SOST10021 Chap.1 Data Sources and Social Measurement
Data Sources and Social Measurement
Data Sources and Social Measurement begins with the evidence chain that supports every later comparison. Surveys can measure attitudes and self-reported outcomes, censuses provide broad population coverage, and administrative records describe contacts created through services or organisations.
The chapter shows why a research question must name its population, unit and time period before a convenient dataset is selected. Metadata, coding rules, coverage and repeated records determine what a variable means. A large file can still be poorly matched to a population claim, while a smaller probability survey can support carefully bounded inference.
The practical goal is to distinguish the people or events recorded by a system from the wider social condition a researcher wants to understand.
What this chapter covers
- 01
Survey evidence
- 02
Census coverage
- 03
Administrative records
- 04
Units of analysis
- 05
Variable metadata
- 06
Coverage and duplication
- 07
Claim boundaries
Worked example · free
Audit an administrative count before calling it prevalence
- 2Identify each row as a recorded resident request under the system's eligibility and coding rules.
- 2List mechanisms separating records from need, including non-application, use of another service, screening and duplicate entries.
- 2Report the administrative count accurately and narrow the prevalence claim until population evidence is available. This is an original revision allocation.
Key terms
- Target population
- The target population is the full set of units to which a research conclusion is intended to apply.
- Unit of analysis
- The unit of analysis is the person, household, event, place or other entity represented by each observation.
- Metadata
- Metadata describe how variables, categories, dates and records were created and should be interpreted.
Data Sources and Social Measurement FAQ
When is administrative data most useful?
Administrative records are useful for studying service activity, operational processes and large recorded populations over time. Their interpretation must follow the system that created them. They may omit people who never enter the service, include repeated events and reflect changing eligibility or coding rules.
Why does the unit of analysis matter?
A file can contain visits rather than patients, applications rather than households or repeated observations rather than unique people. The unit determines the denominator and prevents event counts from being described as population counts. Write the unit explicitly before comparing values.
Can a large dataset still be biased?
Yes. More observations can reduce random variation while leaving systematic undercoverage, duplication or misclassification unchanged. Size cannot repair a mismatch between the recording system and target population. Evaluate provenance, inclusion rules and missing groups alongside the case count.
How do repeated records change a claim?
Repeated records may represent genuine events, such as several appointments made by one person, rather than erroneous duplicates. Decide whether the question concerns events, episodes or unique people, document the deduplication rule, and report both event and person counts when each illuminates a different aspect of inequality.
Exam move
Create a source audit for every substantive example. Write the research question at the top, then add the target population, observational unit, reference period, source type and operational variable. For surveys, record sampling and response; for censuses, check coverage and timing; for administrative sources, document eligibility, access, duplication and coding changes.
Practise converting broad claims into source-faithful statements, such as replacing population prevalence with recorded service demand when no population denominator exists. Reverse the exercise by choosing the evidence needed for a population question and explaining why the available source falls short. Use small diagrams to trace question, source, variable and conclusion.
When comparing sources, ask whether they measure the same construct, population and period rather than assuming similar labels imply comparability. Keep a log of errors involving people versus events, individual versus household measures and current versus cumulative records. Finish each practice response with a decision: fit for purpose, usable for a narrower claim or unsuitable without another source.
That closing judgement turns a checklist into research reasoning and prepares the evidence evaluation used throughout the module. Add a provenance card for one dataset each week. Record who produced it, why records were created, how eligibility changed, which identifiers define unique cases and whether values can be linked across time.
Then imagine one person who should be represented but could be absent and one record that could be counted twice. Explain the likely effect on the result rather than merely naming a risk. Compare a survey estimate and an administrative count on the same topic, identifying why their numerators and populations differ. Finish by writing a public-facing title that communicates the true scope without internal technical notes.
This exercise develops concise reporting while keeping the complete audit available for a methods paragraph. Before the final, practise choosing among source types for new questions and defending the choice against a plausible alternative.
Working through Data Sources and Social Measurement in SOST10021? Sia is AskSia’s AI Statistics tutor — ask any SOST10021 Data Sources and Social Measurement question and get a clear, step-by-step explanation grounded in how SOST10021 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.