1.3 Research Methods in Criminology

Research Methods in Criminology

Criminologists use a range of research methods to study crime and criminal behavior. Surveys, experiments, and field studies each offer different strengths, but they also come with real trade-offs. Knowing how these methods work helps you evaluate published research and understand why criminologists choose one approach over another.

Beyond choosing a method, researchers also need to analyze their data and assess whether their findings are actually valid. This section covers both: the major research designs and the statistical and evaluative tools used to make sense of the results.

Research Methods in Criminology

Research methods in criminology

Surveys collect data from a sample of individuals using questionnaires or interviews. They're the most common method in criminology because they can reach large numbers of people relatively cheaply.

• Questionnaires can be self-administered (the respondent fills them out alone) or interviewer-administered (a researcher asks the questions directly)
• Interviews vary in structure:
  • Structured interviews use predetermined questions asked in a fixed order
  • Semi-structured interviews follow a general guide but allow the researcher to probe deeper on certain answers
  • Unstructured interviews are open-ended conversations that let the subject direct the discussion

Experiments test causal relationships by manipulating an independent variable and measuring its effect on a dependent variable. They're the strongest design for establishing cause and effect.

• Laboratory experiments take place in controlled environments where researchers can isolate the effects of specific variables. For example, a researcher might show participants different sentencing scenarios and measure their perceptions of fairness.
• Field experiments manipulate variables in real-world settings. The Minneapolis Domestic Violence Experiment (1981–1982) is a classic example: police randomly assigned one of three responses (arrest, separation, or mediation) to domestic violence calls to test which best reduced future offending.

Field studies involve observing and collecting data in natural settings, without manipulating any variables.

• Observational studies can use participant observation (the researcher actively engages with the group being studied) or non-participant observation (the researcher watches from a distance without interacting)
• Ethnographic studies require researchers to immerse themselves in a specific culture or community over an extended period. Sudhir Venkatesh's work embedded with a Chicago gang is a well-known criminological example. These studies produce deep understanding of social dynamics but take significant time and access.

Strengths vs. limitations of research designs

Data Analysis and Interpretation

Statistical analysis for criminological data

Once data is collected, researchers use statistics to make sense of it. There are two broad categories.

Descriptive statistics summarize and describe data without drawing broader conclusions:

• Measures of central tendency tell you what's typical in your data:
  • Mean: the average (add all values, divide by the number of values)
  • Median: the middle value when data is ordered from lowest to highest
  • Mode: the most frequently occurring value
• Measures of dispersion tell you how spread out the data is:
  • Range: the difference between the highest and lowest values
  • Standard deviation: the average distance each data point falls from the mean
  • Variance: the average of squared deviations from the mean (standard deviation squared)

Crime data is often skewed (a few extreme values pull the mean), so the median is frequently more useful than the mean. For instance, if most burglaries in a neighborhood result in losses under $\$500$ but a few involve losses over $\$10{,}000$, the mean loss will be misleadingly high.

Inferential statistics allow researchers to draw conclusions about a larger population based on a sample:

• Hypothesis testing compares a null hypothesis ($H_0$: no effect or no relationship) against an alternative hypothesis ($H_1$: an effect or relationship exists)
• The significance level ($\alpha$, typically set at 0.05) and the p-value determine whether results are statistically significant. If the p-value is less than $\alpha$, you reject the null hypothesis.
• Common statistical tests include:
  • t-tests: compare means between two groups
  • ANOVA: compare means across three or more groups
  • Regression analysis: examines relationships between variables and can control for multiple factors simultaneously

Data visualization communicates findings through tables, bar charts, line graphs, and scatterplots. Choosing the right visual matters: bar charts work well for comparing categories (crime types across cities), while scatterplots reveal relationships between two continuous variables (poverty rate vs. crime rate).

Evaluation of research validity

Not all research is equally trustworthy. Evaluating studies requires understanding both validity and reliability.

Validity refers to whether the research actually measures what it claims to measure.

• Internal validity asks: did the independent variable truly cause the observed effect? Threats include:
  • Confounding variables: extraneous factors that weren't controlled for
  • Selection bias: non-random assignment that creates unequal groups
  • Maturation effects: natural changes in participants over time that get mistaken for treatment effects
• External validity asks: do these findings apply beyond this specific study? It depends on:
  • Generalizability: whether results apply to other populations or settings
  • Sample representativeness: how closely the sample mirrors the target population
  • Ecological validity: whether the study conditions resemble real-world situations

Reliability refers to the consistency and reproducibility of findings.

• Test-retest reliability: does the same measure produce similar results when administered at different times?
• Inter-rater reliability: do multiple observers or coders agree on what they're seeing?
• Internal consistency (measured by Cronbach's alpha, $\alpha$): do all the items in a survey scale actually measure the same underlying concept? Values above 0.7 are generally considered acceptable.

A study can be reliable without being valid (you get the same wrong answer every time), but it cannot be valid without being reliable. This distinction comes up frequently on exams.

Critically evaluating research involves a systematic process:

1. Identify limitations and potential biases in the study design
2. Assess methodological rigor (Was the sample large enough? Were variables properly controlled?)
3. Consider alternative explanations for the findings
4. Evaluate whether the conclusions are actually supported by the evidence presented