Preferences Toward Tenants and Beliefs about Eviction

Some aspects of survey design are randomized, including ordering of the opponent in altruism game.

We randomize some aspects of the survey design to permit tests of anchoring or order bias by participants. In particular, participants play each altruism game (described below) twice.

We randomize whether the participants first see that the opponent in the altruism game (explained below) is a tenant vs. a landlord.

We also randomize the order in which choices in this game are presented (e.g., whether participants see an even split first vs. second in the elicitation).

2021-11-11

2021-12-31

(1) Measures of altruism/vindictiveness (both samples).
(2) Beliefs about the eviction process (Memphis sample only).

The key outcome comparisons are both the raw values of (1) and (2), as well as the comparisons of the raw values of (1) and (2) to the measures among the Memphis/Shelby County landlords in #AEARCTR-0008053.

In the future, we may conduct tenant samples. In that case, we would then use the data from the external samples to compare to the tenants as well.

In #AEARCTR-0008053, we implement a modified dictator game and elicit beliefs about the eviction process and the likelihood of recouping money from a money judgment (among other activities). We replicate the modified dictator game and elicit beliefs about the eviction process. See experimental design for explanation of how we implement the dictator game and more details about the precise outcomes.

We only estimate beliefs about the eviction process in Memphis among the Memphis sample.

Heterogeneity by:
Whether the participant is a landlord/property manager
Whether the participant rents vs. owns her primary home
Whether the participant has ever been evicted
Income
Political party

We ask participants whether they are landlords/property managers, rent vs. own, and whether they have been evicted. Behavior in an altruism task where the opponent is a landlord or tenant may differ depending on these characteristics. Similarly, knowledge about the eviction process for the beliefs experiment may depend on these characteristics.

Surveys fielded by Lucid, a survey research company. Precise design of the altruism task and beliefs about the eviction process embargoed but in the experimental design.

The surveys are fielded by Lucid, a survey research company.

Altruism task. In #AEARCTR-0008053, we conduct a ‘modified dictator game’ as follows. We elicit the dollar value X that makes the participant indifferent between a bundle of ($X self, $0 other) vs. ($10 self, $10 other). (In practice, we conduct a series of paired comparisons that give bounds on this indifference point.) We implement this task in both of the two samples described above. For each individual, we elicit the indifference point when the opponent ‘other’ is a random tenant whose landlord participated in the survey in #AEARCTR-0008053, and when the opponent ‘other’ is a random landlord who participated in the survey. Note that the opponent will never be named. This permits us to compare the indifference points for random tenants or landlords in these samples to the landlord sample.

The precise outcomes we will focus on will include:
- The average indifference point.
- The difference in indifference points (i.e., comparing the value of the indifference point when the other participant is a landlord vs. a tenant).
- The share of participants who are altruistic (have X > 10), very altruistic (have X > 20), or vindictive (have X < 10) toward tenants and landlords.

We will then compare these outcomes to the landlord survey sample. Consequently outcomes will include the difference in behavior between a given sample (e.g., the Memphis sample) or pooled sample (both Memphis and general sample) and the participants in the landlord survey.

Belief elicitations. In #AEARCTR-0008053, we elicit a variety of beliefs about the eviction process. In this survey, we elicit the following beliefs (taken directly from that other survey):
- Beliefs about the share of tenants who repay an eviction garnishment (on average) in Memphis/Shelby County
- Uncertainty about the above belief
- Beliefs about the average number of money judgments rendered in a set period of time.

We focus on these beliefs (about the general population) since the participants in these studies may not have tenants, so we cannot elicit beliefs about a particular tenant. The purpose of these belief elicitations is to see whether landlords’ beliefs depart from the Memphis population. Are landlords’ beliefs more accurate than a random person’s?

We may conduct survey experiments with tenants in the future. In that case, we will compare behavior among tenants to the general population samples discussed in this registration.

We only conduct these belief elicitations among the Memphis population and not the general population.

Some aspects of the design are randomized (e.g., the survey order, to reduce anchoring) in Qualtrics.

Individual.

No

Memphis sample: 500 individuals
General population sample: 1000 individuals

Memphis sample: 500 individuals General population sample: 1000 individuals

The probability that the first opponent is a landlord is 0.5. The probability that the first opponent is a tenant is 0.5.

Therefore in the Memphis sample:
- In altruism game, first opponent is tenant: ~250 individuals
- In altruism game, first opponent is landlord: ~250 individuals

Whereas in the general population sample:
- In altruism game, first opponent is tenant: ~500 individuals
- In altruism game, first opponent is landlord: ~500 individuals

Numbers are approximate because randomization occurs at the beginning of the altruism task, and some participants will not complete the survey.

Similarly, across each altruism elicitation, the choice of the even split of the gift card is observed first in the multiple price list with probability = 0.5. For instance, in the game in which the opponent is a tenant, the participant observes an even split of the gift card first in the multiple price list with probability = 0.5; similarly, in the game in which the opponent is a landlord, the participant observes an even split of the gift card first in the multiple price list with probability = 0.5. This will yield identical approximate sample sizes as above, for each of these sub-treatments.