Who Pays, Who Lives, Who Decides? The Respondent-Selection Problem in Household Housing Research

This week, a Chinese social-trend topic—"why adult children who buy a home put their parents in the second bedroom"—sparked intense debate, with one side talking about renovation budgets and childcare needs and the other about dignity and independence in old age. The disagreement itself demonstrates a core fact: household housing decisions rarely have "one correct answer"; they have "whoever answers." If research interviews only the person in each household who is easiest to reach by phone or most likely to complete a survey—usually the adult child—the result is not a household decision but the child's narrative. Treating that narrative as family consensus to steer product and marketing is the most systematic error in housing research.

The right starting point is to identify the decision unit at the household level: who funds, who lives there, who pays the mortgage, who provides care, and who gathers information. Only then choose a single-informant design, a paired two-generation design, or a short full-roster questionnaire, and state in the methodology exactly how respondents were selected and who answered on whose behalf. Housing research is won or lost on respondent structure, not questionnaire length.

Decision-Unit Identification: Housing Is Not a One-Person Purchase Decision

Housing differs from everyday consumer goods in that funding, residence, and decision-making often belong to different people and the cycle spans years. After commissioning, use a screening page to identify five roles within the household:

The second-bedroom debate is a textbook case: children weigh commuting and school districts; parents care about dignity and self-care space; and what actually drives conclusions is who paid the down payment, who lives there, and who will care for whom. Without role identification, a questionnaire only captures the answer of whoever speaks loudest.

Trade-offs Among Three Respondent Designs

After identifying roles, decide whom to interview per household. Three designs have different conditions; "interview the most knowledgeable person" is not a sufficient rule:

• One randomly selected adult per household: suitable for descriptive proportions about households as the unit (for example, the household-level share with multi-generational living intentions). Select respondents by a within-household rule (such as the adult with the nearest birthday) rather than letting the first person reached self-select; otherwise results skew toward younger, digitally fluent members.
• Most-informed respondent: suitable for factual questions (property details, funding structure, mortgage arrangements), but accept proxy-reporting error—one person relaying others' attitudes flattens disagreement into "our family thinks." U.S. Census Bureau materials on respondent selection note that proxy reporting by a primary respondent is standard in household surveys, that "who provides the information" affects results, and that factual and attitudinal questions should be treated differently.
• Paired two-generation design: suitable for attitudes, trade-offs, and conflict. Ask the child and the parent in the same household to answer separately and link responses by household ID. Sample size and quotas are counted in households, not persons, and follow-up must chase both members; otherwise you end up with broken pairs where only one side answered.

In short: for household-level descriptive proportions, randomly select one adult; for verifiable facts, use the most-informed respondent; to understand generational divergence and trade-offs, use paired interviews. The three designs require different methodology write-ups and must not be pooled as the same kind of evidence.

Illustrative Worked Example: How 100 Paired Households Rewrite the Conclusion

The following fully illustrative example shows how different single-informant and paired designs can be; it does not represent WeJot or any real study. Assume 100 urban households recruited on the conditions "bought a home in the past two years, down payment partly funded by parents, and parents have moved in or will within two years," with 100 children and 100 parents answering in pairs:

Interviewing only children yields the optimistic picture that 68% of parents moved in voluntarily and 74% of households reserved independent space for parents; interviewing only parents yields the opposite. The decisive numbers in the paired result are the agreement counts: only 47 households agree that parents chose voluntarily, meaning the two generations do not share the same understanding of "voluntary"; the gap on dedicated living space reaches 36 households—in the same home, parents do not feel the space "belongs to them." For a product team, the actionable denominator is 33 households (both sides recognize dedicated space), not 74; for policy or services, the generational gap in care expectations (57 vs 71) suggests communication and support design should be calibrated separately for each generation.

When computing pair-agreement rates, the denominator is households where both members completed the survey; do not treat single-respondent households as pairs, and do not add up "child says yes, parent says no" as mere error—the disagreement is itself the finding.

Recruiting by Household: Screening and Boundary Cases

Write eligibility at the household level, for example "bought a home in the past three years or plans to move within a year," "down payment or mortgage supported by parents or children," and "parents already co-residing or planning to within two years." Then select respondents from eligible households by rule. Boundary cases to decide in advance include:

• Multi-child families: which child co-resides with parents and how funding is split—do not default to the eldest as the only respondent;
• Parents living elsewhere but deeply involved in the decision—screen "co-residence" and "decision involvement" separately, and record the mode difference when interviewing remote parents by phone or video;
• Families with separate household registrations but shared funding—define the household as an economic-and-care community rather than by the registration booklet, and state the definition in the report;
• Complex structures such as blended families or aging parents whose children live far away—when sample size is insufficient, narrow the scope rather than excluding complex families and then pretending the sample covers them.

The Chinese household backdrop is shifting: the seventh national census showed average household size of 2.62 persons in 2020, down from 3.10 in 2010, with 18.70% of the population aged 60 and over and 13.50% aged 65 and over. Smaller households and deeper aging occurring together mean intergenerational support increasingly happens across separate households or through outsourced services; the household definition and role list in research must track this structure rather than reuse an outdated "three generations under one roof" template.

Design Separate Instruments for Parents and Children

Sharing one question set across generations manufactures fake agreement or fake disagreement. Three adjustments for the parent version: convert rating scales into plain conversational options (such as "not at all / okay / quite good") instead of seven-level agreement; move money-sensitive items to the later part of the questionnaire with a "do you mind saying" buffer; and ask living-arrangement preferences directly—"how would you like to live"—rather than "how satisfied are you with the current household living situation," which presumes the child is the protagonist. For older respondents with limited digital fluency, a short questionnaire with voice or assisted administration usually beats a long online form. Do not use "all parents completed the long online survey" as evidence of a healthy sample; it only proves you screened out people who could not complete it.

Also record proxy reporting actively: when a parent's interview is answered or operated by the child, mark proxy status in the data and report its share and likely direction separately. Many surprising housing findings—"parents actually prefer a senior-living community"—are not measurements of truth but a switch to a response channel that is not filtered by the child.

Do Not Rush to a Questionnaire During Exploration

If the team has not yet built a basic picture of who will live there, for how long, and how the money is arranged, a dozen two-generation interviews to map the decision process will save more budget than a questionnaire. Scenarios that require computing repayment capacity, taxes, or pension coverage are professional calculations, not attitude surveys; a questionnaire cannot impersonate a calculator. What a questionnaire contributes is measurable distributions and disagreement—how many people, how many households, by how much—not the decision itself.

Running This Study in WeJot

WeJot can link paired answers by household ID: a contact person completes the household screening page (identifying funding, residence, and care roles), separate versions are generated for children and parents, and both are grouped under one household ID, with follow-up for missing responses at the household level. Quotas are set by household rather than person, and cross-analysis uses "agreeing vs disagreeing households" as the unit. For the role-structure treatment of multi-informant decisions in organizational purchasing, see B2B survey sample and role methods; for household-level population definitions, see defining the target survey population.

The companion two-generation household housing decision survey template contains the household screener, two-generation versions, and representative items for paired analysis, and can be generated and adapted directly. Get the respondent structure right first; that is the cheapest and most valuable step in housing research.