Since 1973, the US Census Bureau has administered the American Housing Survey (AHS) in odd-numbered years. Surveyors ask questions about the quality and value of respondents’ housing, and have a battery of questions for the subset of respondents who moved recently, asking about their search process. The AHS regularly adds new questions and rephrases old ones from year to year.
In 2021, they rolled out a group of three new questions. One asks recent movers whether they spent more or less than a month searching for their new home. This is the AHS’s first-ever variable dealing with search time. The other two, which are similar to questions asked in previous years, ask about search scope: whether respondents looked for housing in neighborhoods besides the one they ended up moving to, and whether they looked at other housing units in the same neighborhood they moved to.
In analyzing these new questions within the largest 15 metro areas in the US, I found a curious and hard-to-explain relationship. Search time – the binary variable of taking more or less than a month to look for a home – seems to be predicted by the population of a metropolitan area (with r=0.57 and p > |t| = 0.026), whereas search scope in the sense of looking at multiple neighborhoods or multiple units within a neighborhood seems to be predicted by the cost of housing as gathered from 2021 Zillow data (for looking at multiple neighborhoods, r=0.65 and p > |t| = 0.009; for looking at multiple units, r=0.68 and p > |t| = 0.007).
The inverse set of relationships are much weaker. Price and likelihood of taking over a month to search are positively correlated, but the relationship is not statistically significant; the same is true of the relationship between metro population and likelihood of looking at multiple units within their neighborhood. The relationship between metro population and looking in multiple neighborhoods is even weaker. The strong relationships outlined above are robust to controlling for price and population respectively. It’s possible that the vague definition of “neighborhood” could be pushing down the correlation between looking at multiple neighborhoods and metro population: if people in bigger cities imagine “neighborhoods” as larger units than their counterparts in smaller cities, they’d be more likely to search only within one “neighborhood”.
It’s hard to tell a story about this unusual finding, given how new these questions are. Perhaps future iterations of the same variables will firm up the weaker correlations observed in 2021. For that matter, the difference might be just an artifact of region-specific data only being available for the top 15 metro areas. But if years hence it becomes conventional wisdom that people look longer in bigger cities but more widely in more expensive cities, you heard it here first.
