Shallow Inference and Hiring

Last registered on May 18, 2026

View Trial History

Pre-Trial

Trial Information

General Information

Title

Shallow Inference and Hiring

RCT ID

AEARCTR-0018615

Initial registration date

May 12, 2026

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

May 18, 2026, 4:27 AM EDT

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

Locations

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

Request Information

Primary Investigator

Name

Christiane Buschinger

Affiliation

JGU Mainz

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Julian Detemple

PI Affiliation

LMU Munich

Contact Investigator

PI Name

Maria Guadalupe

PI Affiliation

INSEAD

Contact Investigator

Additional Trial Information

Status

In development

Start date

2026-05-13

End date

2027-12-31

Keywords

Behavior, Firms & Productivity, Labor

Additional Keywords

JEL code(s)

Secondary IDs

Prior work

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

Abstract

This study investigates "shallow inference" in hiring evaluations, defined as the failure of observers to account for the difference in socio-economic constraints behind observed performance signals. Using a vignette experiment with MBA and executive MBA students acting as evaluators, the research explores how markers of socio-economic status (SES) influence beliefs regarding underlying ability and organizational fit.
The design utilizes several rounds of relative assessments varying SES indicators such as first-in-family university status and stereotyped extracurricular activities. The study tests whether evaluators naturally correct for SES-driven disadvantages or if explicit contextual information regarding economic constraints and challenges that the applicant had to overcome is required to shift these underlying beliefs. Furthermore, the project examines to which extent overcoming these challenges compensates worse credentials. By leveraging open-ended reasoning data, the study aims to identify the mechanisms driving systematic evaluation biases and the potential for informational interventions to mitigate inequality in elite professional contexts.

External Link(s)

Registration Citation

Citation

Buschinger, Christiane, Julian Detemple and Maria Guadalupe. 2026. "Shallow Inference and Hiring." AEA RCT Registry. May 18. https://doi.org/10.1257/rct.18615-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Intervention Start Date

2026-05-13

Intervention End Date

2027-12-31

Primary Outcomes

Primary Outcomes (end points)

Underlying ability, organizational beliefs, hiring recommendations, and reasoning data.

Primary Outcomes (explanation)

The main outcome variables are the underlying ability and organizational fit beliefs. These beliefs are each measured on a 5-point qualitative scale: (-2) Much higher for Candidate X, (-1) Higher for Candidate X, (0) The same, (1) Higher for Candidate Y, and (2) Much higher for Candidate Y. In the case of organizational fit, the scale uses "Much better" and "Better" respectively.

For the main analyses, we transform these beliefs into a binary variable equal to 1 if the belief is “higher” for candidate X (-2 or -1) for ability. For organizational fit, the values is “1” if the beliefs is “lower” for candidate X (2 or 1). [Candidate X represents the low-SES candidate.]

In further analyses and robustness checks we use alternative specifications for the belief outcomes (see attached pre-analysis plan for details).

Another outcome is a hypothetical hiring recommendation which is elicited as a categorical choice regarding which candidate the subject would recommend for hire. Subjects choose from three options: (1) Recommend Candidate X, (2) Indifferent, and (3) Recommend Candidate Y. For the main analyses, we again transform this into a binary variable with the value “1” if Candidate X is recommended (with candidate X representing the low-SES candidate).

Finally, we employ reasoning data as another outcome. This is unstructured open-ended reasoning data from a follow-up prompt to explain ability and organizational fit assessments (separately). This will be coded using large language models as well as research assistants (for a subsample to check LLM performance) based on a researcher-defined codebook.

Secondary Outcomes

Secondary Outcomes (end points)

Secondary Outcomes (explanation)

Experimental Design

We conduct a vignette study where subjects act as strategy consulting hiring managers to evaluate pairs of candidates on underlying ability and organizational fit. Participants complete ten rounds of relative assessments based on short resumes and occasional personal statements. Three key rounds isolate the impact of socio-economic status (SES) through specific markers like "first-in-family" status and stereotyped extracurricular activities. A Baseline round presents candidates differing only in these SES markers, while a Treatment round adds a personal statement detailing the low-SES applicant's financial constraints and higher signal acquisition costs. An Endline round tests to which extent a similar note as in the treatment round compensates for worse qualifications on paper for the low-SES candidate, while seven interspersed obfuscation rounds mask the study's primary focus and gauge reactions to standard resume features. For three assessments, we also elicit open-ended reasoning and hypothetical hiring recommendations for key rounds. We conclude with the collection of respondent background data.

Experimental Design Details

Not available

Randomization Method

Randomization done by a computer in an online survey.

Randomization Unit

Individual

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

Not applicable, since no clustered treatment assignment.

Sample size: planned number of observations

We do not have a fixed target sample size. Data will be collected in several waves, and we do not yet know how many observations or waves will be possible, as this is determined by the schedule of the MBA courses and how many MBA students give their consent for the research use of their data.

Sample size (or number of clusters) by treatment arms

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