An experiment on robo-advising

Last registered on April 12, 2022

View Trial History

Pre-Trial

Trial Information

General Information

Title

An experiment on robo-advising

RCT ID

AEARCTR-0009159

Initial registration date

April 08, 2022

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

April 12, 2022, 8:14 AM EDT

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

Locations

Country

United Kingdom of Great Britain and Northern Ireland

Region

Primary Investigator

Name

Jörg Oechssler

Affiliation

Universität Heidelberg, AWI

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Marco Lambrecht

PI Affiliation

Hanken School of Economics

Contact Investigator

PI Name

Simon Weidenholzer

PI Affiliation

University of Essex

Contact Investigator

Additional Trial Information

Status

In development

Start date

2022-04-15

End date

2022-09-16

Keywords

Behavior, Finance & Microfinance

Additional Keywords

JEL code(s)

Secondary IDs

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

The current projects seeks to experimentally analyse robo advising and trading. To this end we will first elicit risk preferences and financial literacy of subjects. In a next step subjects will be given the opportunity to participate in experimental financial markets. In some of these markets, they will have the opportunity to receive financial advice on optimal portfolio choice from an algorithm, which is programmed to match risk preferences, and in other markets may delegate their decisions to such an algorithm.

External Link(s)

Registration Citation

Citation

Lambrecht, Marco, Jörg Oechssler and Simon Weidenholzer. 2022. "An experiment on robo-advising." AEA RCT Registry. April 12. https://doi.org/10.1257/rct.9159-1.0

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)

There are three treatments:
1) CONTROL: Subjects have to make the investment decisions on their own
2) SOFT-ROBO: Subjects have available a "robo advisor" which gives them non-binding investment advise
3) HARD-ROBO: Subjects have available a robo-advisor which makes investment decisions for them unless it is overridden by subject

Intervention Start Date

2022-04-16

Intervention End Date

2022-09-15

Primary Outcomes

Primary Outcomes (end points)

1) Market participation
2) Optimality of investment

Primary Outcomes (explanation)

1) is measured in two ways:
a) % of subjects who agree to participate in financial market study
b) % of subjects who supply a valid email address for participation
2) is measured by comparing the optimal investment (calculated by using a CRRA utility function with parameters according to risk elicitation task) to actual investment:
a) how often are redundant assets used?
b) how often are funds rebalanced?
c) As a measure of optimality we plan to compare the certainty equivalent of the optimal portfolio to that of the chosen portfolio and to compare deviations in risk taking in stage 2 from revealed benchmarks in stage 1.

Secondary Outcomes

Secondary Outcomes (end points)

Attrition rates will be used to measure continued and persistent engagement with financial markets.

Secondary Outcomes (explanation)

Experimental Design

Experimental Design Details

Our proposed experiment will run on the online platform Prolific using their subject pool, recruiting subjects from the UK. Our experiment will consist of two parts. In the first part we will elicit risk preferences using the Binswanger-Eckel-Grossman procedure and will measure financial literacy in tests akin to those outlined in Holzmeister et al (2019). The first part will also encompass questions on general numeracy and socio-economic background. Following this, subjects will be asked whether they would like to participate in an experimental financial market. There will be three treatments: In Treatment SOFT-ROBO, subjects will be offered financial advice given by an algorithm. In treatment HARD-ROBO subjects may fully delegate their decision to this algorithm, and in treatment CONTROL subjects will neither receive investment recommendations, nor can they delegate decisions to an algorithm. For subjects who choose not to participate in the financial market stage, the experiment ends after the first stage.

In the second part of the experiment, subjects trade in a simulated financial market. Subjects will start with a virtual endowment of 1000GBP. Trading will take place over ten periods, lasting a week each. In each trading period subject will have to choose how much of their portfolio to invest into different assets, varying in their relative risk and return. Returns will be randomly drawn in each period.
Subjects will receive regular updates on their portfolio via email along with a reminder to make their trading decisions for the next round. Depending on the treatment subjects either have access to robo advice, delegate all decision to the algorithm or trade on their own. Decisions of the algorithm and the nature of advice will depend on the risk attitudes previously elicited. At the end of the 10 trading
periods we will randomly draw three subjects to be paid the final value of their portfolio. Finally, there will be questions about the subjects’ experience of trading.

Randomization Method

Done by Prolific

Randomization Unit

individual

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

1000 subjects

Sample size: planned number of observations

1000 subjects

Sample size (or number of clusters) by treatment arms

333 subjects per treatment

If there is no significant difference between SOFT-ROBO and HARD-ROBO, we will pool the data to obtain 666 combined ROBO observations.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Assuming a participation rate of 10% in CONTROL, the minimum detectable effect size would be 7.5 percentage points (similar when assuming a participation rate of 80%). When comparing combined ROBO data to CONTROL the minimum detectable effect size (assuming no differences between SOFT-ROBO and HARD-ROBO) is 6.4 percentage points. Assuming a participation rate of 50% in CONTROL, the minimum detectable effect size would be about 10 percentage points.

Supporting Documents and Materials

IRB