Back to History Current Version

Housing Expectations and Market Behavior

Last registered on July 19, 2019

Pre-Trial

Trial Information

General Information

Title
Housing Expectations and Market Behavior
RCT ID
AEARCTR-0003663
Initial registration date
July 17, 2019

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
July 19, 2019, 11:58 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Region

Primary Investigator

Affiliation
UC Berkeley

Other Primary Investigator(s)

PI Affiliation
Cornell

Additional Trial Information

Status
In development
Start date
2019-06-10
End date
2021-12-31
Secondary IDs
Abstract
We designed a field experiment to measure how housing price expectations affect real estate transactions in the United States. Our motivation stems from the fact that home price expectations play a prominent role in many accounts of the housing boom that occurred during the mid-2000s. We will launch a large-scale high-stakes information experiment with tens of thousands of households who have recently listed their houses for sale.
External Link(s)

Registration Citation

Citation
Bottan, Nicolas and Ricardo Perez-Truglia. 2019. "Housing Expectations and Market Behavior." AEA RCT Registry. July 19. https://doi.org/10.1257/rct.3663-1.0
Former Citation
Bottan, Nicolas and Ricardo Perez-Truglia. 2019. "Housing Expectations and Market Behavior." AEA RCT Registry. July 19. https://www.socialscienceregistry.org/trials/3663/history/50384
Experimental Details

Interventions

Intervention(s)
We will send letters to tens of thousands of individuals who have listed their houses on the market.
Intervention Start Date
2019-06-10
Intervention End Date
2019-07-31

Primary Outcomes

Primary Outcomes (end points)
The primary outcomes correspond to the market behavior of the subjects: the transaction dates and prices.
Primary Outcomes (explanation)
The main outcomes are: i. the time elapsed from the day of letter delivery until the time the property was sold; ii. the price at which the house was sold.
The variable "time until sale" will be right-censored at the time we collect the data (e.g., if a house has not been sold 3 months after the intervention, we do not know if it will be sold 3, 4, 6... months after the intervention). We can use standard methods to account for the censoring. A second (minor) caveat is that some houses may be actively taken off the market. We do not expect this behavior to be common, but if it is then we may need to include it as a separate outcome.

The variable "price sold" will also be censored for some subjects, because the price is observed only if the home is sold. We will use standard methods to deal with this censoring. Additionally, we will have an "intermediate" variable related to the sales price: the listing price. We will use publicly available data to track changes in the list price, and use it as a separate outcome. The advantages of this alternative outcome are: i. It is not subject to censoring; ii. The list price may reflect the price that the seller "hopes" to sell the house for and thus it may be more elastic to housing expectations. However, ex-ante, we do not know if changes in listing price are frequent enough to give us enough statistical power when considered as the outcome variable.

Secondary Outcomes

Secondary Outcomes (end points)
Survey measures of the housing price expectations of the subjects.
Secondary Outcomes (explanation)
We will collect survey data in two ways. First, the letter sent to the subjects will also include a URL to complete an online survey. A sample of this online survey is attached to this registration. This survey includes questions about the expected future median price in the ZIP code of the respondent (1 year ahead and 5 years ahead).

There is a high likelihood that this survey data will not be useful. First, based on similar surveys, the response rate to the online survey is probably going to be very low. Also, it is possible that treatments affect the response rate, thus challenging the internal validity of the analysis of the survey data. For these reasons, we will conduct a complementary survey experiment with an auxiliary sample (Amazon Mechanical Turk workers). In that auxiliary survey, we will not be able to measure the primary outcomes (i.e., market transactions), but we will be able to measure the secondary outcomes (i.e., survey expectations). We will deploy that auxiliary survey at around the same date in which we will send the letters. And we will also attempt to conduct a follow-up survey with this auxiliary sample a month after the baseline survey. Please find attached screenshots of the baseline and follow-up auxiliary surveys.

Experimental Design

Experimental Design
This is an information provision experiment. Subjects will be randomly assigned to different treatment arms, which result in different information being included in the letter.
Experimental Design Details
Subjects are randomly assigned to one of three possible letter types (and to sub-treatments within some of these types). Samples of each letter type (and sub-types) are attached to this application. These letter types are identical in every respect except for the content of the table included in the middle of the first page:

- Present: the current median price of similar homes in the same ZIP Code.
- Future: the current median price of similar homes in the same ZIP Code, as well as a forecast for the price 1 year ahead. Within this letter type, subjects are randomized to one of three sub-treatments. Each sub-treatment corresponds to a different forecast model. Since the estimated future price change can differ across models, a given individual may see a different signals about the future depending on the model that was randomly chosen for them.
- Past: the current median price of similar homes in the same ZIP Code, in addition to information about past prices. Within this letter type, there are two sub-treatments: the past-1 sub-treatment includes information about the price 1 year ago; the past-2 sub-treatment includes information about the prices 1 and 2 years ago.

The main hypothesis of the study is that higher price expectations affect the transaction date and the transaction price (i.e., a seller who expects his or her house to appreciate more should be willing to wait a bit longer to sell the price for a higher price). We cannot manipulate house price expectations directly, but we do it indirectly through the information provision experiments:

- Within the future letter-type the hypothesis is that being randomly assigned to a higher forecast will increase the future home price expectations, resulting in houses that sell later and for a higher price. For a given individual, the three forecasting models result in forecasts of X%, Y% and Z%. The model that the individual is assigned to will determine the forecast that the individual receives. We will use the Instrumental-Variables regression method from Bottan and Perez-Truglia (2018) to exploit the exogenous variation created by the randomization of the forecast model.

- Within the past letter-type the hypothesis is that being shown higher prices changes in the past will increase the future home price expectations, resulting in houses that sell later and for a higher price. The idea is that giving sellers information about past price changes can influence expectations about the future because they extrapolate from past price changes to future price changes. There are two ways of analyzing this data. The most direct way is to use the randomization of sub-treatments. Consider an individual for whom the price changes were X% two years ago and Y% one year ago. If this individual is shown the past-1 letter, she will observe an average past change of Y% (i.e., over the last year); but if she is shown the past-2 letter, she will observe an average annual change of ((X+Y)/2)% (i.e., over the last two years). Thus, the randomization between these two sub-treatments creates exogenous variation in the information shown to the subject. There is an alternative way of analyzing the data, by using the comparison between the "current-type" and "past-type", exploiting heterogeneity in the past price changes in the ZIP code. However, we will only use this strategy if we do not have enough statistical power with the analysis of sub-treatments.

Note that these letter types (future and past) are testing the exact same hypothesis. The only reason why we included two ways of testing the same hypothesis is that individuals may trust one type of information more than the other (i.e., they may be more comfortable extrapolating from past prices than trusting "black-box" forecasts produced by researchers, or vice-versa). Thus, if the results are consistent between these two approaches, we will use an specification that pools them in order to gain statistical power.


The reason why we will send a large number of letters (60,000) is that we expect a majority of our letters to be ignored. For example, some letters may be lost in the mail, thrown away without being opened like most unsolicited mailing, recipients may not read the whole letter or they may read it but they may distrust the information. For all of these reasons, if we find the average effect to be close to zero, the next step would be to ask whether there is at least some groups of individuals who are influenced by the letters. We would use heterogeneity analysis to disentangle this.

The administrative records already contain some information about the subjects (e.g., whether the seller is living in the property or not, the number of days since the property was put on the market). Additionally, we will merge external data on other characteristics of the subjects, the properties and the local housing markets. For most individual characteristics, we will need to use geographic proxies (e.g., the average education at the census block level).

We would look at subgroups of the populations for which, ex-ante, we expect the letters may have a stronger influence. For example, it is possible that the seller's price expectations have a greater influence when the local market is a seller's market than when it is a buyer's market. Another example is that the letters may affect individuals who are selling their primary residencies less, because they may have less flexibility to delay the sale. We will also use the heterogeneity analysis in the auxiliary survey to guide the heterogeneity analysis in the field experiment. For example, if the auxiliary survey suggests that less educated individuals put more weight on the information we give them about the future, we will then test whether the letters had a stronger effect on the behavior of less educated sellers. When looking at these and other sources of heterogeneity, we'll be using standard methods for joint hypotheses testing. Additionally, we will also consider more modern machine learning methods.

Last, there is a secondary (less important) hypothesis: do sellers have systematic biases in their perceptions about past and/or future home prices? To shed light on this hypothesis, we will measure if there a difference in average behavior between individuals who receive the present, past and future letter types.

References
Bottan, Nicolas L. and Ricardo Perez-Truglia (2018), "Choosing Your Pond: Location Choices and Relative Income," NBER Working Paper No. 23615.
Randomization Method
Randomization done in office by a computer.
Randomization Unit
Individual.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
60,000 individuals
Sample size: planned number of observations
We will send 60,000 letters. It is important to note that the letters will take over a week to be delivered. As a result, by the time the letter was delivered, we anticipate that a significant fraction of these 60,000 individuals will have sold their properties already. We will only analyze the effects on the subjects who did not sell their houses by our estimated delivery date (because it is impossible for our letter to affect a transaction that has happened already). We can use the rest of the sample for a falsification test in an event-study fashion.
Sample size (or number of clusters) by treatment arms
Subjects will be randomized into the following letter types: 20% to the "present" type; 30% to the "past" type and 50% to the "future" type.

Within the "past" letter type, subjects will be randomized with equal probability to 1-year or 2-year sub-treatments. Within the "future" letter type, subjects will be randomized with equal probability to one of the three possible forecast sub-treatments.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Supporting Documents and Materials

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
IRB

Institutional Review Boards (IRBs)

IRB Name
Cornell University Institutional Review Board
IRB Approval Date
2018-12-06
IRB Approval Number
1811008440
IRB Name
Institutional Review Board at University of California Los Angeles
IRB Approval Date
2018-11-16
IRB Approval Number
18-001496

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials