who earned more than 20 percent reductions in their sentences. A year ago I looked into similar claims and found them to be baseless. If DOC has new research on this issue, I am unaware of it.
We took the time to scrutinize and respond to DOC's claims because experience tells us that, once promulgated, such claims become part of people's entrenched beliefs even if they are totally unfounded. That's why I felt it necessary to explain the background of these claims to this committee.
In May of 2002 Dr. de Haan made statements at an interim committee hearing and later to the Salem Statesman Journal to the effect that DOC research showed that the availability of earned time reduces recidivism. After we made several request to see the basis of this claim, DOC put us in touch with Dr. Paul Bellatty of the DOC Research Unit.
I spoke at length with Dr. Bellatty and examined his charts. I found that what Dr. de Haan characterized as research was actually a simplistic analysis of totally uncontrolled statistics culled from the DOC database. These statistics in no way justify the claim that earned time causes lower recidivism. I sincerely doubt that any statistician would make such a claim based on this data, including DOC statisticians.
The problems with the study fall into two main categories. First, the raw data mixes disparate quantities. Earned time is mixed up with time spent in county jail. Data for behavioral, property, and violent offenders is mixed together.
Second, the design of the study lacks the kind of controls that would permit a conclusion of cause-and-effect. There is no control group. A correlation between earned time and reduced recidivism is improperly taken to show cause-and-effect. Alternative plausible explanations are not considered, and contradictory data is ignored.
A more detailed critique is available on the Crime Victims United web site.
On June 21, 2002, Dr. de Haan was kind enough to meet with Crime Victims United President Steve Doell and me to discuss his claims. He acknowledged that the DOC study was incapable of demonstrating cause-and-effect.
I thought that the matter was settled, but after reading the recycled claims in last Friday's Statesman Journal, it appears that I was wrong.
Through my work with Crime Victims United, I have met many victims for whom the subject of earned time credits is a source of anguish. There may be valid reasons for allowing earned time, but the claim that research has shown that it causes lower recidivism is unsubstantiated.
Thank you.
On May 28, 2003, an Associated Press article appeared in the Eugene Register Guard and again recycled the misleading DOC claims. Here is an excerpt:
A recent study by the Oregon Department of Corrections found that the more earned time felony prisoners acquired, with 20 percent being the maximum, the less likely they would return to a life of crime after their release.
"The folks who behave better in prison will behave better when they get out," said Paul Bellatty, manager of research for the prison system.
Dr. Bellatty's statement is common sense. Prisoners separate themselves into groups according to their behavior. Those who are by nature less antisocial behave better in prison than those who are more antisocial, and this dichotomy plays out in the outside world as well. The "recent study" in no way supports the DOC implication that the availability of earned time caused the lower recidivism.
On June 6, 2003, a Salem Statesman Journal editorial recycled the false claim again, using it to call for legislation to increase earned time. Here is an excerpt:
Plan to let prisoners earn time off is good
An Oregon Department of Corrections study found that the more time off felony prisoners earned, the more likely they would avoid crime after release. The study stopped at 20 percent, the current maximum allowed.
The obvious solution is to boost the ceiling to 30 percent and extend the period for transitional release.
On June 12, 2003, Crime Victims United sent the following letter to legislators and media.
Crime Victims United is strongly opposed to increasing earned time beyond the 20 percent and increasing transitional leave beyond the one month that inmates can currently receive. The Sentencing Guidelines system was created to restore the credibility of a sentencing system which had generated the contempt of the public and the offender population during the 1970's and 1980's.
The voters clearly expressed their position on truth in sentencing through Measure 11 (1994), Measure 74 (1999), and Measure 94 (2000).
In addition to our opposition on the substance of the issue, we are also concerned about the manner in which the debate has been conducted. The Oregon Department of Corrections has made statements to the press and in a legislative hearing giving the impression that they have research which shows that increasing earned time causes lower recidivism.
Their research does not support this conclusion, a fact which DOC has acknowledged to us on several occasions.
However this unsupported conclusion continues to circulate in the press and in the legislature. It was recently recycled in the Salem Statesman Journal (4/25/2003 news story, 5/27/2003 news story, 6/6/2003 editorial), in the Eugene Register-Guard (5/27/2003 news story), in the Albany Democrat-Herald (5/24/2003 news story) and more than likely in other media. We heard it recycled by two witnesses in a Ways and Means Public Safety Subcommittee hearing at which I testified on 4/30/2003.
Crime Victims United is concerned that legislation may be passed and criminal justice policy may be formulated based on misleading statements. Even those who favor increasing earned time should blanch at this prospect.
We feel that DOC has a duty to stop the recycling of this unsupported conclusion by making it clear to the press and to the legislature what they have acknowledged several times to us - that they have no research showing that increasing earned time causes lower recidivism.
I have attached material explaining the background of this issue and the basis for our concerns.
Sincerely,
Howard Rodstein
Crime Victims United Director
Dr. de Haan and Dr. Bellatty of DOC use the term "association" for a relationship between the variation of one measured quantity and the variation of another. Such a relationship is also called a "correlation".
Dr. de Haan explained the distinction between an association and cause-and-effect like this: Every spring the birds fly north and the flowers bloom but no one believes that the birds' flight is the cause of the blooming flowers.
In nature and in society, correlations can be found everywhere. The overwhelming majority of these correlations clearly have nothing to do with cause-and-effect, like the birds/flowers example. The remaining small minority of correlations are often assumed to be cause-and-effect relationships, but the vast majority even of this subcategory are also not true cause-and-effect relationships.
An example of how common it is to mistake a correlation for cause-and-effect is the prevailing wisdom about peptic ulcers prior to the 1983. For centuries, people believed that peptic ulcers were caused by psychological problems such as stress. Many people even went through lengthy and costly psychotherapy to try to cure their ulcers. In 1983, Dr. J. Robin Warren and Dr. Barry Marshall presented evidence that peptic ulcers are caused by bacteria, not by stress. This is now the prevailing view.
How did it come about that people held such a misguided notion of cause-and-effect regarding peptic ulcers for so long? Possibly ulcers caused patients to worry and have stress. Doctors noticed this correlation and assumed that the worry caused the ulcers when in fact it may have been the other way around. Or possibly doctors noticed stress in ulcer patients more than in the general population because ulcer patients came under closer scrutiny. There are a myriad of other possible explanations. Accepting an explanation of cause-and-effect without scientific evidence can lead us down a long and costly blind alley.
There is a parallel in the case of earned time and recidivism. Less antisocial prisoners are more likely to behave well in prison than more antisocial prisoners. Therefore they are more likely to get more earned time. Because they are less antisocial, they are also less likely to re-offend once they leave prison. It is a fallacy to assume that the increased earned time caused the lower recidivism rate. A plausible explanation is that both are a result of the prisoner's less antisocial nature. There may be many other plausible explanations.
In order to demonstrate that increasing earned time causes lower recidivism, you would ideally need to design a study that eliminated all other possible explanations. One way to approach this standard of proof is to randomly separate prisoners into two groups, one which received earned time and one which did not. Because the prisoners for each group are chosen randomly, it is likely that other factors will cancel themselves out. For example, there would be no reason to believe that the prisoners in one group were more or less antisocial than the prisoners in the other, because the groups were selected randomly from the same pool.
If you conducted an experiment like this with a sufficiently large number of prisoners in each group, and if the group receiving earned time had a significantly lower recidivism rate, this would be a good indication that there may be a true cause-and-effect. To verify it, you would need to reproduce the study in other jurisdictions to eliminate other explanations for the results, such as bias or error in the conduct of the experiment.
In 1998, the University of Maryland, under commission from the National Institute of Justice, produced a massive report on what works to prevent crime. This report was an exhaustive review of all of the research on preventing crime from all parts of the United States. The report rated each research project on a scale of scientific merit called the Maryland Scale of Scientific Methods.
The hypothetical ideal study outlined in the preceding paragraphs would be at the top of the scale of scientific merit, but it would be very difficult to carry out. The DOC study, while easily carried out, would be at the very bottom of the Maryland scale because it lacks all of the controls of a true scientific experiment. Somewhere in-between, a practical and credible study may be possible, but the DOC study does not approach this standard.
In the Spring of 2005 claims about earned time and recidivism resurfaced, this time in support of Senate Bill 435 which proposed to increase the available earned time from 20 to 33 percent.
By this time Dr. Ben deHaan had left the Oregon Department of Corrections to become the Director of the Criminal Justice Policy Research Institute at Portland State University. He then left that position and Ms. Claudia Black was named interim Director.
On April 12 the Senate Judiciary Committee held a hearing on SB 435. Ms. Black testified about further study into the earned time/recidivism connection. Here is her testimony.
Madame Chair and members of the Committee, for the record, my name is Claudia Black and I am the Interim Director of the Criminal Justice Policy Research Institute at Portland State University.
I am here today to present information on an analysis the Institute performed to identify factors associated with recidivism and to identify those inmates most likely to recidivate. It builds on previous work conducted by the Department of Corrections to determine which inmates were more likely to recidivate, but this analysis uses more historical/demographic variables and more sophisticated analyses. The Department of Corrections asked the Criminal Justice Policy Research Institute to undertake this study in an effort to help them target limited resources to those inmates most likely to recidivate.
The analyses included inmates eligible for earned time who were released from a Department of Corrections' institution between 1995 and 2000. The 8,253 inmates were followed for three years post-release.
The analysis incorporated variables examined in previous Department of Corrections' studies and those variables identified in the literature as possibly affecting recidivism.
We looked at the following variables:
age
previous theft convictions
prior incarcerations
earned time
sentence length
type of crime
revocationsWe also examined combinations of these variables to see if there were interactive effects. The analysts used Cox Regression, also known as proportional hazards regression, to relate inmate characteristics with recidivism.
Recidivism is defined as a conviction for a new felony within three years from prison.
Findings
It is important to note that approximately 75 percent of inmates leaving a Department of Corrections' institution do not get a new felony conviction within three years of release. Twenty-five percent of former inmates in this study were convicted of a new felony within three years of release.
These findings focus on the following variables: earned time, offender age, and sentence length. The relationship between recidivism and sentence length is more pertinent to other legislation (SB 436) being proposed. Therefore, Ben de Haan, the former director of the Institute, has offered to present on sentence length at a future date if the Committee is interested; my testimony will concentrate on earned time and offender age as it relates to this bill.
Inmates with more earned time are less likely to recidivate. Stated differently, those who behave well when incarcerated are more likely to behave well in the community. For example, a 20 year-old with zero percent earned time is 56 percent more likely to recidivate than a 20 year-old with 20 percent earned time.
Recidivism is related to age; younger inmates are more likely to recidivate. The relationship between age and recidivism is strongest with inmates less than 35 years of age. A 20 year-old inmate is 2.7 times more likely to recidivate than a 35-year old.
The relationship between earned time and recidivism is also much stronger for younger inmates. Younger inmates with little earned time are much more likely to recidivate than younger inmates who earn the maximum credit of 20 percent. By age 35, earned time is a less useful predictor of which inmates will recidivate.
The relationship between earned time and recidivism seems to be curvilinear, the more earned time, the less likely an inmate is to recidivate. We could only test this relationship to the maximum of 20 percent earned time; however, it is possible that recidivism rates could drop further with higher levels of earned time credit.
Thank you for the opportunity to present this information.
Later in the hearing, Howard Rodstein of Crime Victims United gave this testimony:
There's been a lot of talk that "research shows that increasing earned time causes lower recidivism". This is not the case.
We have dealt with this myth for years. It started with a DOC study under former Director Dr. Ben de Haan. I had a long meeting with him on this subject. During that meeting Dr. de Haan acknowledged that the study did not provide evidence that increasing earned time causes lower recidivism, but merely that there was an "association".
There is a vast difference between an association and a cause-and-effect relationship. Much of science is devoted to making this distinction but it is completely lacking from the study on recidivism. Our written testimony explains this in detail.
One conclusion of the study is:
"The relationship between earned time and recidivism seems to be linear, but we only tested it until 20 %. We have no reason to believe that it would not continue."
According to this theory, increasing earned time to 100 percent for all offenders would cut recidivism to zero. This is absurd on the face of it and illustrates how fundamentally unscientific this study is.
We should not base public safety policy on pseudo-science. The argument that increasing earned time will increase public safety is without merit.
Mr. Rodstein's written testimony went into greater detail:
There's been a lot of talk that "research shows that increasing earned time causes lower recidivism". This is not the case.
We have dealt with this myth for years. On May 9, 2002, Dr. Ben de Haan, at that time the director of DOC, made a presentation to the interim Senate Budget Committee in which he implied that a study showed that increasing earned time causes decreased recidivism.
I had a long meeting with Dr. de Haan on this subject (June 21, 2002). During that meeting Dr. de Haan acknowledged that the study did not provide evidence that increasing earned time causes lower recidivism, but merely that there was an "association".
Dr. de Haan used an analogy to illustrate the difference between an association and a cause-and-effect relationship. He said "Every spring birds fly north and flowers bloom". This is obviously true yet no one thinks that birds cause flowers to bloom. This is an association with no cause-and-effect relationship.
Both events are obviously driven by the same underlying cause - the change of seasons. In the case of recidivism, again both higher earned time and lower recidivism are obviously driven by the the same underlying cause - the characteristics of the offender. Increased earned time does not cause lower recidivism any more than birds cause flowers to bloom.
Much of science is devoted to separating the rare meaningful cause-and-effect relationship from the mountain of associations. The only way to find the needle in the haystack is to compare comparable groups. This can be done by comparing recidivism of Measure 11 criminals with recidivism of their pre-Measure 11 counterparts. The current study begs the question by comparing those with more earned time to those with less earned time, as if these groups were randomly selected.
The latest version of the study slices and dices the association in a dozen ways. This is like trying to decide if it's the robins that cause the tulips to bloom or if it's the sparrows.
One conclusion of the study is:
"Earned time rates under 20 % are less effective in reducing recidivism. The relationship between earned time and recidivism seems to be linear, but we only tested it until 20 %. We have no reason to believe that it would not continue."
According to this theory, increasing earned time to 100 percent for all offenders would cut recidivism to zero. This is absurd on the face of it and illustrates how fundamentally unscientific this study is.
We should not base public safety policy on pseudo-science. The argument that increasing earned time will increase public safety is without merit.
