Step 5: Research Design
- The research design is the overall plan that links your question to your data: who you study, when, how often, and what you compare.
- Main designs: experimental (you randomly assign a treatment), quasi-experimental (a comparison without random assignment), cross-sectional (one point in time), longitudinal (the same people over time) and case study (one or a few cases in depth).
- Causal claims need three things: association, the cause comes first, and no plausible alternative explanation.
- Random assignment is the most powerful way to rule out alternatives, because it balances the groups on everything, measured or not.
- Internal validity asks whether the conclusion holds in your study; external validity asks whether it holds elsewhere.
What a research design does
Your question says what you want to know. The design says what evidence would answer it. If your question is “how many?”, a single well-sampled survey may be enough. If it is “does X cause Y?”, you need a design that can rule out other explanations. Choosing the wrong design is the most common reason a dissertation cannot answer its own question.
The main designs
| Design | What it is | Part-time jobs example | Can it show causation? |
|---|---|---|---|
| Experiment (RCT) | Randomly assign units to treatment and control | Randomly offer a time-management workshop to half of working students | Yes, strongest |
| Quasi-experiment | Compare treated and untreated groups without randomisation | Compare students at a campus that introduced a workshop with one that did not | Sometimes, with care |
| Cross-sectional | Measure many units once | One survey of hours worked and marks | Rarely: associations only |
| Longitudinal / panel | Measure the same units repeatedly | Survey the same students each term for two years | Better than cross-section: shows timing |
| Case study | Study one or a few cases in depth | An in-depth study of working students at one college | Explains mechanisms; limited generalisation |
Chapter 5 of Research Methods Made Simple covers this in full, with a worked randomised workshop trial and design decision chart.Get the textbook · $15
When can you say “causes”?
To claim that X causes Y you need three conditions:
- Association: X and Y move together.
- Temporal order: X happens before Y.
- No alternative explanation: nothing else could produce the association.
The third condition is the hard one. Students who work long hours may have lower marks because of the work, or because students from poorer families both need to work and have less support at home. Family income is a confounder: it affects both the cause and the outcome.
Why random assignment is so powerful
If you randomly decide who gets the workshop, then by chance alone the two groups will be similar in family income, motivation, ability and everything else, including things you never measured. Any difference in marks afterwards can then be credited to the workshop. That is why the randomised controlled trial is the benchmark for causal questions.
Sixty working students volunteer. Thirty are randomly assigned to a four-week workshop; thirty are put on a waiting list. At the end of the semester the workshop group averages 64.1% and the waiting-list group 60.3%. What can you conclude?
A waiting-list control is a fair ethical choice: everyone gets the workshop eventually.
Threats to internal validity
| Threat | What happens |
|---|---|
| Selection | Groups differ before the treatment (volunteers are keener) |
| History | Something else happens at the same time (a strike, a new grading policy) |
| Maturation | People change naturally over time (students get better at exams anyway) |
| Attrition | People drop out, and those who leave differ from those who stay |
| Testing | Taking a pre-test changes performance on the post-test |
A control group protects against history, maturation and testing, because both groups experience them. Randomisation protects against selection.
Practise until it sticks: Research Methods Made Simple Practice Questions has 1,004 questions, each with a worked answer.Get the practice book · $9
Practise: 6 questions
Try each one on paper first, then check the worked answer.
EasyQ1. Name the design: (a) one survey of 500 workers in March 2026; (b) the same 500 workers surveyed every year for five years. Show worked answer
EasyQ2. State the three conditions for a causal claim. Show worked answer
MediumQ3. Students who eat breakfast get higher marks. Suggest a confounder and explain how it works. Show worked answer
MediumQ4. Why is a waiting-list control often more ethical than a no-treatment control? Show worked answer
MediumQ5. A college runs a workshop and compares marks before and after, with no control group. Name two threats to internal validity. Show worked answer
ExamQ6. Evaluate a cross-sectional survey as a design for answering “Do part-time jobs lower students’ marks?” [6] Show worked answer
Check yourself
1. Random assignment mainly protects against
2. External validity concerns
3. A variable that affects both cause and outcome is a
Your score: 0 out of 3
The textbook and the practice book together: the Research Methods Made Simple bundle for $19 (was $24).Both books · $19
The short version
- The design links the question to the evidence.
- Experiments, quasi-experiments, cross-sections, panels and case studies each answer different questions.
- Causation needs association, temporal order and no alternative explanation; random assignment is the strongest tool.
- Name your design, justify it, and state its limits.
Studying more than one subject? The Complete Book Bundle includes every Economics, Statistics, Mathematics, Econometrics and Research Methods book.