A consumer research firm wants to estimate average household spending on groceries in a metropolitan area with 10 distinct neighborhoods that differ significantly in income level. To ensure each neighborhood is represented, they use stratified random sampling with equal allocation (50 households per neighborhood). What potential issue arises with EQUAL (rather than proportional) allocation?
- ASmaller neighborhoods will be overrepresented and larger neighborhoods will be underrepresented relative to their actual size.
- BThe sample will be biased because households self-select into the study.
- CBoth A and DCorrect
- DA simple unweighted average of all 500 households will not produce an unbiased citywide estimate without adjusting for neighborhood sizes.
Why C is right
With equal allocation (50 households from each of the 10 neighborhoods regardless of size), each neighborhood contributes the same number of observations. Relative to the population, smaller neighborhoods are overrepresented and larger ones are underrepresented (choice A). Consequently, a simple unweighted average of all 500 observations would give too much weight to smaller neighborhoods; to estimate the citywide average the strata must be weighted by each neighborhood's actual share of the population (choice D). Both issues stem from equal (rather than proportional) allocation, so the correct answer is 'Both A and D.' Choice B is wrong because the sampling is random within strata, not self-selected.
Why the other 3 choices miss
- A
- This is correct — equal allocation with unequal-sized neighborhoods does create disproportionate representation.
- B
- All neighborhoods are sampled with equal numbers (50 each). The issue is whether this equal number is proportional to their actual sizes in the population.
- D
- This is also correct — a simple average of all 500 observations would give too much weight to smaller neighborhoods; weighting is needed.