POLS1009 Chap.6 Observational Designs, Datasets and Drafting
Observational Designs, Datasets and Drafting
Most political questions cannot be answered by experiment, so this chapter turns to observational designs, in which the researcher records outcomes as they occur and must work to make comparisons fair because confounding is always possible.
It sets out the two observational families: qualitative research studies a few cases closely and is good at showing how an outcome came about, while quantitative research measures many cases and is good at average patterns, and neither ranks above the other.
Within the qualitative family it covers process tracing inside one case, comparative case studies across a few, and content analysis of texts, along with Mill's logic that cases should differ on the outcome yet resemble each other otherwise. On the quantitative side it explains dataset anatomy and the three data structures: cross-sectional snapshots, time series of one unit, and panels that track many units over time.
It closes with the four-step research-design section of a paper, four questions for choosing a design, and the order in which a draft should be revised.
What this chapter covers
- 01
Observational design and why confounding is always possible
- 02
Qualitative and quantitative families as complements
- 03
Process tracing, comparative case studies and content analysis
- 04
Mill's comparable-cases logic
- 05
Rows, columns and the unit of observation in a dataset
- 06
Cross-sectional, time-series and panel data
- 07
Four questions for choosing a design
- 08
Writing and revising the research-design section
Worked example · free
Restructuring a council panel: from the full grid of councils and years to a snapshot or a time series
- 1Unit and rows. Each row is one council in one year, a council-year, and five councils over five years give 25 rows.
- 1Snapshot. Keeping only the 2023 column leaves one row per council, 5 rows, all sharing the same year.
- 1Time series. Keeping one council across all five years leaves another 5 rows, now with the council fixed and the year changing.
- 1Why the panel. The 2023 snapshot cannot show anyone's past, and the single-council series has no comparison council; the panel holds both the before-and-after change and the non-streaming councils.
Key terms
- Observational design
- A research design in which the researcher does not control who receives the treatment but records outcomes as they occur, using careful comparison to make up for the absence of random assignment.
- Process tracing
- A within-case method that follows the chain of events inside one case step by step, checking whether the order and content of what happened match the proposed causal story.
- Comparable cases
- Cases chosen so that they differ on the outcome of interest while resembling each other on other relevant factors, which makes the remaining difference the leading explanation.
- Panel data
- Data that record the same set of units at several points in time, so that variables change both across units and over time; it is also called longitudinal data.
- Revising
- Reworking a draft's structure, argument and organisation before polishing sentences and finally correcting typos and grammar, as distinct from proofreading alone.
Observational Designs, Datasets and Drafting FAQ
Is observational research less scientific than an experiment?
No. Observational designs answer most of the questions political scientists care about, because many causes cannot be assigned at random. They do demand extra care: the researcher has to select comparable cases, hold rival explanations constant and use statistical controls so that a comparison approaches the fairness an experiment would give automatically.
What is the difference between cross-sectional and panel data?
Cross-sectional data observe many units at a single moment, so variables differ only between units. Panel data follow the same units over several periods, so variables differ between units and within each unit over time. For a panel of countries, each row describes a country in a particular year rather than a country alone.
How do I choose between a qualitative and a quantitative design?
Start from the question rather than the method. Questions about meanings, processes or how something happened tend to suit qualitative work; questions about average patterns or effects across many cases tend to suit quantitative work. Also weigh how many cases you have, what data you can realistically collect and whether your claim covers one case or many.
What should a research design section in a paper include?
It should explain which cases you study and why, how each concept becomes a measurable indicator, where the data come from, and what analysis you will carry out and why it fits the question. At this point you are justifying a plan, not reporting findings, so the section reads as an argument for your choices.
Why compare cases that are similar rather than very different?
If two cases share almost everything except the outcome, the few factors on which they differ become strong candidates for explaining it. Cases that differ in many ways at once leave too many possible explanations, and cases with the same outcome offer nothing to explain. Similarity provides control while the outcome provides variation.
Exam move
Work through this chapter by building small examples of your own. Invent a question, then sketch one version of the study as process tracing, one as a comparison of two similar cases and one as a dataset, and note what each could and could not show. For every dataset you meet in readings, write down what one row is and whether the data are a snapshot, a time series or a panel.
Finally, draft a four-part research-design paragraph for a question you care about and revise it twice, first for structure and ideas and only then for wording. Doing this a few times across the semester makes the distinctions automatic rather than memorised.
Working through Observational Designs, Datasets and Drafting in POLS1009? Sia is AskSia’s AI Arts and Humanities tutor — ask any POLS1009 Observational Designs, Datasets and Drafting question and get a clear, step-by-step explanation grounded in how POLS1009 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.