Back to History Current Version

Causal Evidence on Expectation Formation and the Role of Reactive vs Proactive Learning

Last registered on May 09, 2021

Pre-Trial

Trial Information

General Information

Title
Causal Evidence on Expectation Formation and the Role of Reactive vs Proactive Learning
RCT ID
AEARCTR-0006860
Initial registration date
December 22, 2020

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
December 23, 2020, 6:43 AM EST

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

Last updated
May 09, 2021, 5:07 PM EDT

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

Locations

Primary Investigator

Affiliation
David Eccles School of Business, University of Utah

Other Primary Investigator(s)

PI Affiliation
David Eccles School of Business, University of Utah
PI Affiliation
David Eccles School of Business, University of Utah

Additional Trial Information

Status
In development
Start date
2021-03-01
End date
2022-03-01
Secondary IDs
Abstract
We contrast two fundamental ways to learn and form expectations for businesses. On the one hand, we will provide evidence on firms’ ability to adapt through feedback, sometimes also called “local search” or “adaptive expectations”. This approach primarily reactive in nature.
On the other hand, we will investigate the effectiveness of proactive learning and experimentation, an approach that is connected to concepts such as “hypothesis-driven search” and “anticipatory expectations”.
External Link(s)

Registration Citation

Citation
Gaulin, Maclean, Nathan Seegert and Mu-Jeung Yang. 2021. "Causal Evidence on Expectation Formation and the Role of Reactive vs Proactive Learning." AEA RCT Registry. May 09. https://doi.org/10.1257/rct.6860-1.3000000000000003
Sponsors & Partners

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

Request Information
Experimental Details

Interventions

Intervention(s)
Intervention (Hidden)
Respondents previously agreed to participate in business surveys run by the University of Utah, which ask questions about business sentiment. In this context, we will provide the following two treatments to a random subset of survey participants. Each treatment consists of a set of additional questions in the survey that participants are free to either answer or ignore. These additional questions will appear every month for the upcoming month and every month for the past month. We will randomize treatments initially and then persistently bring up the treatment questions in ongoing surveys.

TREATMENT 1: Reactive Learning Treatment
PRE-PERIOD TREATMENT 1
Are your targeted revenues for the next month different from your best guess of you expected revenues? If so, what are your targeted revenues over the next month last year?
POST-PERIOD TREATMENT 1
Last month you predicted that your revenues would be [X] over the coming month. You reported a revenue of [Y] for the last month. This implies a forecast error of [X/Y]*100 % (Positive errors mean you overestimated growth, negative errors mean you underestimated growth). If your forecast error was more negative than -10% or more positive than +10%, what is the most likely reason for this deviation from your goal?

TREATMENT 2: Proactive Learning Treatment
Are your targeted revenues for the next month different from your best guess of you expected revenues? If so, what are your targeted revenues over the next month last year?
This section will ask you to specify how exactly you intend to achieve your set revenue growth goal in the previous question. We will obtain your permission in advance if we are interested in using anonymized quotes in scientific publications. As a reminder, your data will remain completely confidential and will not be released in any way that can be linked to you. Data from this study will be kept locked or password-protected, and will be destroyed when no longer needed for research purposes.

PRE-PERIOD TREATMENT 2
(1) A "competitive advantage" is a strength your company has, which distinguishes you from your competitors and which is hard to copy.
Often such “competitive advantage” results from exploring previously untested ideas. Do you have a unique idea or belief that differs from “conventional wisdom” in your industry? If you hold such a contrarian belief, what is it and how could it help with your growth goal?

For example, you might own a sandwich shop and no other sandwich shop in your neighborhood might offer breakfast, because “conventional wisdom” is that there is not enough foot-traffic in the morning. A contrarian belief might be that many office workers are open to purchasing breakfast, but do not currently do so, because they want to avoid fatigue after eating a heavy and unhealthy breakfast sandwich.

(2) What are the most important problems that prevent your unique idea from being realized? Put differently, what are the reasons your belief is contrarian instead of being widely accepted in your industry?

In the sandwich shop example, among the problems preventing you from offering breakfast is that you do not know demand by office workers for healthy breakfast options. Another problem might be that office workers do not know that healthy breakfast options are available for purchase.

(3) Please list two possible plans that might solve the problems that prevent your unique idea from being realized and which can help with your growth goals. Ideally, these two plans would be two different ways that help you solve a problem that other competitors in your market are not solving.

We recommend that these two plans include: (1) What customer or market segment you will target (2) A list the activities that you will use to deliver the intended results, (3) Optional: What advantage you intend to use or create to solve the problem

In the sandwich shop example, one plan might be to offer healthy breakfast smoothies with caffeine, which prevent customers from being tired after breakfast. The targeted customer segment are nearby office workers, which are more likely to be repeat customers. To deliver such smoothies you would need equipment and freshly purchased ingredients. One potential advantage might be your knowledge of tasty smoothy recipes.

(4) What would have to be true for each of the two plans you listed in the last question, to achieve your growth goal for the next month?

For each of the two plans, please make up a list of conditions, which you could potentially observe, and that can either assure you that your plan worked or make you confident that the plan did not work. Such a list of conditions can enable you to pay attention to the relevant business information in a targeted way and more accurately learn from your experiences.

One way to express this is an IF-THEN statement: IF your conditions are met, THEN your profit increases because of the problem the plan solves.

In the sandwich shop example, your condition might be "IF I can at least attract 45 office workers at $5 per breakfast smoothie every weekday, THEN offering breakfast is profitable". One way this condition might fail is that there are not enough office workers interested to purchase breakfast smoothies every weekday.

(5) For each of the conditions you specified in the previous question, how would you test whether this condition is true?

A "test" involves figuring out if the underlying REASON your plan works is correct or incorrect, just like a “business scientist” would. Understanding the reason your plan works can be important to ensure that you can repeat your success and do not rely on “luck”. It will also ensure you that you solved the problem that prevents other firms from doing the same.

Let’s return to the sandwich shop example with the condition "IF I can at least attract 45 office workers at $5 per breakfast smoothie every weekday, THEN offering breakfast is profitable". Your test might involve offering healthy breakfast smoothies with caffeine and advertise these healthy options in neighboring office buildings. Keeping track of how many of your breakfast smoothie customers are office workers and how many of your office workers are repeat customers can then tell you if you can repeat your success.

For more detail, see this article (which will open in a new tab and not interrupt your survey responses on this tab).

POST-TREATMENT 2
(6) Last month we asked you to come up with two alternative plans that might help you meet your growth target. We also asked you to specify "what would have to be true", for these two plans to succeed and to come up with ways to test whether these conditions are true for your business. Did you have an opportunity to conduct a test of the "what would have to be true" conditions?
(7) Last month you predicted that your revenues would be [X] over the coming month. You reported a revenue of [Y] for the last month. This implies a forecast error of [X/Y]*100 % (Positive errors mean you overestimated growth, negative errors mean you underestimated growth). If your forecast error was more negative than -10% or more positive than +10%, what is the most likely reason for this deviation from your goal and how do you know that this is the most likely reason?
Intervention Start Date
2021-03-01
Intervention End Date
2022-03-01

Primary Outcomes

Primary Outcomes (end points)
Revenue forecasts, realized revenues, accuracy of revenue forecasts, firm growth, firm profitability, firm growth path, hiring, intended hiring, elasticity of intended and realized hiring in response to revenue changes, firm productivity, new product development, measures of strategic initiatives from text data, measures of overconfidence (overestimation, overprecision and overplacement), measures of misattribution bias, measures of digital technology adoption, measures of different types of perceived risk, measures of operating leverage, local returns to scale and (variable and fixed) costs
Primary Outcomes (explanation)
One outcome that needs to be constructed are measures of strategic initiatives from text responses for which we will use natural language processing techniques, such as Neural Language Networks as well as simple count models such as TF-IDF.

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Treatments encourage participants to learn from past outcomes as well as provide structured ways to think about strategies to realize set goals.
Experimental Design Details
We will use a cumulative treatment design. Treatment 1 includes information on past forecast errors, while treatment 2 will include treatment 1 and add encouragements to think about strategic initiatives in a structured way.
Randomization Method
Computer-generated random numbers
Randomization Unit
Firm-level
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
1000
Sample size: planned number of observations
1000 firms, 12 observations per firm. However, we will replenish firms to offset sample attrition. We will run the treatments in the replenished firms.
Sample size (or number of clusters) by treatment arms
333 in control group, 333 firms in treatment 1, 333 firms in treatment 2
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB

Institutional Review Boards (IRBs)

IRB Name
Institutional Review Board University of Utah
IRB Approval Date
2020-12-22
IRB Approval Number
00139342

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