including economics, social and demographic characteristics, culture, politics,

and incarceration rates. To date, policymakers’ emphasis on incarceration for

reducing crime has been premised, largely, on theories about its influence in

incapacitating active offenders and deterring would-be offenders. However,

thanks to rapid increases in crime and imprisonment through the 1970s and

1980s, followed by a sharp decrease in crime in the 1990s, we now have a

large body of recent empirical work on the effects of incarceration to draw on

as well.

Much of this research seeks to quantify the association between the size of

a jurisdiction’s incarceration rate and its crime rate.10 Led primarily by economists,

these analyses have become increasingly sophisticated, examining

many factors and looking at data across jurisdictions and over long periods

of time.11 Much of this research has confirmed a relationship between higher

incarceration rates and lower crime rates. However, as Table 1 summarizes,

recent studies vary widely in their conclusions about how strong this relationship

is and, in some cases, whether it really exists after all.

For example, using national-level data, researchers have found that a 10 percent

higher incarceration rate is associated with anywhere from a 9 percent to a

22 percent lower crime rate.13 In contrast, analyses using state-level data found

a weaker association, concluding that a 10 percent increase in incarceration is

associated with a crime rate that is anywhere from 0.11 percent to 4 percent

lower.14 Similar estimates have been generated from studies using county-level

data, ranging from a 2 percent to a 4 percent crime-rate difference.15 Moreover,

several studies have found no relationship between incarceration rates and

crime rates.16 One study even found that higher incarceration rates were associated

with higher crime rates in states with already high incarceration rates

(incarceration rates above 325 inmates per 100,000 population).17

As these disparate findings suggest, the impact of incarceration on crime

is inconsistent from one study to the next. One could use available research to

argue that a 10 percent increase in incarceration is associated with no difference

in crime rates, a 22 percent lower index crime rate, or a decrease only in

the rate of property crime.18 Therefore, to be guided by the available empirical

research, policymakers clearly need to develop a more nuanced understanding

of the literature.

Devine, Sheley, and Smith (1988)19 National — 1948–1985

–28.4 (violent offenses)

–19.9 (property offenses)

–22.0 (index offenses)

Marvell and Moody (1997, 1998)20 National — 1958–1995

–7.9 (violent offenses)

–9.5 (property offenses)

–9.3 (index offenses)

Marvell and Moody (1994)21 49 states — 1971–1989 –1.6 (index offenses)

Besci (1999)22 50 states and D.C. — 1971–1993

–0.46 (violent offenses)

–0.91 (property offenses)

–0.87 (index offenses)

Rapheal and Winter-Ebmer (2001)23 50 states — 1971–1997 not significant (violent offenses)

–1.1 (property offenses)

Donahue and Levitt (2001)24 50 states — 1973–1997 not significant (violent offenses)

–1.6 (property offenses)

Levitt (2001)25 50 states — 1950–1999 –0.76 (property offenses)

–1.3 (violent offenses)

DeFina and Arvanites (2002)26 50 states and D.C. — 1971–1998

not significant (murder, rape, assault, robbery)

–1.1 (burglary)

–0.56 (larceny)

–1.4 (auto theft)

Kovandzic and Sloan (2002)27 57 Florida counties — 1980–1998 not significant (index offenses)

Washington State Institute

for Public Policy (2003)28 39 Washington counties — 1980–2001 –2.4 (index offenses)

Liedka, Piehl, and Useem (2006)29 50 states and D.C. — 1970–2000

–0.118 (index offenses)

(states with incarceration rates <325)

+0.05 (index offenses)

(states with incarceration rates >325)

Kovandzic and Vieraitis (2006)30 58 Florida counties — 1980–2000 not significant (index offenses)

Levitt (1996)31 50 states and D.C. — 1971–1993 –3.8 (violent offenses)

–2.6 (property offenses)

Spelman (2000)32 50 states and D.C. — 1971–1997 –4.0 (index offenses)

Spelman (2005)33 254 Texas counties — 1990–2000 –4.4 (violent offenses)

–3.6 (property offenses)

are not equal in design and value, of course. Different findings often arise

because of crucially important methodological choices that researchers make

and which the literature has only recently begun to address and settle. These

include:

• the choice of whether to study the relationship between incarceration

and crime at the county, state, or national level (i.e., the level of aggregation

used);34

• whether a study accounts for the fact that incarceration and crime

occupy a two-way street in which each influences the other (i.e., whether

an analysis controls for simultaneity); and

• the number of other factors affecting crime that a study takes into

account (i.e., the specification of other factors potentially associated with

crime rates).

Such technical decisions can produce tremendous differences in results, leading

researchers and policymakers to strikingly different conclusions.

On which findings should policymakers rely, then? “[I]f...results are to

converge on a defensible estimate,” argues William Spelman, perhaps the

leading researcher in this area, “[t]hey must measure effects at the lowest level

of aggregation possible,” “account for simultaneity,” and control for as many

other factors as possible.35 Accordingly, the most reliable studies are those conducted

at the state or county level that account for simultaneity and consider

a significant number of crime-related factors. To date, only three studies—two

by Spelman and one by Steven D. Levitt—include all three of these criteria.36

Interestingly, all three have produced a fairly consistent finding, associating a

10 percent higher incarceration rate with a 2 to 4 percent lower crime rate.37

Levitt’s and Spelman’s findings have garnered a great deal of attention from

both supporters and opponents of a continued emphasis on incarceration.

Supporters take the findings as a confirmation that prison works, concluding

that every 10 percent increase in incarceration rates will produce a 2 to 4 percent

decrease in crime rates. Opponents, on the other hand, see the findings

as a confirmation that prison does not work very well at all. They maintain that

even a 4 percent decrease in crime is not much for a 10 percent increase in

incarceration.38 Indeed, Spelman himself characterizes a 2 to 4 percent crime

reduction as a fairly limited impact given the sizable financial obligation states

have incurred in incarcerating so many people.

An even more complex picture emerges from analyses focusing on the

neighborhood level.39 A growing body of research examining specific neighborhoods

finds that more incarceration can actually lead to increasing crime rates.40

Several recent studies maintain that communities may reach an incarceration

“tipping point.” Dina Rose and Todd Clear, for example, found that the level of

crime in several Florida communities increased after the incarceration rate of

individuals from those communities reached a certain level.41 Rose and Clear

argue that high rates of imprisonment break down the social and family bonds

that guide individuals away from crime, remove adults who would otherwise

nurture children, deprive communities of income, reduce future income potential,

and engender a deep resentment toward the legal system. As a result, as

communities become less capable of maintaining social order through families

or social groups, crime rates go up.42

between crime and incarceration were established, a precise ratio could not be

applied to policymaking in every location. This is because empirical research

on incarceration and crime is not meant to be a prescription for future imprisonment

policies at the local level, even though this is how some seek to use

it. Research only provides an estimate of average relationships between incarceration

and crime rates across jurisdictions. Thus, such findings are a blunt

instrument whose applicability to any specific jurisdiction is dubious.

Research, therefore, cannot predict the impact of future prison increases

in a given state. What happens in any particular jurisdiction will depend on a

variety of factors that have yet to be contemplated by crime and incarceration

research. Some of these are environmental, such as social preferences, economic

changes, and political activity. Others are practical: the size of a state’s

current prison population, the way offenders are sanctioned or incarcerated,

and the types of offenders a state chooses to incarcerate. In the debate about

the impact of incarceration on crime, such issues are generally overlooked by

both sides and largely absent in the academic literature they draw upon. This

is surprising given the practical influence these considerations can exert on

the bigger issue of how crime and incarceration interact. The following brief

discussion of each of these factors may help illustrate the limitations.

The size of a state’s prison population and crime rate will influence the impact

of increases in incarceration rates. To date, most arguments for more incarceration

to further reduce crime have relied on Spelman’s estimate that a 10

percent increase correlates with a 2 to 4 percent lower crime rate. Even if that

estimate is accurate, however, for a state with an already high incarceration rate

the costs of increasing incarceration by 10 percent to achieve a 2 to 4 percent

reduction in crime could be tremendous. For example, California and Nebraska

had very similar crime rates in 2003 of approximately 4,000 index offenses

per 100,000 people in the population. To achieve a 2 to 4 percent reduction,

California, with a prison population of 162,678 inmates, would have to incarcerate

an additional 16,089 inmates.43 To achieve the same rate of reduction,

Nebraska, with a prison population of 3,976, would have to incarcerate just 400

additional inmates. If the average cost to incarcerate an offender for one year is

$22,650, California would spend $355 million more than Nebraska to achieve

the same level of public safety.44 The cost incurred per unit of crime reduction,

then, is substantially larger for California. Thus, an increase in incarceration

in a state with an already large prison population may require a huge boost in

actual prison populations that may be difficult to sustain economically.

Raymond Liedka, Anne Piehl, and Bert Useem have confirmed, moreover,

that increases in prison populations in states with already large prison populations

have less impact on crime than increases in states with smaller prison

populations.45 States experience “accelerating declining marginal returns, that

is, a percent reduction in crime that gets ever smaller with ever larger prison

populations,” they argue.46 Thus, increases in incarceration rates are associated

with lower crime rates at low levels of imprisonment, but the size of that association

shrinks as incarceration rates get bigger. Eventually, they say, there is

an “inflection point” where increases in incarceration rates are associated with

higher crime rates. This inflection point occurs when a state’s incarceration

rate reaches some point between 325 and 492 inmates per 100,000 people. In

other words, states with incarceration rates above this range can expect to experience

higher crime rates with future increases in incarceration rates. (This

state-level phenomenon recalls the neighborhood effects of high incarceration

rates cited by Rose and Clear.)

Incarceration is not the only punishment that may reduce crime rates. Other

types of punishment, including fines, probation, community service, drug

treatment, or other sanctions have also been shown to suppress crime. These

alternative sanctions are not considered in the crime control studies noted

above, however. Had they increased at the same time as the expansion of imprisonment,

these sanctions may have contributed—in part or even completely—to

the effects found in those studies. Similarly, the content of incarceration—the

quality of inmates’ experience in prison—may matter greatly as well. Studies

so far have only considered how the size of prison populations affects crime

rates. Although some research has examined how related factors like the

length of stay in prison or changes in supervision policies after release influence

recidivism, no studies apparently have considered how or whether such

factors affect crime rates.

The type of offenders a state decides to incarcerate may also be a relevant factor.

Franklin Zimring and Gordon Hawkins argue that continued expansion of the

prison system does little more than increase the number of individuals in the

criminal justice net without reducing criminal offending or crime rates.49 U.S.

prisons already housed the most serious violent offenders in the early 1980s,

they argue; prison expansion since then has resulted in nothing more than the

imprisonment of large numbers of nonviolent, “marginal” offenders. Thus,

since the worst offenders had already been incarcerated, Zimring and Hawkins

contend that increasing incarceration rates through the 1990s did nothing to

impact the crime rate.

Others have noted as well that the increased incarceration of drug offenders

has helped reduce the effectiveness of incarceration.50 Between 1980 and

2005, the number of inmates incarcerated for drug possession in state prisons

or local jails grew by more than 1,000 percent.51 By 2004, 419,000 drug

possessors were incarcerated in state prisons or local jails at a cost of nearly

$8.3 billion annually.52 Ilyana Kuziemko and Steven D. Levitt argue that the

continued increase in the number of drug offenders in prisons may lead to a

“crowding out” effect, in which the high number of incarcerated drug offenders

prevents the incarceration of offenders prone to more serious crime, thereby

reducing the effectiveness of incarceration to reduce crime. 52a

Analysts agree with apparent unanimity that future increases in incarceration

rates for such offenders will do less and cost more.53 Washington State,

for example, concluded that while more incarceration has led to less crime

in the state, the benefits of additional prison expansion will be smaller and

more expensive to achieve.54 Specifically, an increase in the incarceration rate

in 2003 prevented considerably fewer crimes than did previous similar-size

increases. The state further concluded that while incarcerating violent and

high-volume property offenders continued to generate more benefits than

The research on crime rates has not examined the impact of prison-based

programming, either. States vary a great deal in the amount and content of programs

offered to inmates. A large body of literature has found that the design

and content of specific programs can reduce individual recidivism rates.47 For

example, drug treatments using therapeutic community models or rehabilitative

programs tailored to the risks and needs of offenders have been shown to

reduce recidivism, while boot camps and unstructured rehabilitation programs

have been found to have no positive influence on recidivism.48 Given this demonstrated

impact on recidivism, it stands to reason that programming during

incarceration could also affect incarceration’s influence on crime rates.

Costs, in the future each additional person incarcerated will result in fewer prevented

crimes. Washington even found that increasing the incarceration rate

for drug offenders in the 1990s actually had a negative impact overall, as it

now costs more to incarcerate additional drug offenders than the average value

of the crimes prevented by their imprisonment. Money should not be the sole

measure by which policymakers evaluate the effectiveness or attractiveness of

a policy, of course. However, financial implications are a relevant concern of

officials facing limited public resources and a seemingly endless list of areas in

need of investment.

Between 1990 and 2005, the crime rate in the United States fell dramatically to

its lowest point in 30 years.55 However, as noted earlier, according to Spelman

only 25 percent of this crime drop through the 1990s could be explained by

increasing incarceration rates.56 The remaining 75 percent, therefore, must be

due to factors other than incarceration. Indeed, researchers have identified a

number of such factors including, for example, fewer young persons in the

population, smaller urban populations, decreases in crack cocaine markets,

lower unemployment rates, higher wages, more education and high school

graduates, more police per capita, and more arrests for public order offenses.57

An examination of just a few of these indicates that future investment in other

policy areas may be not only more effective but also more cost effective than

continued investment in increased prison populations (see Table 2).

the number of police per capita and lower crime rates. For example, using city level

data, Levitt found that a 10 percent increase in the size of a city’s police

force was associated with an 11 percent lower violent crime rate and a 3 percent

lower property crime rate.58 Thomas Marvel and Carlisle Moody similarly

found a 10 percent increase in the size of a city’s police force associated with

a 3 percent reduction in index crime rates.59 Using county-level data, Tomislav

Kovandzic and John J. Sloan also found associations between the size of police

forces and crime, concluding that a 10 percent increase in the number of

police was associated with a 1.4 percent lower index crime rate.60 At the state level,

however, Marvel and Moody found no association between the size of the

police force and crime rates. This disparate finding suggests that the impact of

increased police presence may be felt only at the local level.61

Using Marvel and Moody’s estimate tying a 10 percent increase in the size

of a city’s police force to a 3 percent decrease in the index crime rate, we can

imagine how a crime reduction policy focusing on policing might operate in,

say, New York City, where the 2004 index crime rate was 2,800 offenses per

100,000 people in the population. To achieve a 3 percent reduction in the crime

rate by increasing incarceration, New York City—with a prison population of

33,564 inmates—would have to incarcerate an additional 3,300 inmates at a

cost of approximately $121.5 million per year.62, 63 With a police force of 39,110

sworn police officers, the city could achieve the same reduction in crime by

hiring 3,911 more police officers at a cost of $97.2 million per year.64, 65 Compared

to policing, then, incarceration would cost the city $24.3 million more to

achieve the same level of public safety.66

It is important to note that in the same way the content of incarceration

policy may affect how imprisonment relates to crime, so too may the content

of policing policy influence how crime develops. Poorly structured policing

policies could negate the positive potential of an increase in officers. Also, an

increase in the number of police would not necessarily translate directly into

more law enforcement and arrests. Provided such concerns are taken into

account, however, this example illustrates the promise of increasing policing

as an alternative to increasing incarceration.67

interact on several levels in the United States. Incarceration creates problems

of low earnings and irregular employment for individuals after release from

prison by dissuading employers from hiring them, disqualifying them from

certain professions, eroding job skills, limiting acquisition of work experience,

creating behaviors inconsistent with work routines outside prison, and

undermining social connections to good job opportunities.68 Research has

Marvell and Moody (1996) 56 U.S. cities — 1971–1992 –3 (index offenses)

Marvell and Moody (1996) 49 states — 1971–1992 not significant (index offenses)

Levitt (1997) 59 U.S. cities — 1970–1992

–11 (violent offenses)

–3 (property offenses)

Kovandzic and Sloan (2002) 57 Florida counties — 1980–1998 –1.4 (index offenses)

Levitt (1996)83 50 states and D.C. — 1971–1993 not significant (violent offenses)

10 (property offenses)

Levitt (1997) 59 U.S. cities — 1970–1992 not significant (violent offenses)

10.4 (property offenses)

Rapheal and Winter-Ebmer (2001) 50 states — 1971–1997 not significant (violent offenses)

16.3 (property offenses)

Gould et al. (2002) 705 counties — 1979–1997 not significant (violent offenses)

16.6 (property offenses)

Gould et al. (2002) 705 counties — 1979–1997

–25.3 (violent offenses)

–12.6 (property offenses)

–13.5 (index offenses)

Raphael and Winter-Ebmer (2001) 50 states — 1971–1997 –1.6 (violent offenses)

not significant (property offenses)

Grogger (1998) Individual survey data (1980) –10 (index offenses)

Lochner and Moretti (2004) 50 states — 1960, 1970, 1980 –9.4 (index offenses)

shown that men with criminal records experience no growth in earnings and,

therefore, have few choices other than day labor.69 At the community level,

neighborhoods with high incarceration rates may be shunned by employers.

Further, as John Hagan argues, incarceration can generate social connections

to illegal rather than legal employment, thus, potentially increasing crime.70

Indeed, researchers have found that economic shifts have significantly

affected crime rates in the last decade. Using state-level data, Steven Raphael

and Rudolf Winter-Ebmer found that a 10 percent decrease in a state’s unemployment

rate corresponded with a 16 percent reduction in property crime rates;

the researchers concluded that, between 1992 and 1997, “slightly more than

40 percent of the decline [in the overall property crime rate] can be attributed

to the decline in unemployment.”71 Liedka, Piehl, and Useem produced similar

findings, also using state-level data.72 Analyses using county-level data have

produced estimates that find a similar association between unemployment and

crime; Eric Gould, Bruce Weinberg, and David Mustard, for example, found

that a 10 percent reduction in unemployment rates was associated with a 16.6

percent reduction in property crime.73 Both studies found no association, however,

between unemployment and violent crime.74 Raphael and Winter-Ebmer

conclude that “the magnitudes of the crime-unemployment effects...suggest

that policies aimed at improving the employment prospects of workers facing

the greatest obstacles can be effective tools for combating crime.”75

Research has also considered the relationship between real wages and

crime. Using national-level data, Gould, Weinberg, and Mustard determined

that a 10 percent increase in real wages saw a 13 percent lower index crime

rate—specifically, a 12 percent lower property crime rate and a 25 percent lower

violent crime rate.76 At the state level, Raphael and Winter-Ebmer found that a

10 percent increase in per capita income saw a 1.6 percent lower violent crime

rate; Liedka, Piehl, and Useem found similar associations at the state level.77

Using individual-level data, Jeffrey Grogger found that a 10 percent increase in

real wages was associated with a 10 percent decrease in crime participation at

the individual level.78 In examining the crime drop of the 1990s, Grogger and

Michael Willis suggested that a better economy allowed more young people

opportunities in the labor market rather than in crime. They attributed the

crime drop largely to the expanding economy.79

attract research attention. Lance Lochner and Enrico Moretti, for example,

have shown an increase in citizens’ education levels to be associated with lower

crime rates: specifically, a one-year increase in the average education of citizens

results in a 1.7 percent lower index crime rate. In addition, they associated

a 10 percent increase in graduation rates with a 9.4 percent lower index crime

rate. Combined with Gould and his colleagues’ findings on the link between

wages and crime, Lochner and Moretti argue that “a 10 percent increase in

high school graduation rates should reduce arrest rates by 5 to 10 percent

through increased wages alone.”80 “[A] 1 percent increase in male high school

graduation rates would save as much as $1.4 billion” nationally through crime

reduction, they conclude.81

Moreover, prison-based education programs have been found to significantly

reduce recidivism rates for offenders after release.82 In their recent

examination of the effects of prison education participation across three states,

researchers Stephen Steurer, Linda Smith, and Alice Tracy did not consider the

impact of such education programs on overall crime rates, but their findings of

significantly reduced recidivism for individuals who participated in such programs

underscore the impact of education on reentry success.

Given the demonstrated influence of these other factors on crime and the

decreasing impact of incarceration, criminal justice policymakers appear to have

placed undue emphasis on incarceration. As William Spelman has cautioned, “It

is no longer sufficient, if it ever was, to demonstrate that prisons are better than

nothing. Instead, they must be better than the next-best use of the money.”84

Yet, in the past two decades, spending on these other factors has been cut.

Corrections expenditures were the only state budget category other than Medicaid

to increase as a percentage of total state spending over the past 20 years.85

Between 1985 and 2004, states increased corrections spending by 202 percent.

By comparison, spending on higher education grew by just 3 percent,

Medicaid by 47 percent, and secondary and elementary education by 55 percent;

spending on public assistance decreased by more than 60 percent during

the same period (see Figure 1).86 Public opinion appears to be in harmony with

a move away from incarceration spending. Whereas 75 percent of Americans

believed that too little money was spent on halting rising crime rates in 1994,

by 2002, this had declined to 56 percent. In contrast, during the same period

the proportion of Americans responding that too little money was spent on

welfare increased from 13 percent to 21 percent.87

Such approaches are not new to criminal justice policy discussions. Many

commentators have argued for a crime reduction policy that focuses on the

labor market by addressing employment, wages, and education.89 In the lim-

ited areas where such policies have been implemented, they have targeted

individuals believed to be at high risk of crime involvement, such as high

school dropouts, inner-city youth, offenders, and ex-offenders. But a broader

approach would require a shift in criminal justice policy away from reactive

responses to criminal offending and toward a proactive attempt to address the

underlying causes of criminal offending. It also requires a more general shift

in policymaking attitudes toward the view that public safety may best be served

by protecting citizens from the factors that can contribute to crime.

Moreover, the public favors a policy that addresses the underlying causes of

crime rather than simply responds to crime after it occurs. Polling by Peter D.

Hart Research Associates, for example, shows the public questioning whether

incarceration is the best crime control policy. In 1994, 42 percent of Americans

favored responding to crime with stricter sentencing; by 2001, this had

decreased to just 32 percent. Conversely, in 1994, only 48 percent of Americans

said they favored addressing the underlying causes of crime; by 2001, this

had increased to 65 percent (see Figure 2).88

After 15 years of declining crime rates, many analysts are claiming that “prison

works.” But, as Elliot Currie notes, “if ‘prison works’ is the answer, what was

the question?”90 If the question is whether it is possible to prevent individuals

from committing crimes by putting them in prison, then prison certainly

works; it works to punish and incapacitate those who have committed crimes.

But if the question is what is the best way to reduce crime, “prison works”

may not be the most helpful response. Does a five-year prison sentence “work”

better to reduce crime than a two-year prison sentence? Does a two-year prison

sentence for nonviolent offenders “work” as well as a two-year prison sentence

for violent offenders?

The most salient question of all may be, Do the resources devoted to prison

“work” better to ensure public safety than if those resources were devoted to

something else?91 Prisons are not the only way to fight crime. Policymakers

could spend money on more judges, better staffed or equipped law enforcement,

or better-trained probation and parole officers. They could invest, as this

paper indicates, in other, non-criminal-justice areas shown to affect crime: education,

employment, economic development, etc. The impact of incarceration

on crime is limited and diminishing. The public’s support for reactive crime

control is also in decline. It is therefore fitting that we reconsider the continued

emphasis on and dedication of resources to incarceration.

Public safety cannot be achieved only by responding to crime after it

occurs; research shows that it may also depend on protecting people against

those factors that have been shown to be associated with high crime rates,

such as unemployment, poverty, and illiteracy. By pursuing crime reduction

chiefly through incarceration, states are forgoing the opportunity to invest in

these other important areas. As state policymakers continue to feel pressure

to introduce measures to keep crime rates low, they would therefore do well to

look beyond incarceration for alternative policies that not only may be able to

accomplish the important task of protecting public safety, but may do so more

efficiently and more effectively.