Experimental Design
To measure the effect of treatment on various outcomes, we estimate the following:
Y_i = β_0 + β_1 TREAT_i + X'_i γ+ ε_i
where Y_i is the outcome of interest for user i and β_1 is the coefficient of interest on the treatment variable.
If HUD data are available, covariates in X_i will include, total annual income, an indicator for whether the user’s housing voucher is from an authority with “Moving to Work” designation, indicators for race (white, black, and hispanic), indicators for users with children in the following groups: 0-4, 5-10, 11-14, and 14+, indicator for “Hard to House” status (as defined by HUD), and commuting zone fixed effects. HUD data may not be available to us, however, in which case our covariates will be voucher status, an indicator submitted rental app pre-treatment, has accessibility needs, intended move-in date, indicator for mobile/desktop browser, week fixed effects, missingness indicator + imputation for missing values Because people may visit the site with minimal engagement, we will also examine our results for the subgroup formed by restricting the sample to those who have a mobile/desktop indicator, indicating the user entered a search term and didn’t leave the site before ever searching for a property (and so seeing a treatment). Another subgroup is people with rental applications at baseline, we can also add: baseline tract atlas and pollution scores, age, reason for moving, income, employment, and whether they have children. Within this group, we will explore effects for those users with children.
We also collect information from two surveys given to AffordableHousing.com users: Air Quality Survey This survey was offered to 25,000 user accounts stratified by experimental treatment assignment. We describe the purpose of the survey as aiding researchers to “understand how people perceive and prioritize environmental factors, like air quality, when searching for housing.” This survey includes questions on the respondent’s background (gender, age, education, etc.), their reasons for moving, and their prior beliefs about air quality. We then provide an information intervention where we randomly provide the following information about the difference between high and low air quality areas in the US to some users:
"This is an equivalent to smoking 36 cigarettes a year. Breathing in fine particulate matter makes it harder for your body to get the oxygen it needs. Over time, this can lead to serious health problems, including heart issues and even damage to your brain. Studies show that even short exposure to these tiny particles can make you feel tired, make it harder to concentrate and think clearly, and, over time, increase the risk of diseases like Alzheimer’s and dementia."
Respondents are then asked to interpret air quality information provided via a chloropleth map and an informational badge. Economic Mobility Survey This survey was offered to 25,000 user accounts stratified by experimental treatment assignment. We describe the purpose of the survey as aiding researchers to “understand how people perceive and prioritize certain factors, when searching for housing.” This survey includes questions on the respondent’s background (gender, age, education, etc.), their reasons for moving, and their prior beliefs
about economic opportunity. We then provide an information intervention where we randomly provide the following information about the difference between high and low economic opportunity areas in the US to some uses.
"If a child moved at birth from the average low-economic mobility neighborhood (bottom 25% in the US) to the average high economic mobility neighborhood (top 25% in the US), the expected income of that child as an adult would increase by $17,000 per year”.
Respondents are then asked to interpret economic mobility information provided via a chloropleth map and an informational badge.