Questionnaire length is an important factor in survey cost and response rates. Reducing respondent burden through improved questions can potentially result in higher data quality and even prevent breakoffs. Advances in survey technology allow the measurement of overall interview length as well as precise capture of response times to individual questions. An emerging literature has examined the effects on response times of various item features as well as respondent and interviewer characteristics in a multilevel analytic framework in web, in-person, and telephone surveys. This paper analyzes in-person interviews from the New York City Housing and Neighborhood Study, a randomized natural experiment evaluating the impact of affordable housing on overall wellbeing. Householder interviews (affordable housing applicants with no co-resident children) took on average 60 minutes; caregiver interviews (applicants with co-resident children) took on average 90 minutes and included all householder questions plus additional caregiving and child-related questions. Interviews were conducted variously in respondents’ homes, at project offices, and in public locations like coffee shops, libraries, or parks. This paper builds on previous work by examining the variability in response time due to interviewers, respondents, and items. We explore which characteristics of each are associated with response time as well as investigating interview setting through a sociocultural as well as a cognitive perspective. Results provide insight into which interview settings may be more encouraging to respondents as well as which question types and answer features pose challenges, and which may be easier to answer. This analysis supports improvements in study design that may lower respondent burden and fatigue, produce higher data quality, and reduce break-offs.