Associative Memory and Overreaction in Expectations

Last registered on January 09, 2020

Pre-Trial

Trial Information

General Information

Title
Associative Memory and Overreaction in Expectations
RCT ID
AEARCTR-0004247
Initial registration date
May 28, 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
June 10, 2019, 5:23 PM EDT

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

Last updated
January 09, 2020, 4:09 PM EST

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Region

Primary Investigator

Affiliation

Other Primary Investigator(s)

PI Affiliation
PI Affiliation

Additional Trial Information

Status
In development
Start date
2019-05-31
End date
2020-12-31
Secondary IDs
Abstract
Laboratory experiments to study the effect of (associative) memory on expectation formation.
External Link(s)

Registration Citation

Citation
Enke, Benjamin, Frederik Schwerter and Florian Zimmermann. 2020. "Associative Memory and Overreaction in Expectations." AEA RCT Registry. January 09. https://doi.org/10.1257/rct.4247-3.3000000000000003
Former Citation
Enke, Benjamin, Frederik Schwerter and Florian Zimmermann. 2020. "Associative Memory and Overreaction in Expectations." AEA RCT Registry. January 09. https://www.socialscienceregistry.org/trials/4247/history/60324
Experimental Details

Interventions

Intervention(s)
Laboratory experiments to study effect of associative memory on expectation formation
Intervention Start Date
2019-05-31
Intervention End Date
2020-12-31

Primary Outcomes

Primary Outcomes (end points)
The study contains two outcome variables, explained in the attachment:

• Subjects’ second belief, i.e., their guess about the value of a company in Part 2.
• Subjects’ recall, i.e., the number of positive and negative signals that they recall having seen. We combine these recall data into one measure by computing the difference between the number of positive signals and the number of negative signals that a subject reports to recall. To make this measure directly comparable with the beliefs data, we normalize it so that
Recall measure = (# positive signals recalled - # negative signals recalled) * 10
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Different treatment variations to manipulate subjects' memory constraints when they forecast the value of hypothetical companies.
Experimental Design Details
Also see attachment.

I. DATA COLLECTION:

In the BonnEconLab at the University of Bonn, we conduct laboratory experiments. The basic structure of the experimental setup is as follows. Subjects are asked to estimate the stock market value of 12 hypothetical companies. Each company has a baseline value of 100, and subsequently receives multiple pieces of news that either imply an increase in value of +10 or a decrease in value of -10. The value of the company is a deterministic function of the news and is given by 100 plus the sum of all news that a subject observes throughout the experiment.

The experiment consists of the following stages:

Part 1:
• For each of the 12 companies, a subject first observes a sequence of news (+10 or -10 each). The news appear sequentially on a subject’s computer screen. The number of signals per company is 0, 1, 2, or 3.
• The subject is then asked to guess the current value of the company (we will henceforth refer to this as “first belief”). This belief is financially incentivized through a binarized scoring rule.
• This procedure is repeated for all twelve companies.
• The subject completes real effort tasks for 15 minutes.

Part 2:
• A subject potentially observes a final piece of news about the value of a company.
• The subject then guesses the value of the company, where the true value is given by 100 plus the sum of all news in the first and second part combined (we will henceforth refer to this as “second belief”). This belief is financially incentivized through a binarized scoring rule.
• The subject is asked to recall how many positive and negative news they have seen for this particular company throughout the entire experiment (we will henceforth refer to this as “recall”). These recall data are not financially incentivized.
• This procedure is repeated for all twelve companies, except that for two companies no final signal is observed.

Part 3: Questionnaire and Raven IQ test. 


II. TREATMENTS:

1. Treatment Main

In treatment Main, the pieces of news are communicated on subjects’ computer screens along with a context. The context consists of a story and an image. Importantly, there is a one-to-one mapping between type of news for a given company and context. That is, every positive news for company A is communicated with the same context (image and story). Likewise, every negative news for company A is communicated with the same context (albeit a different one than the positive news of course). The same logic holds for all other companies. Thus, it can never happen that a context is communicated with news for different companies, or with both positive and negative news. A context deterministically identifies a piece of news.

In this treatment, the time lag between the first and second part of the experiment is given by 15 minutes, i.e., the time it takes subjects to complete the real-effort tasks.


2. Treatment Reminder

In treatment Reminder, the setup is exactly the same as in Main, except that at the beginning of Part 2 of the experiment (i.e., before a subject observes the last signal for a company), a subject is reminded of their own first belief from the first part of the experiment.


3. Treatment No cue

In treatment No cue, the setup is exactly the same as in Main, except that each piece of news is communicated with a different context. That is, a given context (image and story) never appears twice, even if the company and type of news is identical.


4. Treatment Time lag

In treatment Reminder, the setup is exactly the same as in Main, except that the time lag between the first and second part of the experiment is longer, and given by three days. Subjects complete Part 1 of the experiment on day 1, including the real-effort task. Then, subjects come back into the lab three days later to complete the second part.

In case a subject fails to show up for the second day, we follow up with additional email invitations that offer subjects an opportunity to complete the experiment at a later point in time.



5. Treatment Reminder time lag

In treatment Reminder with time lag, the setup is exactly the same as in Reminder, except that the time lag between the first and second part of the experiment is longer, and given by three days, as in treatment Reminder. All comments from above about treatment Reminder also apply here.


IV. NATURE OF ANALYSES
We analyze our experimental data by means of OLS regressions:
• The main dependent variable is given by a subject’s second belief.
• The ancillary dependent variable is a subject’s recall measure discussed above.
• The main independent variable is the realization of the last signal (-10 or +10), as well as its interaction with various treatment dummies per the discussion below.
• Because we have multiple observations per subject (10), we cluster the standard errors at the subject level.


V. HYPOTHESES:

1. Overreaction in expectations

Restrict attention to treatment Main. Take as dependent variables (i) a subject’s second belief as well as (ii) their recall measure. Regress each of these on the value of the last signal, controlling for the value of the company after the first period. We hypothesize that in both cases the OLS coefficient is significantly larger than one.

2. The role of memory

Restrict attention to treatments Main and Reminder. Take as dependent variables (i) a subject’s second belief as well as (ii) their recall measure. Regress each of these on (i) the value of the last signal, (ii) a treatment dummy, and (iii) the interaction of the value of the last signal and a treatment dummy (where treatment Main is coded as 1), controlling for the value of the company after the first period. We hypothesize that in both cases the OLS coefficient of the interaction term is significantly larger than zero.

3. The role of associative memory

Restrict attention to treatments Main and No cue. Take as dependent variables (i) a subject’s second belief as well as (ii) their recall measure. Regress each of these on (i) the value of the last signal, (ii) a treatment dummy, and (iii) the interaction of the value of the last signal and a treatment dummy (where treatment Main is coded as 1), controlling for the value of the company after the first period. We hypothesize that in both cases the OLS coefficient of the interaction term is significantly larger than zero.


4. Variation in the number of cued signals

Restrict attention to treatment Main. Take as dependent variables (i) a subject’s second belief as well as (ii) their recall measure. Regress each of these on (i) the value of the last signal, (ii) the number of signals in the first part that took on the same value as the signal in the second part, and (iii) a corresponding interaction term, controlling for the value of the company after the first period. We hypothesize that in both cases the OLS coefficient of the interaction term is significantly larger than zero.


5. Longer time lag

Restrict attention to treatments Time lag and Reminder time lag. Take as dependent variables (i) a subject’s second belief as well as (ii) their recall measure. Regress each of these on (i) the value of the last signal, (ii) a treatment dummy, and (iii) the interaction of the value of the last signal and a treatment dummy (where treatment Time lag is coded as 1), controlling for the value of the company after the first period. We hypothesize that in both cases the OLS coefficient of the interaction term is significantly larger than zero.



VI. HETEROGENEITY ANALYSES
Restrict attention to treatment Main. Take as dependent variables (i) a subject’s second belief as well as (ii) their recall measure. Regress each of these on (i) the value of the last signal, (ii) a measure that is of interest for heterogeneity analyses, and (iii) a corresponding interaction term. The coefficient of interest is the interaction term.

We use the following variables for these heterogeneity analyses:
• Subjects’ score on a Raven IQ test that is conducted as part of the final questionnaire.
• A proxy for the strength of memory that is estimated from the recall data in the following way: for those signals that do not get cued (i.e., that do not equal the last signal in the second period), we regress reported recall of these signals on the true number of signals in the first period, controlling for the belief in the first period. The regression coefficient of the true number of signals is the proxy for the strength of memory.
• Subjects’ response time.

VII. EXCLUSION CRITERIA
By design of the experiment, we exclude subject-company observations where the subject did not receive a final signal about a company in the second part of the experiment. These subject-company observations are only meant to identify the relationship between second belief and first belief in case of absence of further signals. Potential differences between second and first belief allow us to verify that subjects do not (or not perfectly) recall their first belief when stating their second belief.
After subjects read the experimental instructions, they answer a series of comprehension questions. In case a subject makes more than one mistake, they are excluded from the experiment.
Randomization Method
Random draw of numbered card
Randomization Unit
Individual
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
n/a
Sample size: planned number of observations
340
Sample size (or number of clusters) by treatment arms
Treatments are randomized within session according to the following logic:
• Block A: Treatments Main, Reminder, No cue
• Block B: Treatments Time lag, Reminder time lag

The sample size will be given by:
• Treatments Main, No cue, Time lag: 80 subjects each
• Treatments Reminder, Reminder time lag: 50 subjects each
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
Harvard IRB
IRB Approval Date
2019-01-09
IRB Approval Number
IRB16-1753
Analysis Plan

Analysis Plan Documents

Memory_PAP_Part4_2019-11-19.docx

MD5: 0e73759b046189decba3fc86a956071c

SHA1: e0e3061e0f8829f9edc605a0c42a8dd98a0747b1

Uploaded At: November 19, 2019

Memory_PAP_2019-05-17.docx

MD5: 3c66152ff15185ff6e8c983cdc4af453

SHA1: e69680942980a4e96aa5d7b77a952e593c991e81

Uploaded At: May 28, 2019

Memory_PAP_Part2_2019-07-18.docx

MD5: 1cd7d3ded406bd3352620e1be600bb03

SHA1: 51ad23b896073c184a5abf8fe44d7f69dc906e4a

Uploaded At: July 18, 2019

Memory_PAP_Part3_2019-10-01.docx

MD5: 2023939bdfcfffd3c20cad852bbd0027

SHA1: aebbf9772bf75ef6dcdcff3ee10f914303f0487a

Uploaded At: October 01, 2019

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