class: center, top, title-slide # Rational Choice and Deterrence ## SOC371 ### Chuck Lanfear ### Jan 27, 2021<br>Updated: Jan 26, 2021 --- # Overview 1. Sherman & Berk (1984) 2. Routine Activity Theory 3. Extensions * Behavioral Economics * Situational Opportunity --- class: inverse # Sherman & Berk (1984) --- ### Minneapolis Domestic Assault Experiment Competing Theoretical Hypotheses * *Specific Deterrence*: Arrest will reduce reoffending * *Labeling*: Arrest will increase offending ??? This is nice case where two theories predict opposite effects Labeling says that once marked as an offender (primary deviance), you will alter self-image and increase offending (secondary deviance) Can use an experiment to adjudicate between them -- Treatments 1. Separate 2. Advise or mediate dispute 3. Arrest offender ??? Officers had randomized color coded notepads; were to treat based on sheet color. Nice setup as separation takes them out of home like arrest but without the actual arrest component Have to do something so there's no untreated (e.g. do not respond) -- Sample * Misdemeanor assaults * Victim and offender present * Probable cause for recent assault ??? Limited to cases where they are sure something happened but not dangerous to pick random treatment -- Measurement (Next six months) * Re-arrest records * Interviews with partners ??? Checked police records for re-arrest while simultaneously interviewing victims every two weeks Nice because two ways of verifying; might deal with unwillingness to report in future after initial arrest --- # Results Arrest had specific deterrent effect * Significantly reduced re-arrest and reported reoffending * No evidence for labeling effect * Does not appear to be incapacitation ??? About half as much recidivism under arrest as mediation Confident not incapacitation as there was no immediate reoffending in any treatment -- # Policy Implications Mandatory arrest * Arrest *required* if evidence of domestic violence * ~ 19 states in some form * Others *recommend* arrest * Some mutual arrest ??? Evidence provided strong support for mandatory arrest--seemed to be consistently reducing reoffending Policy enacted in many states subsequently, frequently citing this vein of research In other states common for arrest to be default position; some require both parties to be arrested if both claim other contributed --- # Issues Departures from randomization * Offenders refusing separation or mediation * Victims demanding arrest ??? Could not always apply target treatment; sometimes offender uncooperative or victim insisting on arrest This should be conservative bias, with arrest treatment disproportionately affecting most belligerent or severe; would expect weaker effect -- Follow-Up * Some victims could not be reached * Reduction in response over time ??? Common issues with reaching a disadvantaged, vulnerable population This is why they also used official records--self-report likely to catch everything but offense reports always obtainable -- Labeling * Most were past offenders already * Process may be slow ??? Possible labeling effects already took place Also possible takes longer for arrest to exert an effect -- *Need for replication* ??? Pretty robust study but issues combined with single context mean you need replication --- # NIJ Replications **Spouse Assault Replication Program** (SARP) Locations * Omaha, NB * Milwaukee, WI * Dade County, FL * Colorado Springs, CO * Charlotte, NC N = 4,032 ??? NIJ funded replications in 1981-1991 Larger sample size allowed for additional controls and examination if interactions Replication across many contexts increases generalizability -- Randomization by Dispatcher (e.g. Milwaukee) * Standard Arrest (held until morning or bail, ~ 11 hrs) * Short Arrest (release on recognizance, ~ 3 hrs) * Police Warning Only ??? Many randomizations, this is Milwaukee example. In total there were about 7 treatments. --- # Replication Results Effects * Inconsistent deterrent effect + Modest (<30%) effect where present * Prior records and age better predictors of reoffending * Most had no reoffending regardless of treatment + Small number had very high repeat offending ??? Initial examination of experimental results looked very mixed: Half beneficial, half negative When reanalyzed together with more power, careful statistical design, general moderate deterrent effect found Other individual factors had stronger effect on recidivism 80% of DV concentrated in around a tenth of those studied; this is a common sort of Pareto or power law distribution you see throughout crime and many other phenomena e.g. On average 80% of wealth in society is held by 20% of people e.g. 80% of crime occurs in 20% of locations; Iron Law of Troublesome Places or Law of Crime Concentration -- Conditional Hypotheses * Arrest deters married and employed * Arrest may increase offending for single and unemployed * Proposition: More to lose if married or working ??? Some evidence in analysis of experiments that deterrent effect more relevant for these groups This will come up when we talk about control theory -- Mandatory arrest may be ineffective or counterproductive * Other recent evidence seems to support this * May be better to target serious offenders --- class: inverse # Routine Activity Theory --- # Context Through 1970s, most theories of crime focused on individual characteristics, socialization, and social structures * Motivation and decision-making * Social learning * Subcultural theories * Social organization of neighborhoods -- These explain cross-sectional variation between individuals and groups well They predict changes in nationwide and area crime trends poorly ??? To a large degree this is a limitation of data More practical to study offenders and crime in specific locations and in cross-section Analyzing large-scale trends is hard because you need to observe many places over long period of time Effects may also have long lag--might be bad conditions a decade or more ago -- Cohen & Felson: (Potential) offenders are only part of the story. * Regardless of motivation, crime requires an **opportunity** ??? This is not a new insight--they note lots of work recognizing this for ages Contribution here was creating an analytical framework that clarified how these things are related So elegant it seems obvious; immensely impactful and one of the few conceptions in criminology that approaches a law --- # Opportunity Predatory crimes *require* elements to converge in time and location--criminogenic *situations* ??? Possible for things to overlap in location but not time, in time but not location; -- 1. Likely offenders ??? We know a lot about motivation of individuals and contexts that generate likely offenders -- 2. Suitable targets ??? Suitable targets is often taken as given or unlimited in many theories But suitability *varies*! Some things easier to take, worth more. And changes in society and technology impact this. -- 3. Absence of capable guardians ??? Obvious that crime doesn't occur when police are watching But guardianship is complex and pervasive--we are guardians of our property but also targets ourselves of some crimes Different contexts and behaviors facilitate or impede guardianship -- *Aggregate* changes in these can explain crime rates *Social and technological changes in routine behavior can produce drastic changes in crime* ??? Key insight here is that changes in *any* of these can change crime -- * More high-value portable goods or conspicuous consumption * Less time spent at home, more at night and at drinking venues * Dual income households--less supervision of home, neighborhood, and children --- # Routine Activity Theory .image-full[  ] ??? This is simplified. Note that theory requires all three components to be present. In a situation, if any one of these is missing, no crime can occur. No many how many suitable unguarded targets there are, no crime will occur if there are no offenders. But in aggregate, independent changes in amount of one factor changes quantity of crime --- # Levels of Explanation Rational choice theory is micro-level * Describes individuals * Describes situations -- Routine activity theory is macro-level * Describes groups * Describes trends -- Rational choice is often used as individual-level foundation for routine activities * In aggregate, people behave rationally * More suitable targets means greater rewards * Fewer guardians means lower costs -- *Rational choice and routine activity suggest crime can increase without change in motivation or number of offenders* --- # Policy: Hot Spots Policing Crime requires offender, target, absence of guardians * Hard to change motivation * Hard to harden many targets * Police (guardians) can be redirected easily ??? We've been trying to act on motivation forever--it is real hard, has structural roots Hardening targets requires people to change their behavior. Sometimes works, but generally want to avoid putting burden on people. Both because it means those most capable of protecting will benefit most and also because people won't do it Individual target hardening helps individuals most: If you're being chased by a bear you don't need to be fastest, just not the slowest. Locking your bike isn't about preventing bikes from being stolen. It is about getting them to steal someone else's bike instead. -- Identify high crime areas, send police there * Assumption: Police can act as capable guardians * Assumption: Criminals will not relocate (displacement) -- Effects * Good evidence for crime reduction * Not much evidence for displacement + High crime = high opportunity + Nearby locations may lack opportunity * Long-term and large-scale effects uncertain ??? Experiments targeting high crime areas show substantial effects There were worries crime would go up nearby--displacement, but not much evidence Focusing on hot spots for long period might have declining effectiveness Effect of targeting one hot spot won't be same as effect of targeting all at once--policy may not scale This is an important point about all experiments and policies: Effects at small scale may look very different at large scale. --- class: inverse # Rational Choice and Routine Activities ## Extensions --- # Behavioral Economics Pogarsky, Roche, & Pickett (2018) "Offender Decision-Making in Criminology" ??? Behavioral econ is about systematic departures from rationality Emerged from cognitive psych contributions to econ research -- * Bounded rationality ??? Short & Strodtbeck (1964) had handle on this; not really part of behavioral econ exactly We often have very little information when acting; can't assume perfect info Also rule out many possible decisions immediately and consider a few options -- * Prospect Theory (Kahneman & Tversky) + People risk averse in gains, risk seeking in losses + Underweight highly likely, overweight unlikely events + Expectations matter ??? Strong evidence that gains/losses to crime are treated differently if a gain vs a loss Nonlinearity in effects Criminals likely to focus on easy guaranteed gains, but will prefer risk of larger loss to smaller certain loss, ex. resisting arrest For same level of punishment, expectations shape later recidivism: Expecting worse, punishment deters less. -- * Dual-Process Models + Much behavior is intuitive and rule (heuristic) based + Rational, deliberative behavior is costly ??? People generally like to avoid effort; rational choice thinking takes lots of effort Most behavior is habitual and intuitive If you want to act on rational choice, need to trigger deliberative system; evidence those using deliberation more are also averse to crime -- * Heuristics and Biases + Shortcuts in uncertain environments + Replace hard question with easier one ??? May make large mistakes in reasoning by applying shortcuts to new situations --- # Situational Opportunity Routine activities focuses on aggregates * Not concerned with what micro-level theory is used * Suggests importance of opportunities to micro-level ??? Some theories take opportunity as a given--there's always some crime that can be committed -- Situational opportunity is a "meso" theory of *places* * Weisburd et al.'s *Criminology of Place* * St. Jean's *Pockets of Crime* * Crime Prevention Through Environmental Design (CPTED) * Place Management Theory ??? These are approaches based on considering attributes of relatively small areas like intersections or street segments Characteristics of places can have large influence on crime Interaction between offenders, guardians, and places Offenders want to use high opportunity places, may conflict with people who live or work there -- Multi-level theory * Micro-level behavioral theory (ex: rational choice) * Meso-level place theory (ex: situational opportunity) * Macro-level structural theory (ex: social disorganization) ??? Some approaches link theories at different levels together into micro-macro models Lot of this emerges from Coleman's *Foundations of Social Theory* approach --- .image-tall[  ] ??? Example of a multilevel framework with rational choice model of individuals, situational opportunity model of places, social disorganization / CE model of neighborhoods Neighb's influence place opportunities which influence individual actions Crime rate for neighborhood is aggregate of street-level crime rates which are result of individual criminal actions --- class: inverse # Questions --- # Exam * Exam will go out today. * Due 11:59 PM Saturday + Don't spend the entire time on it! + Email me if you have time concerns * Email if clarifications are needed + I may may forward clarifications to entire class --- # For Next Time * Blumstein, Alfred, and Jacqueline Cohen. 1987. “Characterizing Criminal Careers.” *Science* 237:985-991 Things to pay attention to: * Different insights gained from considering trajectory of individuals versus aggregate crime rates * Participation vs. Frequency + Different factors may predict these! * Difference between age-crime curve and criminal career trajectories ??? This article is challenging and math heavy, but short Focus on insights, don't dwell heavily on the methods but do read them, see if you can get intuition