L21: Risk Perception, Risk Taking and Risk Compensation

Affect heuristic

"bad" negative emotive events = higher risk

Risk acceptance: Willingness to accept risks will depend on...

- Expected benefits of risky behaviour alternatives e.g., time saved by speeding, thrill seeking - Expected costs of risky behaviour alternatives e.g., speeding tickets - Expected benefits of safe behaviour alternatives e.g., safe driver insurance discounts - Expected costs of safe behaviour alternatives e.g. peer disapproval, time loss

Gender differences in risk taking Meta-analysis of 150 studies: Largest effects for (3)

- Physical skills - skill games, e.g. shuffleboard, ring tossing - Risky experiment - participating in experiments with potential for harm - Intellectual risk taking - trying harder maths/spatial problems

Risk taking and personality UK study: Personality predictors of overall risk taking:

- Positive relationship with Extraversion, Openness - Negative relationship with Neuroticism, Agreeableness, Conscientiousness - Similar patterns for all domains, with varying effect sizes (except Openness was not a significant predictor for Safety domain)

Risk taking and personality US study of ~3,000 college students: Correlations between domains:

- moderate correlations between all of drinking, smoking, drugs and sex - small-medium correlations between gambling and drinking, sex (esp M) - small correlation between driving and drinking

Cultural Theory: Egalitarian

Both similarities and strong bonds between people (voluntary helping) Risks: technology and environment

Cultural Theory: Fatalists

Differences between people, little bonding Risks: none of the above

Risk taking and personality US study of ~3,000 college students: Personality scales (5)

Impulsive Sensation Seeking (ImpSS), Neuroticism-Anxiety, Aggression-Hostility (Agg-Host), Activity, Sociability

What are the four categories of people in cultural theory?

Individualists Hierarchists (Collectivists) Egalitarian Fatalists

Cultural Theory: Individualists

People are similar, but not obligated to assist each other (dismissive of central authority) Risks: war and other threats to commerce

Predictors of risk perception

Risk perception is complex and multi-faceted, many aspects of it: 1. Objective risk 2. Heuristics and biases 3. Risk target 4. Psychometric properties of threat (dread, new/old) 5. Culture 6. Individual differences in risk sensitivity, attitudes and specific fear

Cultural Theory: Hierarchists (Collectivists)

Strong bonds between people, but variations between people - social hierarchies Risks: law and order

Representativeness heuristic

decisions based on categories (e.g., this is the type of person who would die of lung cancer)

Gender differences in risk taking Meta-analysis of 150 studies: Smallest effects for

drinking drug use sexual activities

Risk taking and personality US study of ~3,000 college students: Risk taking (6)

drinking, smoking, drugs, sexual behaviour, driving, gambling

White male effect Swedish study

found no white male effect, but differences between native vs. foreign born respondents - suggests social inequality plays a key role

Gender differences in risk taking Meta-analysis of 150 studies: Small/moderate effects for (3)

gambling driving physical activities (climbing banks, balance beams, donkey riding)

Availability heuristic

people judge risks as more likely if they are easy to imagine

Anchoring heuristic

people start with known information ("anchor") and adjust it (insufficiently/inappropriately) to estimate unknown risk

Gender differences in risk taking Meta-analysis of 150 studies: No effect for

smoking

Helmets and crashes: case-control study

• 22,814 cyclist crashes in Spain during 1990-1999 • Voluntary helmet use - overall frequency 13% • Helmet use negatively correlated with non-speeding infractions (i.e., rule breakers less likely to wear helmets) • Does not support risk compensation - but cannot necessarily contradict it • Because helmet use is voluntary, participants self-select into helmet group - may have pre-existing differences (lower risk tolerance) that obscure risk compensation effects

Risk compensation: economic perspectives

• Before Wilde popularised risk homeostasis theory, Peltzman made a similar point from an economic perspective • Vehicle safety improvements yielded smaller than expected improvements • Trade offs - if we have more of a resource (safety) than we require, we can trade some to gain another benefit (e.g., time saved)

Risk homeostasis theory supporting evidence (5)

• Cars with ABS driven in riskier style (faster, shorter headway, later braking) - offsets benefits of ABS • Drivers adopt more "aggressive" driving style with airbags • Higher travel speeds on safer roads - e.g. better lighting, less traffic • Childproof medicine bottles did not reduce accidental poisoning - parents became less careful in storage • Negative effects of road safety training for children and skid training for novice drivers

Psychometric model

• Classic study - 76 participants rated risks, benefits and nine characteristics of risks: 1. Voluntariness 2. Immediacy 3. Known to exposed 4. Known to science 5. Controllability 6. Newness 7. Chronic 8. Common/Dread 9. Severity of consequences • Used factor analysis to group characteristics in two dimensions:newness and dread • Psychometric model accounts for 20-30% of variance in risk perception

Cultural Theory

• Cultural Theory of risk perception argues there are four categories of people (cultures), who are especially sensitive to different hazards/risks: • Little empirical testing of this theory, but suggests these general cultures and values predict <10% of variance in risk perception

White male effect and cultural differences

• Demographic differences in risk perception arise from cultural differences • For gun risks, interaction between demographics and worldview i.e. rate guns as less risky if white male is hierarchical or individualist

Hedlund's criteria for no risk compensation

• Don't know it's there • It doesn't affect them (i.e., don't feel safer) • No reason to change their behaviour (i.e., safety was not a factor influencing behaviour to begin with) • Behaviour is tightly controlled by other factors Relevant to bicycle helmets? No Maybe Maybe No

Heuristics and biases: old and early work

• Early work suggested heuristics influenced risk perception • More recent work has shown risk perception to be multifaceted • These heuristics influence probability judgements, which is one of multiple factors that determine risk perception

Helmet law evaluation: NZ

• Evaluation of NZ data for exposure, injuries and deaths for cyclists and pedestrians, 1989-2009 (helmet legislation introduced 1994) • Both cycling and walking decreased - cycling to a greater extent • Pedestrians: large reductions in deaths and injuries per hours walked • Cyclists: slight increases in deaths and overall injuries per hours cycled, decrease in serious injuries • Risk ratio, ratio of cyclist:pedestrian crashes increased ... NZ's helmet laws "failed in aspects of promoting cycling, safety, health, accident compensation, environmental issues and civil liberties"

Bicycle helmets - final thoughts

• Evidence for risk compensation is mixed • Mandatory cycling laws in NZ and Australia were associated with decreased cycling - so any changes in injury rates are confounded • The key question is whether helmet laws reduce head injuries - but in NZ there were not many to begin with (small N, hard to show effect) • European studies suggest underlying differences between cyclists who do vs. do not wear helmets • Some experimental studies find risk compensation (e.g., riding faster with vs. without helmet) - but not all do

White male effect Survey of 1844 Americans

• Hypothesised white male effect is an artefact of cultural worldview • Rated risk of abortions, guns and environment much lower than other groups • Individualist vs. Communitarian • Egalitarian vs. Hierarchical

Risk taking and personality US study of ~3,000 college students: Correlations between personality scales and domains

• ImpSS correlated with drinking, smoking, drugs, sex • Soc correlated with drinking, gambling; drugs & sex (M); smoking (F) • Agg-Host correlated with drinking; driving (M); smoking & sex (F)

Conditions against risk compensations

• Individuals will not compensate for a safety measure if: - They don't know it's there - It doesn't affect them (i.e., I don't feel safer) - They have no reason to change their behaviour (i.e., safety was not a factor influencing their behaviour to begin with) - Their behaviour is tightly controlled by other factors

Objective risk

• Laypersons' subjective risk perceptions are positively correlated with objective and expert assessments • But there are some discrepancies (remember yesterday's lecture!) • Most accurate when: - individuals have knowledge of the objective risks (direct or indirect) - individuals are rating the general risk to others, not themselves • Accurate: people have a general sense of what is more/less risky • Causes of death • Road features (other than intersections) • Inaccurate: tendency to under- or over-estimate specific risks Key issue: are there systematic differences that we can predict? • What are these predictors? • Nature of the risk? • Individual differences?

Helmet law evaluation problems with evaluations: (2)

• Low quality evaluations • Lack of control groups or appropriate comparisons

Risk taking and personality UK study: Other effects seen (3)

• Males higher than females on Finance, Safety, Recreational, Health risks • Risk taking declined with age for all risk taking domains - Largest effects for Finance, Recreational, Social risks • Sensation-seeking (a facet of Extraversion) was especially important

Personality and sexual risk taking

• Meta-analysis of 53 studies on personality and sexual risk taking • (Impulsive) Sensation Seeking [SS] most studied (64% of effects studied) • SS, ImpSS and Impulsivity consistently positively associated with risk taking • Five factor model: Extraversion positively associated, Agreeableness and Conscientiousness negatively associated • Aggression-hostility and neuroticism-anxiety had larger effects than ImpSS - but only based on 1 study (extreme sample) • Results suggest complex association between risk taking and risk appraisal - high SS --> risk taking --> ↓ risk appraisal

What about other road users?

• Naturalistic study of driver/cyclist overtaking interactions • Researcher cycled 320km within two UK cities (Salisbury and Bristol) • Varied appearance, attire • Passing distances were smaller when the cyclist wore a helmet • Remember safety is a complex system - interactions are important!

Risk target

• Optimism bias: we rate our own personal risk differently than general risk. Large discrepancy between how we rate ourselves and others • People exhibit risk denial or unrealistic optimism - own risk are lowest - general risk are highest - ratings for family and friends in between • Extent of risk denial depends on the specific risk (perceptions of whether it can be controlled?)

Risk compensation experiment

• Oslo field study of cyclist behaviour with and without helmet - 19 normally wear a helmet, wore own helmet for study - 16 do not normally wear a helmet, borrowed a helmet for study • Cycled the same route with and without helmets • Borrowed helmet group showed no difference in pace, HR, discomfort, excitement, insecurity, or accident likelihood, with vs. without helmet • Own helmet group showed decreased pace, increased insecurity, accident likelihood, and pNN10 (short intervals between heart beats) when not wearing a helmet • Suggests degree of long-term risk compensation

Risk homeostasis theory

• People have a target level of acceptable risk (preferred risk) • They assess perceived risk at each point in time • If perceived risk differs from acceptable risk, they adjust their behaviour

Criticisms of risk homeostasis theory

• Plausible in some cases but not a universal theory of behaviour • Selective evidence - several studies suggest airbags reduce death, childproof medicine bottles have prevented deaths • Airbags reduce specific injury types in event of crash but there are other disincentives that should deter crashing (e.g. property damage) • Unlikely that individual behaviour is influenced by population-wide injury and accident statistics (which Wilde claims) • Many safety improvements have had large gains - overall improvements in vehicle standards, seatbelt laws, motorcycle helmets

Individual differences

• Previous proposed factors cannot fully explain risk perception • Sjöberg (2000) suggests adding three individual difference predictors: 1. Attitudes - specific to the risk (i.e., perceived benefits) 2. Risk sensitivity - general tendency to perceive things as risky 3. Specific fear - associated with the risk, e.g. fear of radiation • These variables combined can predict up to 60% of variance • Note that attitudes and specific fear could overlap with other proposed predictors of risk perception

Risk compensation

• Risk compensation is a special variant of behavioural adaptation i.e., changing behaviour in response to perceived risk levels • Passionately debated due to implications - "do no harm" • Has been used as argument against - Seatbelt laws (but evidence supports restraint use) - Motorcycle helmet laws (but evidence supports helmet use) - Driver training (more likely overconfidence than risk compensation) - Bicycle helmets - hmm, let's look at that...

Take home messages

• Risk perception is multi-faceted - depends on objective level of risk, characteristics of the risk, and individual differences - "White male effect": some social groups (especially white males in US) rate risks consistently lower • Risk taking varies according to risk domain, personality, culture • Risk homestasis theory posits that people have a "target" level of acceptable risk - if you increase safety, they will compensate by engaging in more risky behaviour; if you decrease safety, they will be more cautious • Inconsistent evidence for risk homeostasis and risk compensation - more plausible in some contexts than others

Psychometric model criticisms (3)

• Some results don't entirely fit the claims - e.g., nuclear power is no longer new, and is common in some countries (Sweden), but opposition remains • Conceptual issues with interpreting the factor analysis - factors do not necessarily predict risk to the same extent - other factor structures may fit the data - other data sets yield different factor structures • There may be other attributes not covered in the scales that influence risk - e.g., unnatural risk

Helmet law evaluation 2009 Cochrane Systematic Review:

• Studies from USA and Canada with partial laws - e.g., under 18yo only (use adults as a control group) • Found helmet laws significantly increased helmet use, decreased brain injury • Did not assess impact on cycling rates

White male effect Study of 1512 Americans (Flynn et al., 1994)

• Study of 1512 Americans (Flynn et al., 1994) • Environmental health risks rated on scale from 1 (little or no risk) to 4 (high risk) • Among non-white respondents, men and women gave similar ratings • White men gave consistently lower ratings than white women and non-whites • This group were also on average better educated, more politically conservative and had higher household incomes

Helmets and cycling behaviour: survey

• Survey of ~1400 Norwegian cyclists aged 16-65 • Always use helmets: 48% females, 58% males • Helmet use did not predict accidents, but fast cycling did • ...but fast cycling was correlated with helmet use • Helmet use was correlated with other equipment use • Suggests different types of cyclists -> indirect effect of helmets on crash risk

Risk taking and personality UK study: Six risk domains (and what did compare with?)

• UK study N = 2041, Five Factor Model of personality and six risk domains: Recreational Health Career Finance Safety Social

Risk taking domain

• Who actually takes risks? It varies depending on the risk domain... • Correlations between specific risk domains range from small/none to large Variation suggests risk taking domain specific

Bicycle helmets Arguments for: (1)

• head injuries are less severe = potential to save lives, reduce brain injury

Bicycle helmets Arguments against (3)

• risk compensation • mandating helmets deters cycling • helmets don't fully protect against brain injury