This article summarizes and expands on the paper “Do Sex Offender Registration and Notification Laws Affect Criminal Behavior?” by Prescott and J.E. Rockoff, Journal of Law and Economics, Vol. 54, No. 1 (2011).

State legislatures enacted SORN laws with the explicit and exclusive aim of reducing sex offender recidivism. Registration laws were designed solely to enable more effective law enforcement supervision (and apprehension, if necessary) of previously convicted sex offenders, who were assumed to be at serious risk for reoffending. Similarly, notification laws were passed with the singular goal of helping potential victims protect themselves from known and nearby sex offenders by facilitating the public monitoring and physical avoidance of these individuals. Proponents of registration and notification laws continue to defend them entirely on these grounds. The U.S. Supreme Court has concluded, at least with respect to early versions of SORN laws, that state legislatures intended merely to “regulate” released offenders who may prove to be dangerous, not to punish them for previously committed crimes.

The general idea that we ought to “regulate” released offenders—of any type—to reduce the likelihood of their returning to crime is an attractive one, at least in theory. Criminal recidivism generates significant social harm. As many as two-thirds of all released felony inmates are readmitted to prison within a few years, and the public generally views individuals convicted of sex offenses as among those most likely to reoffend upon release. It is not surprising, therefore, that the public’s impression of sex offender dangerousness, when catalyzed by a few gruesome and headline-grabbing crimes in the late 1980s and early 1990s, produced the first SORN laws, which in turn led to the federal acts that extended SORN laws nationwide.

Nevertheless, despite their now-widespread use, SORN laws became the norm without any systematic study of their consequences. Admittedly, the logic underlying these laws seems at first difficult to gainsay: if a known sex offender poses even a small risk to a potential new victim, how can it hurt if the police are keeping better tabs on that offender or if the offender’s neighbors are made aware that he is a threat so they can take measures to reduce their own risk of victimization? But this question and its implied answer presume that SORN laws have no influence on whether released sex offenders opt to pursue new victims in the first place. If the enforcement of notification laws imposes significant financial, social, and psychological costs on released sex offenders, as an avalanche of evidence suggests it does, then notification may in fact be criminogenic. The result may well be many more attempted attacks by convicted sex offenders and therefore higher recidivism rates on the whole, even if every individual attack attempted becomes somewhat less likely to succeed.

Furthermore, the logic offered by most SORN advocates ignores the potentially significant, yet unintended, consequences that these laws may have on many other distinct facets of sex offender behavior. For one example, SORN laws may function as a deterrent to potential sex offenders, i.e., those with “clean” records who want to avoid the prospect of being publicly branded a sex offender if they are caught and convicted of committing a sex crime. For another, released offenders who are subject to notification laws may react by attacking different (i.e., uninformed) rather than fewer victims, a phenomenon generally known as “crime displacement.”

A comprehensive study of the consequences of SORN laws is necessary, therefore, to improve our understanding of not only whether but also how these particular “regulatory” post-release methods affect sex offender behavior. Such research is also indispensable to determining whether these laws are ultimately wise policy.

Fortunately, analysis of the experiences of 15 states over almost 10 years provides needed insight into the effects that SORN laws may have on criminal behavior. Findings from these data indicate that registration alone may reduce sex offense recidivism against friends, acquaintances, and neighbors by improving local law enforcement supervision of sex offenders. But the evidence runs against the notion that notification reduces recidivism. Although notification laws do appear to deter nonregistered offenders, these tools may well increase recidivism among those the government publicly identifies as sex offenders by reducing their ability or desire to live without crime. Indeed, the idea that notification regimes may make registered offenders more dangerous is consistent with the fact that notification causes these individuals significant financial, social, and psychological harm.

Modeling the Behavioral Response of Criminals to SORN Laws

An empirical examination of the effects of registration and notification laws on criminal behavior requires the development of policy-relevant, empirically testable hypotheses. In order to catalog and organize the range of possible offender reactions to SORN laws, it pays to begin with the underlying determinants of criminal decisionmaking. Registration and notification differ in how they operate on these central inputs, and SORN laws also apply differently to convicted (registered) sex offenders than they do to potential (nonregistered) sex offenders, making it essential to understand the distinct behavioral dynamics in each of these two cases.

The likelihood that an individual commits a crime against a particular victim turns in significant part on four factors:

• The difficulty or cost (c) to the potential offender—unrelated to any punishment—of targeting and attacking the victim.
• The perceived probability (p) the potential offender will be caught and punished for any attack on the victim.
• The severity of the punishment the potential offender believes will be imposed for any crime against the victim (f).
• The benefits to the potential offender of engaging in other (presumably lawful) activities relative to the benefits he would enjoy from committing the offense (u).

By assumption, the likelihood that an offender commits a crime is increasing in the relative utility of crime commission (or decreasing in the relative attractiveness of choosing only legal behavior) and decreasing in the difficulty of targeting a victim, punishment probability, and punishment severity.

Findings from these data indicate that registration reduces sex offense recidivism by providing law enforcement with information on local sex offenders. But the evidence runs against the notion that notification reduces recidivism.

Registration laws, by enhancing the efficacy of police supervision, seem likely to reduce recidivism of registered offenders by increasing the probability of detection and punishment of any who commit new sex offenses. But registration alone appears unlikely to increase the actual difficulty of committing new crimes (assuming the identities of registered sex offenders are available only to local authorities and otherwise remain confidential), to alter the level of punishment (given convicted offenders are already registered), or to affect the relative attractiveness of participating in only noncriminal activities, at least if the registration procedures themselves are not too burdensome (and registration details remain secure).

Notification laws may also reduce recidivism among known sex offenders by offering potential victims information they can use to protect themselves. But this publicity, by generating significant negative collateral consequences for released offenders such as the loss of employment, housing, and social ties, may also increase their proclivity for crime. Therefore, although disseminating the identities of convicted offenders may indeed make it more difficult for them to attack “informed” individuals (e.g., neighbors and acquaintances) and may increase the probability of detection and apprehension if they do, notification may also significantly reduce the relative attractiveness of staying on the straight and narrow, leaving the overall effect on recidivism ambiguous.

Potential offenders—those who have never been convicted of a covered crime and so are not presently required to comply with SORN laws—may also be deterred by the prospect of becoming subjected to the demands of registration or notification in the future if they are caught and convicted of committing a sex offense. Otherwise, however, these individuals should be unaffected by SORN laws that by definition apply only to convicted sex offenders. As between the two types of SORN laws, potential offenders seem much more likely to be deterred by notification given the manifold and well-documented difficulties that released sex offenders are known to face.

Capitalizing on the Geographic and Temporal Variation in SORN Laws

A careful study of the effects of SORN laws on criminal behavior and crime frequency requires a reliable empirical strategy capable of differentiating between and evaluating these various hypotheses. The striking legal variation in SORN laws across states and over time (as well as the peculiar sources of this variation) offers a flexible approach to overcoming this hurdle. The implementation and evolution of state SORN laws during the 1990s and 2000s comprise, in effect, a whole series of natural experiments. By comparing the experiences of jurisdictions implementing or amending their SORN laws at particular points in time to the experiences of “control” jurisdictions under similar conditions, it is possible to disentangle and measure the behavioral consequences of registration and notification laws.

Of course, attributing causal significance to the close correspondence between legislatively enacted laws and near-contemporaneous changes in outcomes of interest is rightly viewed with suspicion. Correlation is not causation. Just as new criminal laws may influence criminal behavior, new prohibitions, more severe penalties, and higher police spending often follow from rising offense rates.

SORN laws, however, do not fit this mold. Although the timing and content of SORN laws vary considerably across the country, they do so for reasons unrelated to offense trends. Enacted during a period of declining sex crime frequency, state registration and notification laws were responses either to one or two highly atypical but well-publicized local incidents or to federal mandates (with significant grace periods), which were themselves inspired by a small number of conspicuous crimes. In addition to the random sparks of legislative activity generated by individual sex offenses, exogenously fixed state and federal constitutional constraints affected whether criminal laws like registration and notification were applied retroactively to individuals on the basis of crimes they committed in the past, and variation in preexisting legislative schedules partially dictated when SORN laws were enacted and became effective.

Not surprisingly, precisely measuring the timing and content of state SORN laws is critical to an accurate assessment of their consequences. But deciphering the meaning and history of these laws is no easy task. Every state’s laws evolved over time as a result of legislative amendments and judicial decisions, and SORN laws cannot be divided easily into one or even two types. Registration and notification laws are obviously designed to work in two different ways, but a canvassing of state laws imparts significant diversity even within these categories, especially in the form notification laws take. While early notification laws merely allowed public access to paper registries, later laws required that registration data be posted to publicly accessible web registries. Some states also enacted “active” notification laws under which officials must take affirmative steps to inform at-risk individuals (particularly neighbors) of relevant released offenders—typically by written notice or even a visit from a police officer.

SORN laws thus fit into roughly four categories: registration, public access to a paper registry, web registry availability, and active community notification. Even so, states followed different legislative paths. For example, Iowa began registration and (limited) public access notification simultaneously in 1995, but did not have any form of active notification until 1998 or an internet site until 2000. Texas, by contrast, began the registration of sex offenders in 1991, instituted public access and active notification in 1995, and launched an internet site in 1999. And, as jurisdictions moved from basic registration laws to more restrictive and varied forms of public access and from restrictive public access to the publication of sex offender information on the internet, states had to determine whether to apply such changes retroactively and, if appropriate, specify which former offenders were to be subject to the new laws.

This cross-state variation in SORN laws—in exact timing of enactment and effective dates, in content and scope, and in retroactive application—can be used to identify the causal effects of registration and notification laws on the commission of sex crimes. Simply combining these legal data with the number of individual sex-offense incidents by state and time period as a measure of the outcome of interest (using, perhaps, the FBI’s Uniform Crime Reports) would make it possible to estimate at least the average effects of SORN laws on the overall frequency of sex offenses.

Unfortunately, this approach is less useful than it might seem because it cannot separately identify the deterrence and recidivism effects of SORN laws. Furthermore, it fails to take advantage of the fact that registration and notification laws, by design or otherwise, may operate to produce specific predictable consequences. For instance, notification laws, if they work as intended, should not only cut the total frequency of sex crimes, but should also reduce the relative frequency of attacks by registered offenders on their neighbors (as opposed to strangers). An empirical strategy that tests for the existence of these additional patterns can provide a more complete view of the data and allows for more robust interpretation of any findings than would be feasible with simpler reduced-form approaches and any amorphous results they might produce.

Contrasting Crime Frequency Effects from Shifts in the Types of Victims Targeted

SORN laws may influence the number and nature of the sex offenses committed during any period of time in a variety of complementary and offsetting ways. Although registration and notification laws function differently, supplying information to diverse parties, both types of laws have the potential to influence at least a few of the key determinants of criminal behavior (e.g., c, p, f, and u)—presumably discrete sets and by distinct mechanisms. Because some of these inputs are victim-type specific, SORN laws seem likely to affect the relative frequency of particular offender-victim combinations (e.g., offenses against neighbors), not merely overall offense or recidivism rates. Thus, any comprehensive investigation into how SORN laws influence criminal behavior ought to exploit the specific means by which these laws operate.

Consider notification laws, which target recidivism by providing individuals located “near” potential recidivists (like significant others, neighbors, acquaintances, etc.) with identifying information that can be used to monitor or avoid potential threats. Notification seeks in effect to make it more difficult for a released sex offender to attack “nearby” individuals, many of whom it is assumed were previously unaware of the offender’s criminal history. If notification laws work as their proponents suggest, one would expect to observe relatively fewer attacks against the neighbors and acquaintances of released offenders. Likewise, under registration laws, police supervision of released sex offenders should, in theory, lead to fewer crimes against victims who the police view as more probable targets—such as family members, significant others, neighbors, and acquaintances—as these crimes would be easier for authorities to detect and prosecute.

These are testable hypotheses and they can be tested using federal National Incident-Based Reporting System (NIBRS) data, the only high-quality, multi-state crime data that include offender-victim relationship information. NIBRS is a relatively new data collection effort in which only a subset of states participates. Registration and notification laws were enacted primarily in the 1990s and only 15 states were contributing to NIBRS as of 1998: Colorado, Connecticut, Idaho, Iowa, Kentucky, Massachusetts, Michigan, Nebraska, North Dakota, Ohio, South Carolina, Texas, Utah, Vermont, and Virginia. Fortunately, these states are broadly representative in important relevant respects and NIBRS contains sufficient crime data both pre- and post-SORN laws in each of these 15 states to conduct reliable analysis.

The NIBRS offender-victim relationship variable has one geographic element (the offender was a neighbor), but it is otherwise organized by the level of familiarity that the victim has with the offender (e.g., the offender was a friend or acquaintance) or by family ties (e.g., the offender was a spouse or sibling). Offenses that may respond in discernible ways to registration and notification laws fall into three categories: “close” (involving family members, significant others, and friends), “near” (involving neighbors, acquaintances, and offenders otherwise known), and “stranger” crimes. Potential victims who are neighbors or acquaintances seem more likely to learn from and respond to an offender’s information being made public than do family members or complete strangers. Accordingly, notification may reduce the relative frequency of “near” offenses. Similarly, registered offenders may be more easily monitored when they are around their families and in their neighborhoods, and more easily located when a nearby crime occurs. Hence, registration ought to reduce the relative number of “close” and “near” offenses.

But even if registration and notification laws succeed at reducing the relative incidence of crimes against particular classes of victims believed to be especially vulnerable, one cannot assume that the overall recidivism rate—much less the sex offense rate across the board—has also fallen. This counterintuitive claim is true for at least two reasons.

First, sex offenders may respond to registration and notification by seeking out victims who are not protected by these laws. A registered offender who visits another jurisdiction or perhaps even another neighborhood will not be “known” there by either the public or the police. In theory, SORN laws may, as forensic psychologist Robert Prentky has observed, “accomplish nothing more than changing the neighborhood in which the offender looks for victims.” If the offenses that would have been committed against the now-informed or newly supervised potential victims are simply displaced onto “strangers,” overall recidivism levels will not change.

Second, notification may reduce the relative capacity or willingness of released sex offenders to live without crime by rendering them pariahs wherever they reside or work. As a consequence, even if committing crimes against neighbors becomes relatively less attractive, returning to crime may still become more attractive to released offenders on the whole, so much so that recidivism rates against all types of victims may actually increase.

Distinguishing Changes in Recidivism Levels from Deterrence Effects

If the only questions of interest were whether SORN laws reduce the total frequency of sex offenses and whether these laws influence the relative frequency of sex offenses committed against certain classes of victims, using interstate and intertemporal variation in these laws to explain changes in sex offense patterns across jurisdictions and over time would be an entirely suitable empirical approach.

But SORN laws are defended as a means of inhibiting or restraining released offenders from reoffending, not as a way of deterring potential first-time sex offenders (or non-registered offenders more generally) from committing sex crimes by threatening to enforce such laws against them if they do. This makes sense: reducing recidivism by providing potential victims with the identities and criminal histories of latent recidivists requires, by assumption and design, the collection and sharing of such information. By contrast, deterring offenders may be achieved in many other conceivable ways (such as by increasing prison sentences, perhaps), and one or more of these alternative strategies may prove to be more cost effective or superior on other grounds once all costs and benefits have been tallied.

Therefore, separately measuring the deterrence and recidivism effects of SORN laws is essential to any discussion of the general desirability of these laws (particularly in the case of notification, it turns out). Unfortunately, isolating these effects is difficult to do persuasively when working with comprehensive multi-state crime data because these data simply do not indicate whether a crime was committed by a registered sex offender or by an offender with no prior sex crime record.

This challenge has an intuitive solution, however. Although NIBRS data also lack information on whether crimes were committed by registered sex offenders or by “new” offenders, a simple fact bridges this gap: registration and notification laws cannot significantly affect the frequency of sex offenses by altering recidivism levels (by, for example, informing potential victims of “known” threats) when nobody or only a few convicted sex offenders are subject to these laws—i.e., when registries are close to “empty.” But, in theory, both registration and notification laws can reduce the frequency of sex offenses even when private and public registries are “empty” by deterring potential sex offenders who fear becoming subject to these laws in the future.

In application, this fact means that the significant differences across states and over time in the retroactive coverage of SORN laws can be used to distinguish their deterrence effects from their recidivism effects. This variation in coverage led to dramatic disparities in the numbers of affected offenders as these laws became effective: some states applied their laws only prospectively to individuals convicted or released after the relevant effective date; others made their laws retroactive, applying them to individuals convicted or released over a range of different pre-effective date time frames. As a result, as all states began to enforce their SORN laws, certain states had large catalogs of “registered” offenders while others had empty or nearly empty registries.

A straight-forward means of implementing this approach involves collecting data on the actual number of offenders on each state’s registry (i.e., registry size) at many different points in time and combining this information with county-level registry size data at a single point in time (August 2007) to estimate the registry size in each county for every month in the sample under the assumption that registry sizes grew smoothly once established.

Putting It All Together: The Many Consequences of SORN Laws

Using these ideas and a detailed coding of state SORN laws, NIBRS crime data, registry size information, and county-level economic and demographic data, the effects of SORN laws on overall and victim-specific sex offense frequencies—and whether these consequences, if any, resulted from changes in deterrence, recidivism, or both—can be estimated using standard econometric methods.

For all intents and purposes, these regression techniques use the timing and content of SORN laws and registry size information to explain changes in the number and type of sex offenses while at the same time accounting for alternative explanations that might generate spurious associations between these policies and fluctuating offense rates. The registry coverage data make it possible to estimate recidivism and deterrence effects separately, despite not knowing whether offenses were committed by registered offenders. Estimating victim-specific offense frequency effects informs the interpretation of any aggregate frequency estimates and offers a means of testing whether notification simply displaces crime onto unknown victims.

In these regressions, the outcomes of interest are the total number of sex offenses and the number of offenses against particular victim types as recorded by NIBRS reporting agencies by month. The key explanatory variables are:

• whether a registration law has been implemented (capturing the deterrence effect of a registration law),
• how many individuals are registered if a registration law is in place (registration’s effect on recidivism),
• whether a notification law has been implemented (the deterrence effect of notification), and
• how many individuals are registered if a notification law is in place (notification’s effect on recidivism).

All regressions include controls for when the crime occurred (to account for the potential influence of any secular crime trends) and for the agency that reported the offense (to account for any persistent heterogeneity in crime across reporting areas), as well as for annual per-capita income, unemployment levels, poverty rates, and the fraction of the population in five ethnicity categories and five-year age categories at the county level. Ordinary assaults and other non-sex crimes committed per 10,000 people are employed to correct for any broad changes in criminal justice policy or law and for related trends. As one would expect, sex offense rates are positively associated with both variables. Finally, regressions are weighted by an agency’s population coverage, meaning estimates reflect the average changes in risk faced by a typical person in the sample and are also adjusted for differences in crime variability that can result from large disparities in population levels.

What does all of this work tell us?

First, no evidence emerges that the threat of registration alone deters individuals from engaging in sex crime. In fact, the estimated impact of a new (empty) sex offender registry on the frequency of sex offenses is positive, although too small and imprecisely measured to be reliable. There is strong evidence, however, that requiring registration does reduce recidivism, presumably by increasing police monitoring and therefore increasing the likelihood of punishment that potential recidivists face. The data intimate that each additional sex offender registered per 10,000 people reduces the annual number of sex offenses reported per 10,000 people on average by 0.098 crimes (from a starting point of 9.17 crimes). This sizeable reduction (1.07 percent) buttresses the idea that we may be able to use law enforcement supervision to combat sex offender recidivism.

Second, with respect to notification, quite the opposite pattern unfolds. The threat of becoming subjected to a notification regime—and the shame and collateral consequences that accompany being publicly identified as a sex offender—appears to have a sizeable deterrent effect (i.e., it reduces nonregistrant offenses). The data suggest that notification laws deter on average 1.17 crimes per 10,000 people per year (approximately 12.8 percent of all sex crimes). But these same data also warn that the greater the number of released offenders that states actually subject to notification, the higher the relative frequency of sex offenses. In other words, the punitive aspects of notification may have unintended perverse consequences. All else equal, publicly revealing the identity and criminal history of a released offender seems to increase the likelihood of his returning to crime. These results are highly statistically significant: it is unlikely that existing notification laws are reducing recidivism among registered offenders, and it is distinctly possible that these laws are making things worse.

Sex offender notification policies may thus involve a difficult tradeoff: many potential offenders may be deterred by the threat of community notification and its associated costs, but the ex-post imposition of those sanctions on convicted offenders may make them more, rather than less, dangerous.

Notification laws appear most attractive when they apply only to small numbers of offenders, presumably the worst of the worst. According to the analysis, notification precludes an additional 0.78 yearly sex offenses per 10,000 people (p = 0.02) when 4.7 offenders per 10,000 people (only the 10th percentile of registry size) are registered. By contrast, adding a notification law to an existing registration requirement with an average-size registry has an insignificant positive effect on sex offense frequency (0.14 yearly sex offenses per 10,000 people; p = 0.65), although the overall impact of both laws is still a reduction of 1.06 sex offenses per 10,000 people per year (p = 0.11). These numbers are at odds with the idea that expanding the coverage of notification laws will reduce crime. Given the substantial social and individual costs of maintaining a large public registry, one implication of these estimates is that states should employ relatively narrow notification regimes in which most or all sex offenders are required to register with police, but only a small and carefully selected subset is subjected to notification.

Third, consistent with registration reducing recidivism, the data imply that large registries (without notification) make the commission of crimes against “close” and “near” victims—but not strangers—more difficult or otherwise less likely. The effect for victims of “near” offenders is highly statistically significant, whereas the estimate for victims of “close” offenders is slightly smaller in absolute value and is not statistically significant at conventional levels. The estimate for victims of “stranger” sex offenses is, in contrast, positive, though small and statistically inexact. These findings are at least compatible with the idea that registering sex offenders with law enforcement can preclude crimes against “nearby” victims while at the same time being of no consequence to strangers—i.e., neither protecting them nor making them more likely to be victimized through displacement.

Fourth, the evidence suggests that notification’s deterrence effects are similar in percentage terms across all victim types. This parity result is reasonable given that the nature of the penalty that notification imposes on individuals convicted of sex offenses is largely invariant to the type of victim involved.

Finally, notification is estimated to be recidivism-enhancing, and these effects are statistically significant and similar in size in percentage terms across all offender-victim groups. If notification laws were to make targeting local victims more difficult for a registered sex offender, then fewer “near” victim crimes should occur, along with little or no reduction in crimes against strangers (or even an increase in frequency in the case of displacement). Alternatively, if notification laws primarily make life outside of prison distressing for sex offenders and if notification laws fail to raise the relative cost of attacking certain victims, then crime rates should rise and be spread equally across all types of victims. The data support the latter scenario. Although notification may deter potential offenders from engaging in sex crime, prompting victims to protect themselves against known offenders may backfire in the end.

This discussion leaves out many details, robustness checks, and caveats that would accompany a full reporting of this empirical work. Highlights from this work include:

• An analysis of the effects of SORN laws on arrest rates reinforces the interpretation and reliability of the offense results.
• Placebo tests find no evidence of similar associations between SORN laws and crimes other than sex offenses, implicitly rejecting the possibility that unobserved variables or trends may drive these findings.
• Alternative samples and specifications of the regressions, including the employment of state-specific linear trends as well as a series of corrections to the NIBRS data, do not affect the substantive conclusions of this work.
• The possibility that shifts in victim reporting behavior may account for or explain these results can be shown to be inconsistent with other patterns in the data.

Conclusion

Sex offender registration laws appear to reduce the frequency of reported sex offenses, particularly when the number of registrants is large. Consequently, there may be significant social benefits to providing information about most or all convicted sex offenders to local authorities. An average-size registry decreases crime by approximately 1.21 sex offenses per year per 10,000 people, a 13 percent reduction from the sample mean. This drop in the sex crime rate benefits local victims (neighbors and acquaintances, as well as family members, friends, and significant others); there is no evidence that registration affects the frequency of sex offenses against strangers.

Notification regimes also seem to influence sex offense frequency, although not as lawmakers had intended. Notification laws reduce the number of sex offenses when the size of the registry is small, but these benefits dissipate as more offenders become subject to notification requirements. This finding accords with the idea that notification can deter nonregistered individuals, but that it encourages recidivism among registered offenders, perhaps because of the psychological, social, and financial harms they suffer from the public release of their criminal history and personal information. Indeed, for a registry of average size, instituting a notification regime has the aggregate effect in these data of increasing the number of sex offenses by more than 1.57 percent, with all deterrence gains more than offset.

The conclusion that notification appears to increase recidivism may seem counterintuitive. After all, the active identification of sex offenders through public registries is designed to alert potential victims to the threat a nearby released sex offender may pose to them or their households. But for community notification laws to succeed at reducing sex offender recidivism levels, two separate and nontrivial conditions must be met.

The first condition, not surprisingly, is that the public identification of sex offenders must make it more difficult, on average, for a released offender to commit a sex crime. This condition, in turn, requires that three underlying facts be true:

• At least some potential victims must be newly informed of the registered offender’s status by the operation of the notification law.
• These newly informed victims must be capable of reducing their exposure (or the exposure of others) to victimization by the offender, at least on average.
• It must be costly for the offender to identify any alternative victim who is either unaware of his status or who is unable to act on that information.

But as ample research demonstrates, offenders are rarely strangers to their victims. Friends and family members will often already be aware of a sex offender’s criminal history. When an offender’s target is a stranger, it may be hard for the potential victim to determine the offender’s status in a timely way or to reduce her exposure if she does. Even if the victim is able to reduce her exposure, a released offender who becomes at risk for reoffending may find himself surrounded by other potential victims, many or some of whom are likely to be oblivious to the threat he poses. Neighbors seem most likely to benefit from notification so long as they receive the information and are able to behave in ways that reduce their risks. Unfortunately, neighbors as a class make up less than 5 percent of sex offense victims.

The second condition is that notification must avoid aggravating those risk factors that can significantly increase an offender’s likelihood of reoffending. In fact, even if notification were to succeed at making it more difficult for released sex offenders to target and attack a victim, recidivism rates may still rise.

The difficulty of committing a crime is only one factor, among many, that affects an offender’s likelihood of recidivating. Publicly identifying an individual as a sex offender (as well as imposing other significant constraints—residency restrictions, frequent reporting requirements, etc.), however, also influences many other drivers of criminal behavior by dramatically changing a sex offender’s daily life, future prospects, and psychological and financial burdens. While a law that restrains a released offender has the potential to reduce recidivism if that law makes the commission of crime more difficult or if it mitigates various risk factors, such a law also has the potential to increase recidivism if it worsens those circumstances (e.g., unemployment, unstable housing) known to contribute to reoffending.

Notification regimes, with their attendant impositions, appear much more likely to increase the probability that released sex offenders return to crime, all else equal, than to reduce it. These laws and their implementation exacerbate a host of recidivism risk factors. Many compelling studies have established that publicly identifying sex offenders makes it more difficult for them to find employment and housing, residency restrictions make everything more expensive and life less stable, and both make it harder for registered offenders to be with or build their own families. Life as a registered sex offender, by all accounts, is simply much more difficult than the life of the typical former felon—in large part because of SORN and other post-release laws.

In an important sense, notification was and is intended to reproduce an upside of incarceration, essentially by “incapacitating” potential recidivists. Publicly identifying released sex offenders is an attempt to create a barrier between them and the public—i.e., their potential victims. Unfortunately, the analogy of notification laws to prison is too apt: publicly defining individuals as “sex offenders” reproduces many of the deprivations and burdens of prison in addition to partially recreating the incapacitation effect. Plus, by making the world outside of prison more like being in prison, the threat of sending someone to prison should he commit another sex crime is rendered much less severe. Put another way, the more difficult, lonely, and unstable our laws make a registered sex offender’s life, the more likely he is to return to crime—and the less he has to lose by committing these new crimes.

It is easy to see, therefore, that the effect on recidivism of notification laws (and of most sex offender post-release laws generally) is an empirical question: the effectiveness of these laws will depend on how they are structured and applied. If notification and its associated burdens make it more difficult for a registered sex offender to find victims, while at the same time not aggravating the risk factors known to lead to recidivism and not reducing a registered offender’s desire to avoid prison, then recidivism rates should drop. But if these laws impose significant burdens on a large share of former offenders, and if only a limited number of potential victims benefit from knowing who and where sex offenders are, then we should not be surprised to observe more recidivism under notification, with recidivism rates rising as notification expands.