LAWS6032 Chap.5 Experimental and Quasi-Experimental Designs
Experimental and Quasi-Experimental Designs
Experiments compare outcomes after manipulating a treatment, ideally with random assignment that creates a credible counterfactual on average. Criminological settings often make randomisation difficult, so quasi-experimental designs use thresholds, timing, matched groups or policy variation.
This chapter treats every design as an explicit comparison supported by assumptions, and separates internal validity from transfer to other institutions and populations.
What this chapter covers
- 01
Treatment, outcome and control conditions
- 02
Random assignment and allocation concealment
- 03
Counterfactual reasoning
- 04
Selection, history, maturation and regression
- 05
Contamination, noncompliance and attrition
- 06
Internal versus external validity
- 07
Matching, interrupted time series and policy thresholds
- 08
Statistical significance versus design credibility
Diagnosing selection in voluntary treatment
- +2Identify willingness to participate as a selection process that may track motivation, stability or support.
- +2Compare baseline characteristics and report program receipt, follow-up and differential attrition.
- +2Consider stronger variation such as phased rollout, capacity limits or an eligibility threshold, with assumptions stated.
- +2Present the result as promising association unless exchangeability or a stronger quasi-experimental warrant is defended.
Key terms
- Random assignment
- Allocation by chance intended to balance participant characteristics across treatment conditions on average.
- Counterfactual
- The outcome that would have occurred for the treated population under the alternative condition.
- Control group
- A comparison condition used to estimate what may have happened without the studied treatment.
- Quasi-experiment
- A design estimating effects without researcher-controlled random assignment by using structured comparison variation.
- Internal validity
- The credibility of attributing an observed difference to treatment rather than bias or another event.
- Attrition
- Loss of cases during follow-up, especially concerning when it differs across conditions or relates to outcomes.
- Contamination
- Exposure of the comparison group to treatment elements, reducing the distinction between conditions.
Experimental and Quasi-Experimental Designs FAQ
Why does random assignment help causal inference?
Random assignment makes treatment status independent of participant characteristics in expectation, so outcome differences have a stronger counterfactual interpretation. Its value depends on correct implementation, follow-up, measurement and analysis; randomisation does not prevent attrition or treatment crossover.
What is the difference between an experiment and a quasi-experiment?
An experiment assigns treatment under researcher control, commonly at random. A quasi-experiment uses naturally occurring or policy-created variation without controlled random assignment. It can support strong inference when the assignment process and identifying assumptions are credible.
How do internal and external validity differ?
Internal validity concerns whether treatment produced the observed difference in the study. External validity concerns whether that effect would transfer to other populations, institutions or implementation conditions. Tight control can improve attribution while limiting real-world transfer.
Does a statistically significant result prove the program worked?
No. Statistical significance is interpreted within a model and does not repair selection, contamination, attrition or invalid measurement. First establish that the design supports the comparison; then assess estimate size, uncertainty, practical importance and transfer.
What should I report about program implementation?
Report whether assigned participants received the program, how much was delivered, crossover, attrition and fidelity. These details distinguish the effect of offering treatment from the effect of receiving it and help separate an ineffective theory from a program that was never implemented as intended.
Assessment move
Draw treatment and comparison groups before reading any outcome. Write how assignment occurred, which people could cross conditions and who could disappear before follow-up. Use a threat checklist: selection, history, maturation, regression to the mean, contamination, noncompliance, attrition and measurement. For each threat, state whether it would likely inflate, reduce or obscure the difference.
Compare one experimental and one quasi-experimental design for the same policy problem, focusing on the assumptions each needs rather than ranking them by label. Practise separating efficacy under controlled conditions from effectiveness in ordinary settings. When a numerical result is supplied, interpret the effect size and uncertainty only after establishing the design warrant.
Close with a transfer sentence naming the institutional conditions that would need to remain similar. Add an assignment audit to every example: who controlled allocation, could the next condition be predicted, and could staff or participants change entry around the rule? For quasi-experiments, write the identifying assumption in ordinary language and list an observable pattern that would weaken it.
Track exposure separately from assignment so implementation failure is not confused with theory failure. Finally, rehearse a null-result explanation that distinguishes an ineffective intervention, insufficient delivery, insensitive measurement and imprecise estimation without choosing among them before examining the relevant evidence.
Working through Experimental and Quasi-Experimental Designs in LAWS6032? Sia is AskSia’s AI Law tutor — ask any LAWS6032 Experimental and Quasi-Experimental Designs question and get a clear, step-by-step explanation grounded in how LAWS6032 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
