Case study 01 Master's thesis · Hunter College, CUNY · 2026

Fear of Crime, Intolerance of Uncertainty & Trait Worry

A mediation model of perceived risk and victimization concerns

Sample
538 adults · Trinidad & Tobago
Methods
Regression · Mediation · SPSS
Timeline
Jan 2025 – Apr 2026
Role
Sole researcher
Survey research Quantitative analysis Mediation analysis Cross-cultural Behavioral data Mixed methods
01

Context & research question

Fear of crime doesn't follow crime statistics. People living in the same neighborhood report very different levels of fear. Communities with falling crime rates often remain highly fearful. This disconnect suggests that psychological factors — not just environmental conditions — shape how people assess threat and respond emotionally.

This study asked: do stable personality traits — specifically trait worry (a tendency toward repetitive, future-oriented negative thinking) and intolerance of uncertainty (IU, a disposition to react negatively to ambiguity) — influence how much people fear crime? And critically: does this influence operate by changing how risky people cognitively perceive their situation, or does it bypass that step and influence fear more directly?

Primary question: Does perceived risk of victimization mediate the relationship between psychological vulnerability and fear of crime — and does this process work the same way across different cultural groups?

The Trinidadian context was significant: a nationally diverse population with two large ethnic groups (Afro-Trinidadian and Indo-Trinidadian) allowed for a meaningful test of whether the same psychological process operated differently across sociocultural contexts — a question rarely examined in fear of crime research.

02

Approach

The study used archival data from a nationwide cross-sectional survey conducted in Trinidad and Tobago — administered face-to-face in participants' homes by trained interviewers using a standardized questionnaire.

538
Adult participants, nationally stratified by age, gender, ethnicity & region
413
Used in subgroup analyses (Afro-Trinidadian n=181, Indo-Trinidadian n=232)
5,000
Bootstrap resamples for bias-corrected mediation confidence intervals
Why quantitative survey
Testing a theoretically specified model at scale
The research question required testing specific directional pathways between psychological constructs across a large, diverse sample. Quantitative survey design was the only method capable of doing this with statistical power adequate to detect mediation effects.
Why subgroup analysis
Aggregate data masks group-level differences
Analyzing Afro- and Indo-Trinidadian participants separately allowed the study to detect whether the psychological mediation pathway was universal or culturally contingent — a question that aggregate-level analysis would obscure.
Why Hayes' PROCESS macro
Robust, non-parametric indirect effects
Bootstrapped confidence intervals don't require assuming a normal distribution of the indirect effect — making them more appropriate for mediation testing than traditional approaches like the Sobel test.
Why personal vs. property distinction
Crime type affects fear differently
Research consistently shows that violent/personal crimes elicit more fear than property crimes. Running parallel models for both allowed for a more nuanced picture of where and how psychological traits exert their influence.
03

Execution

Four validated instruments were used, each split into personal crime and property crime subscales. All scales demonstrated strong to excellent internal consistency.

Scale reliability (Cronbach's α)
Personal Fear scale (16 items)α = .949
Personal Risk scale (11 items)α = .947
Property Fear scale (5 items)α = .881
Property Risk scale (5 items)α = .890
Trait Worry (Penn State, adapted)α = .865
Prospective IU (IUS-12)α = .820
Inhibitory IU (IUS-12)α = .813

Data preparation in SPSS included reliability verification, reverse scoring of 5 trait worry items, Pearson correlation matrices to screen for multicollinearity, and composite score construction by averaging validated subscale items. Mediation was then tested using Hayes' PROCESS macro Model 4 with 5,000 bootstrap samples and 95% bias-corrected confidence intervals — run separately for Afro-Trinidadian and Indo-Trinidadian subgroups.

04

Findings

Finding 01

Perceived risk was the strongest and most consistent predictor of fear — by a wide margin

Personal perceived risk correlated strongly with personal fear (r = 0.555, p < .001). Property risk correlated strongly with property fear (r = 0.509, p < .001). Personality traits, by contrast, had far weaker associations with perceived risk (r = 0.067–0.121). The data clearly placed cognitive threat appraisal — not personality — as the primary driver of emotional fear responses.

Correlation strength with fear outcomes
Perceived personal risk → personal fearr = 0.555
Prospective IU → personal fearr = 0.195
Trait worry → personal fearr = 0.156
Trait worry → personal riskr = 0.093
Finding 02

Psychological traits influenced fear directly — not through perceived risk

Prospective IU (the tendency to fixate on uncertain future outcomes) was a significant direct predictor of both personal fear (β = 0.137, p = .035) and property fear (β = 0.142, p = .031) in regression models. But it did not significantly predict perceived risk. This matters: people high in prospective IU don't necessarily believe they're more likely to be victimized — they simply feel more afraid. The emotional and cognitive responses were decoupled. Trait worry and inhibitory IU showed weaker or non-significant patterns throughout.

Finding 03 · Most significant

The mediation pathway worked for one ethnic group — but not the other

This was the most important and unexpected result. Among Afro-Trinidadian participants, perceived risk fully mediated the relationship between all three psychological traits and fear of crime — indirect effects were statistically significant across both personal and property crime models. Among Indo-Trinidadian participants, no mediation was found; all confidence intervals crossed zero. Fear was equally present in both groups, but the cognitive pathway producing it was fundamentally different.

Mediation results: perceived risk as mediator
Afro-Trinidadian
Mediation supported
Prospective IU indirect effect = 0.073, 95% CI [0.034, 0.111] — CI excludes zero. Psychological traits → perceived risk → fear pathway confirmed.
Indo-Trinidadian
No mediation
Prospective IU indirect effect = −0.001, 95% CI [−0.028, 0.028] — CI crosses zero. Psychological traits did not significantly influence perceived risk in this group.

This finding suggests that the fear of crime is not a singular psychological experience. Cultural, social, and contextual variables determine how individuals assess and emotionally respond to potential threats — a result that challenges universalist models of fear.

05

Impact & recommendation

The practical implication of this research is that interventions designed to reduce fear of crime — whether public safety campaigns, community programs, or policy design — should not assume a one-size-fits-all cognitive model. The mechanism through which psychological vulnerability produces fear differs across groups, and so must the approach to addressing it.

What this means in practice

For groups where the cognitive mediation pathway is active, targeting perceived risk directly — through accurate crime information, neighborhood transparency tools, or uncertainty-reduction messaging — may reduce emotional fear.

For groups where fear is not cognitively mediated in the same way, interventions focused purely on changing risk beliefs are unlikely to work. Culturally specific frameworks — drawing on community structures, social trust, or alternative threat-appraisal models — are more likely to be effective.

More broadly: any research or product that attempts to model how users perceive and respond to risk should account for the possibility that the psychological process itself differs by cultural background. Aggregate-level findings can mask the most important variation.

06

Reflection

What I'd do differently with more time or resources:

Add longitudinal design. The cross-sectional design limits causal claims. A longitudinal study would let me test whether psychological traits actually precede changes in perceived risk and fear over time, establishing temporal ordering rather than just correlation.

Measure media exposure directly. Media exposure was flagged in the literature as a likely amplifier of risk perception but wasn't captured in this dataset. I'd want to test whether high-IU individuals are disproportionately sensitive to crime-related media — a practically important moderator for any intervention design.

Pre-register the subgroup hypothesis. The ethnic group finding was the most important result, but it was exploratory. I'd pre-register this hypothesis in future research to strengthen the confirmatory claim and reduce the risk of capitalizing on chance.

Pair survey data with behavioral indicators. All measures were self-report. Adding behavioral data — avoidance behaviors, mobility patterns — would provide convergent validity and make the fear construct more actionable for applied audiences.