There are studies that have been done of recidivism based upon various factors, including but not limited to, the nature of the offense. Most of these factors (e.g. age of offender at time of release) have heavy overlap across all offenses.
For example:
A Florida Study ranked the relative importance of different factors in
the likelihood that a released offender will commit another crime. Not
surprisingly, a poor disciplinary record while in prison, serving
prison time in a high security facility, and a history of prior
recidivism are connected with future criminal acts. People who commit
economicly motivated crimes (burglary, property offenses and drug
offenses) are also more likely to commit crimes again, than violent
criminals and sex offenders. Crimes are more likely to be committed
again by released prisoners who are young, have served shorter
sentences, and are less educated. Blacks are more likely to reoffend,
Hispanics are less likely to reoffend. Close post-release supervision
also lowers the reoffense rate.
Boil it all down and you see basically two forces at work. First,
someone is more likely to reoffend is crime was a way of life for
them, rather than an isolated incident (often driven by economic
factors and often illustrated by a history of offending), and second,
crime is predominantly something done by men of a certain age who tend
to cease to reoffend once they age and mellow.
One study reports that: 12.2% of non-sexual violent offenders
reoffend, 13.4% of sex offenders reoffend, and 36.9% of other
offenders reoffend. This doesn't mean that more care with serious
offenses isn't in order. A repeat sex offense or aggrevated assault is
far worse for society than a repeat burglary or car theft. But, it
does confirm the Florida statistics and the trend in national
statistics. Texas statistics show a less marked distinction between
different offenses, but similar trends except with regard to sex
offenses, for which it found high recidivism rates.
(Discussed here on September 29, 2005; this post's links to the original sources are rotted.)
Indeed, the U.S. Sentencing Commission itself has done exactly these kinds of studies for basically this purpose. For example:
The fewer run-ins with the law you have had before, the less likely
you are to reoffend, according to a May 2004 U.S. Sentencing
Commission report:
The analysis [of empirical data on re-offending] delineates recidivism
risk for offenders with minimal prior criminal history and shows that
the risk is lowest for offenders with the least experience in the
criminal justice system. Offenders with zero criminal history points
have lower recidivism rates than offenders with one or more criminal
history points. Even among offenders with zero criminal history
points, offenders who have never been arrested have the lowest
recidivism risk of all.
Federal courts deal with many criminal defendants with little criminal
history:
The annual proportion of federal offenders with zero criminal history
points is substantial. In fiscal year 2001, more than four out of
every ten federal offenders (42.2%) had zero criminal history points
for their Chapter Four computation. This proportion was even higher in
fiscal year 1992, when nearly half (49.9%) of citizen offenders had
zero criminal history points.
About 29.8% had no prior arrests or convictions, 8.4% had been
arrested but not convicted, and 1.5% had only minor prior criminal
convictions, in 1992. Defendants with no criminal record are much more
like to be women, to be white, to be older than 41 years old, less
likely to have used drugs, more likely to be employed, more likely to
be married, more likely to be high school graduates or have some
college, and are more likely to have dependents. Non-violent drug,
fraud and theft offenses are the most common charges, making up about
three quarters of cases, with fraud and theft much more common than
for those with prior offense records. There is no significant
difference in mental illness rates.
The recidivism rate of those with no criminal history points is 11.7%,
for those with 1 criminal history point is 22.6% and for those with
more criminal history is 36.5%. Within those with no criminal history
points:
Among the potential first offender groups, group A [no prior
convictions or arrrests] has the lowest primary recidivism rate at 6.8
percent, followed by offender group C [minor convictions, no prior
arrests] with a rate of 8.8 percent. Offenders in group B have a
recidivism rate of 17.2 percent, [prior arrests but not prior
convictions] which is more than twice the rate of group A and nearly
twice the rate for group C. Even with the comparatively higher
recidivism rate, group B’s rate is still several percentage points
lower than offenders in the remaining zeroes group (21.7%) [zero
criminal history points but some prior conviction for a non-minor
matter].
(From a July 21, 2008 post whose links have rotted).
The Pew Center on the States released a report in late May of 2009 arguing for evidence based sentencing which noted that:
Sixty to 80 percent of state felony defendants are placed on
probation, fined or jailed in their local communities. Although the
United States has the highest incarceration rate in the world, there
are nearly three times more offenders on probation than in state
prisons. Recidivism rates among these felony defendants are at
unprecedented levels. Almost 60 percent have been previously convicted
and more than 40 percent of those on probation fail to complete
probation successfully. The high recidivism rate among felons on
probation pushes up state crime rates and is one of the principal
contributors to our extraordinarily high incarceration rates. High
recidivism rates also contribute to the rapidly escalating cost of
state corrections, the second fastest growing expenditure item in
state budgets over the past 20 years.
An August 31, 2022 analysis, citing other academic studies with an international reach has also looked at the issue. It notes that recidivism follows a power law relationship (with a small proportion of offenders committing a disproportionate share of crimes), that substance abuse and mental health are important factors in the likelihood of reoffending, that the economic return to non-criminal activity for offenders is an important factor, and that criminals do not, in general, behave like "rational actors" in response to incentives. The article doesn't mention the fact, but a history of traumatic brain injury is another major risk factor for recidivism.
See also a January 2011 paper on the subject (evaluating a Dutch recidivist sentencing law), and U.S. and Colorado recidivism data collected here.
It also, however, recognizes that not all crimes are equally harmful to society, and likewise, that not all new crimes committed by past offenders are of equally concerns to society. Someone who dines and dashes over and over again is not as much of a threat to society as someone who murders someone even once upon being released.
It is not, in general, practicable to base sentences solely upon this criterion, however.
In part, this is because even if someone who is released from prison has a high probability of reoffending, very long sentences for minor crimes are considered to be disproportionate. In one famous case that the U.S. Supreme Court declined to overrule a cruel and unusual under the 8th Amendment to the U.S. Constitution, a man was sentenced in California under a three strikes and your out law to life in prison without possibility of parole for shoplifting something despite not having particularly heinous prior strikes. See Ewing v. California, 538 U.S. 11 (2003) and Lockyer v. Andrade, 538 U.S. 63 (2003). Much later, the California's statute in question was repealed by a citizen initiative backed by a majority of California voters.
Another problem is that criminal sentences (including those under the U.S. Sentencing Guidelines) consider two kinds of factors: one related to the nature of the offense, and the other related to the characteristics of the offender. Recidivism is significantly more closely related to the characteristics of the offender than it is to the nature of the offense. But, undue emphasis on offender characteristics to the point that they overshadow offense characteristics deviates from the sentencing principle that the punishment should fit the crime.
Both of these issues are well illustrated by the habitual offender statutes in Mississippi and Louisiana, adopted in 1994 and 1995 respectively, which have caused about 2,000 people in those two states combined to serve very long sentences of twenty years or more in prison, to de jure life in prison sentences, to de facto life in prison sentences, for comparatively minor offenses like possession of small quantities of illegal drugs for personal consumption.
While these statutes are designed (rather crudely but designed nonetheless) to reduce recidivism, they do so at the expense of other sentencing values and are widely criticized as result.
The logic behind using sentences for prior convictions is that while notions of what sentences are appropriate are inherently subjective, since they involve competing considerations, there is a stable intersubjective consensus which past exercises of judicial discretion (especially in the federal system which prior to the Sentencing Guidelines provided judges with very little guidance) reflects.
In general, it makes sense to capture the status quo in a clear fashion before attempting to reform it.
