The Social Signal of Jobs: An Audit Study on How Jobs Influence Housing Decisions

2025-10-06

2026-03-16

Whether an inquiry to a “roommate wanted” listing receives a response, and whether the response is affirmative.

Constructed as binary variables (1 = any response received within 10 days, 0 = no response; 1 = the response is affirmative, 0 = no response or the response is negative). For analysis, occupations will primarily be pooled into “socially desirable” and “less socially desirable” categories, rather than considered individually.

We will respond to “roommate wanted” postings on a popular classified advertisements website with short, standardized messages on behalf of real participants. Messages will include participant-provided information (name or alias, occupation, age range). Each listing will receive two inquiries.

We are screening participants who hold a set of occupations chosen to vary in perceived roommate desirability. If we cannot reach the target number of survey completions with the survey firm with whom we are contracting, we may expand the list of occupations to increase feasibility. The list includes:

Socially Desirable Occupations
-Chief executives
-Lawyers
-Software developers
-Computer programmers
-Registered nurses
-Licensed practical and licensed vocational nurses
-Nurse midwives
-Nurse practitioners
-Nurse anesthetists
-Financial managers
-Project management specialists
-General and operations managers
-Administrative services managers
-Management analysts
-Financial and investment analysts
-Accountants and auditors
-Postsecondary teachers
-Education and childcare administrators
-Computer and information systems managers
-Network and computer systems administrators
-Computer systems analysts
-Information security analysts

Less Socially Desirable Occupations
-Food preparation workers
-Maids and housekeeping cleaners
-Stockers and order fillers
-Packers and packagers, hand
-Janitors and building cleaners
-Cashiers
-Waiters and waitresses
-Retail salespersons
-Parts salespersons
-Customer service representatives
-Laborers and freight, stock, and material movers, hand
-Receptionists and information clerks
-Driver/sales workers and truck drivers
-Industrial truck and tractor operators
-Insurance sales agents

When a participant’s profile is used, we use their complete set of characteristics – we do not randomize occupations independently from other characteristics.

We will select both which listings are contacted and which participant profiles are used to contact each listing using computer randomization.

Each weekday that the study is active, we will collect all postings from the “roommates wanted” listing board for several major metropolitan areas and randomly select a subset to contact. We will send each of the chosen listings two inquiries, with messages that reference the profiles of different survey participants.

We will randomize profiles across listings. If the post indicates a gender preference (e.g., “women only”), we will randomize among profiles that meet this criterion. We will also randomize within city – e.g., for postings in New York City, we will randomly draw profiles from participants who indicated this as their location.

Individual “roommate wanted” listings, and profiles. Each listing will receive two inquiries, each with different random participant profiles.

No

N/A

We will recruit roughly 500 survey participants whose real profiles will be used to contact the subjects. We will contact approximately 20 listings for each survey participant. To reach our target of 3,500 sent messages, we will need to use 175 profiles (35% of participants). The excess profiles (beyond the minimum of 175) give us flexibility to choose profiles that are demographically balanced by gender and age, have occupations in our experimental set, and have common names. This allows us to avoid introducing confounds with individual occupation in our analysis.

Jobs will be analyzed in pooled categories of “socially desirable” vs. “less socially desirable” occupations. We will send approximately the same number of responses for these pooled categories.