Abstract
Die Messung von Kriminalitätsfurcht ist Gegenstand der vorliegenden Arbeit. Warum ist dieses Thema gesellschaftlich und wissenschaftlich relevant? Die Kriminalitätsfurcht der Bürger ist eine Größe, mit der Politik gemacht wird, die aber oft auf methodisch unsicheren Beinen steht. Diese Arbeit untersucht die Reliabilität und Validität der Messinstrumente, die in der Sozialwissenschaft zur Erfassung dieses zentralen Konstrukts eingesetzt werden.
Introduction
In the realm of criminology and public policy, "fear of crime" is a pivotal metric used to justify legislative changes, police resource allocation, and urban planning. However, as Frauke Kreuter demonstrates in her 2002 methodological treatise, this metric is frequently misunderstood and poorly measured. This paper investigates the reliability and validity of survey-based measurements of fear, arguing that reported public anxiety is often as much a product of survey design—question wording, order, and interviewer behavior—as it is a reflection of actual safety concerns.
The Fear of Crime Paradox
A central motivation for this research is the "Fear of Crime Paradox": the recurring observation that demographic groups with the lowest objective risk of victimization—such as the elderly—often report the highest levels of fear. Kreuter suggests that this paradox may not be a psychological reality but rather an artifact of measurement error.
Because fear of crime is a "latent construct"—something that cannot be observed directly—researchers must rely on survey indicators. These indicators are prone to several systematic errors:
- Non-attitudes: Respondents may provide answers to questions about topics they have not previously considered, constructing an opinion on the spot.
- Vague Terminology: The inconsistent use of terms like "fear" (a specific reaction to a threat) versus "anxiety" (a general state of unease).
- Context Effects: The influence of preceding questions in a survey on a respondent's mindset and subsequent answers.
Methodological Framework
To deconstruct these errors, Kreuter employs a sophisticated multi-method approach using data from the DEFECT project, alongside other major German datasets like the ALLBUS (German General Social Survey) and the Konstanz Victimization Survey (KVS). Her methodology includes:
- Experimental Survey Design: Utilizing "vignettes" (short stories) to test how different contexts change respondent reactions.
- Split-Ballot Experiments: Systematically varying question wording (e.g., "Furcht" vs. "Angst") and the order of questions to measure "questionnaire effects."
- Reaction Time Measurement: Using response latency to determine how "available" or deeply held an attitude is; shorter reaction times usually indicate a more stable, genuine opinion rather than a non-attitude.
- Interviewer Effect Analysis: Applying multilevel modeling to isolate how much variance in the data is caused by the individual interviewer rather than the respondent.
Key Insights and Results
Kreuter’s analysis reveals that "fear of crime" is highly sensitive to the survey environment. One of the most significant findings is the impact of interviewer variance. A substantial portion of the recorded "fear" in certain surveys was found to be interviewer-driven, meaning specific interviewers systematically elicited higher or lower fear responses. This introduces significant bias when making regional or demographic comparisons.
Furthermore, the research highlights that:
- Wording Matters: The choice between "fear" and "anxiety" significantly shifts reported levels of concern.
- Stability: Through test-retest experiments, the work demonstrates that "fear of crime" is less stable over time than many other social attitudes, suggesting it is highly sensitive to recent media reports or immediate environmental cues.
- Vulnerability vs. Risk: Fear is often a proxy for a sense of general social vulnerability rather than a calculated assessment of the probability of a specific crime occurring.
Conclusion and Future Directions
The study concludes that the field of criminology lacks a "gold standard" for measurement, making it difficult to compare studies across different countries or decades. Kreuter suggests moving away from vague questions about general "fear" and toward more concrete measures of perceived risk and the subjective costs of victimization.
While the 2002 work focused on CATI (Computer-Assisted Telephone Interviewing) and face-to-face surveys, its lessons remain vital for the modern era of digital surveys, where non-sampling errors continue to challenge the validity of social science data.
Summary updated