From problem to question: answer key
Introduction to Social Research Methodology
Task 1: three questions, three goals
There is no single correct set of questions; any question grounded in the vignette that genuinely serves its stated goal and passes the four criteria (answerable, focused, feasible, non-trivial) earns full credit. Model answers:
Figure 1 shows them as the deck’s model-answer slide does, with the commonest error underneath.
Exploratory: What do employees who have voluntarily left Meridian in the past twelve months say about their reasons for leaving? Justification: nobody at Meridian yet knows which factors matter, so the question’s purpose is to build initial familiarity and generate the categories and hunches — pay, rotas, management, something unanticipated — that later, more structured research will measure and test.
Descriptive: How does voluntary staff turnover at Meridian vary across stores, roles, tenure bands, and the store/e-commerce divide? Justification: the question asks precisely what is happening, where, and to whom — it maps the pattern of the problem without yet claiming anything about its causes, which is exactly the descriptive goal.
Explanatory: Does perceived workload under the new rota system affect front-line store employees’ intention to quit? Justification: the question relates a presumed cause (perceived workload) to a presumed effect (intention to quit), so answering it would tell the board why people leave and hence which lever to pull — the defining ambition of explanation.
Marking guidance for Task 1
- Goal fit is the main criterion. The commonest error is an “exploratory” question that is really descriptive (e.g. “how many employees are dissatisfied?”) or an “explanatory” question that is really a prediction or a decision (“will turnover fall?”, “what should Meridian do?”). Ask of each submission: does the exploratory question tolerate not knowing the categories in advance? Does the descriptive question ask only what/where/how much? Does the explanatory question name a candidate cause and an outcome?
- Check the four criteria explicitly. Each question should name a defined population/setting drawn from the vignette (answerable, focused), be doable with the access the board granted — records, staff, customer feedback data (feasible) — and have an answer that is genuinely uncertain (non-trivial).
- Justifications must name the kind of knowledge produced, not restate the question. “This is exploratory because it explores” earns nothing; “this builds the categories a later survey would measure” earns full credit.
- Customer-satisfaction versions of any question are equally acceptable (e.g. descriptive: “how have satisfaction scores changed by channel and store since the e-commerce launch?”); do not require the turnover framing.
Expected discussion points
- The three questions form a programme, not a menu: exploration feeds description (what to measure), description feeds explanation (what needs explaining). Strong pairs will have noticed that their three questions chain together; make this explicit in the plenary.
- Turnover vs satisfaction: some pairs will split their questions across the two halves of the board’s worry. Point out that this is legitimate but is itself a scoping decision — a real team would flag to the board that one project cannot chase both problems with equal depth.
- Intention to quit vs actual quitting: if a pair uses “intention to quit”, credit it and note the design advantage (you can survey current staff) and the limitation (intention is not behaviour) — a preview of measurement issues in session 8.
- The board deliverable framing: the best answer to “why these three questions?” is managerial — together they tell the board where the problem lives, what its vocabulary is, and which lever might move it.
Task 2: diagnose and repair
For each item: the flaw, why it fails, and a model rewrite. Accept any rewrite that removes the named flaw without introducing another; the rewrites below are models, not the only acceptable answers. Several items commit more than one flaw — credit any defensible diagnosis, but the primary flaw listed below should be found.
Figure 2 and Figure 3 summarise the primary flaw, the criterion it violates and the model rewrite for each question, as on the deck’s two model-answer slides.
1. “How satisfied are Meridian’s employees with their pay and their opportunities for promotion?”
- Flaw: double-barrelled.
- Why: it asks two questions at once (pay; promotion) which may have different answers — an employee content with pay but blocked from promotion has no single answer to give. Violates focused.
- Model rewrite: split into two: How satisfied are Meridian’s employees with their pay? and How satisfied are Meridian’s employees with their opportunities for promotion? (Both descriptive; measured separately, they can later be compared as rival predictors of quitting.)
2. “Why does Meridian’s outdated rota system force our best salespeople to resign?”
- Flaw: leading (primary); also value-laden (“outdated”).
- Why: it presupposes the very thing at issue — that the rota system causes resignations, and that it is outdated — so any study built on it has convicted the accused before the trial. Violates answerable (the real question was never asked).
- Model rewrite: first ask whether: Is there an association between working under the new rota system and voluntary resignation among Meridian’s sales staff? — and only if yes, proceed to why.
3. “Is it fair that e-commerce staff can work from home while store staff cannot?”
- Flaw: value-laden.
- Why: “fair” asks for a moral verdict, which no data collection can deliver; fairness is a judgement, not a measurable property. Violates answerable.
- Model rewrite: research the empirical parts: How do store-based and e-commerce employees differ in their perceptions of the fairness of Meridian’s working arrangements, and how are those perceptions related to intention to quit? The verdict on fairness itself stays with the board, informed by this evidence.
4. “Would Meridian’s turnover have stayed at 14% if the company had never launched its e-commerce arm?”
- Flaw: unanswerable (bare counterfactual).
- Why: the Meridian-without-e-commerce does not exist to be observed; no evidence could settle the question as posed. Violates answerable.
- Model rewrite: rebuild around an observable comparison: Have turnover rates risen more in the stores most affected by the e-commerce expansion (e.g. through rota changes or staff transfers) than in the stores least affected?
5. “What influences how people behave at work?”
- Flaw: too broad.
- Why: no population, place, period, or outcome — it names an entire field (organisational behaviour), so no single study could address it. Violates focused and feasible.
- Model rewrite: narrow by component, type, place, and relationship: How does perceived workload relate to intention to quit among Meridian’s front-line store staff with under two years’ tenure?
6. “Do Meridian employees whose overtime goes unpaid feel less happy about their overtime than those who are paid for it?”
- Flaw: trivial.
- Why: it is answerable, focused, and feasible — but the answer is known before the study begins, so nothing is learned. Violates non-trivial.
- Model rewrite: find the genuinely uncertain version: How much does unpaid overtime contribute to Meridian employees’ intention to quit, relative to workload and relations with line managers? — the ranking of causes, unlike the direction of the preference, is genuinely unknown.
7. “Should Meridian close its five worst-performing stores?”
- Flaw: not researchable (a decision question posed as an empirical one).
- Why: “should” mixes evidence with strategy, cost, and values, which research can inform but cannot settle; no data pattern equals a decision. Violates answerable.
- Model rewrite: extract the researchable core: To what extent is the underperformance of Meridian’s five worst-performing stores explained by store-specific factors (location, local competition, staffing) rather than company-wide ones? — evidence the board would need before deciding.
8. “Will staff turnover at Meridian fall next year?”
- Flaw: prediction without mechanism.
- Why: a bare forecast has no explanatory content — whatever happens next year, nothing is learned about why, so nothing transfers to any other decision. Violates non-trivial (and, as posed, answerable: only waiting answers it).
- Model rewrite: anchor the forecast in a mechanism: Does perceived workload drive resignations at Meridian, such that the planned increase in store staffing should reduce turnover over the next year?
Marking guidance for Task 2
- Award (a), (b), and (c) separately. A correct flaw label with a weak rewrite, or a good rewrite with a mislabelled flaw, each earn partial credit. The explanation (b) should reference one of the four criteria; that link is the learning objective.
- Multiple flaws: items 2 (leading + value-laden) and 3 (value-laden; some will say “not researchable”, which is defensible — the two flaws overlap wherever “should/fair” language appears) legitimately admit more than one label. Accept any label the pair can defend; in the plenary, use item 2 to make the “flaws hunt in packs” point.
- Watch for rewrites that introduce new flaws — the commonest failure: repairing a leading question into a double-barrelled one, or repairing “too broad” so aggressively that the result is trivial or an unresearchable micro-incident. A rewrite is only complete when it passes all four criteria.
- Rewrites should stay loyal to the manager’s intent. A rewrite of item 7 that abandons the store-closure issue entirely (e.g. pivoting to turnover) has dodged the translation task rather than performed it.
- Do not require the model wording. In particular, item 4 can be repaired by any before/after or more-affected/less-affected comparison; item 8 by any mechanism-anchored version; item 6 by any question whose answer is genuinely uncertain.
Expected discussion points
- Where does the flaw live? Items 1 and 2 fail in their wording; items 4 and 7 fail in their logic. Wording flaws can be edited away; logical flaws require rebuilding the question — a useful distinction for the plenary.
- Goal shifts in repair: the model rewrite of item 2 turns a “why” question into a “whether” (descriptive/comparative) question. This is the right repair, not a retreat: description must establish the pattern before explanation can be attempted. Expect some pairs to resist “downgrading” — this is the moment to reinforce the explore-describe-explain sequence.
- Items 3 and 7 together teach the session’s managerial lesson: research serves values-soaked decisions by supplying the empirical parts, never by delivering the verdict. Link back to the “managerial question ≠ research question” slide.
- Item 8 previews session 3: the repaired version — a mechanism plus a forecast — is one step short of a hypothesis with an independent and a dependent variable. Say so explicitly; it makes the next session feel inevitable rather than new.
Task 3: plenary discussion
The worksheet’s three plenary questions draw on the discussion points above. Figure 4 gives the model answers, as on the deck’s last model-answer slide: different flaws in the same question are often both right (flaws hunt in packs); a repair that turns why into whether is the right repair, because the pattern must be established before it is explained; wording flaws can be edited away, but logic flaws must be rebuilt; and for values-laden questions such as 3 and 7, research supplies the evidence a decision needs, never the verdict itself.