Discrimination in the Danish labour market: Belief Elicitation

Last registered on August 22, 2025

Pre-Trial

Trial Information

General Information

Title
Discrimination in the Danish labour market: Belief Elicitation
RCT ID
AEARCTR-0016265
Initial registration date
August 19, 2025

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
August 22, 2025, 5:52 AM EDT

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

Locations

Region

Primary Investigator

Affiliation
University of Milan

Other Primary Investigator(s)

PI Affiliation
University of Copenhagen
PI Affiliation
University of Trento
PI Affiliation
University College London

Additional Trial Information

Status
In development
Start date
2025-08-18
End date
2025-09-30
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
This is a follow-up study to the project registered under AEARCTR-0005301 that uses an RCT in Denmark to estimate a demand curve for native vs. migrant labour and, in doing so, obtain a monetary valuation of customer preferences for different types of workers. The findings are that the demand for a migrant operator is 45 percent lower than demand for a native, but the gap is sensitive to price. It grows larger as the price increases since the demand for a migrant falls more steeply than the demand for a native. This follow-up project studies probabilistic beliefs of a representative sample of Danish households about advertised worker characteristics on the domains of trustworthiness, professionalism, flexibility, diligence, education level, work experience, and language fluency.
External Link(s)

Registration Citation

Citation
Bartos, Vojtech et al. 2025. "Discrimination in the Danish labour market: Belief Elicitation." AEA RCT Registry. August 22. https://doi.org/10.1257/rct.16265-1.0
Experimental Details

Interventions

Intervention(s)
We orthogonally manipulate the name [Peter / Mohammad], price [140kr / 185kr; inflation adjusted to prices in field experiment], and quality [high / low] of domestic cleaning operators using naturally-looking flyers.
Intervention (Hidden)
Intervention Start Date
2025-08-18
Intervention End Date
2025-09-30

Primary Outcomes

Primary Outcomes (end points)
We generate a standardized index across the following belief measures: trustworthiness, professionality, flexibility, trustworthiness in working alone, timeliness, experience, fluency.
Primary Outcomes (explanation)
We construct the index as follows: First, we assume that higher values are considered better (e.g., more flexibility is better than less flexibility; the fluency variable is reverse coded). Second, we calculate weights with a likert scale centered around 0 at the neutral option (for experience, we take the 3-5 years option as equal to 0) according to elicited beliefs for each option for a given variable. Finally, we combine all the variables in a single standardized index. Higher values imply more positive beliefs.

Secondary Outcomes

Secondary Outcomes (end points)
We study the respective probabilistic belief measures separately: trustworthiness, professionality, flexibility, trustworthiness in working alone, timeliness, experience, fluency, education
Secondary Outcomes (explanation)
We examine both mean and variance of probability distributions

Experimental Design

Experimental Design
Sample: A nationally representative (based on gender, age, geography and education) sample of 500 adult Danish residents. Participants are invited to a YouGov survey on decision making.

Experimental manipulation. We elicit participants’ beliefs about the flyers advertising house cleaning operators. We orthogonally manipulate flyer design, name of the worker to signal migration or native background, the hourly price requested, and the rating awarded (using stars) to the operator.

Measures. We elicit probabilistic beliefs about advertised worker characteristics on the domains of trustworthiness (in general and in working alone at clients’ homes), professionalism, flexibility, diligence, education level, work experience, and language fluency. See attached survey material. Each participant evaluates 2 flyers, with different designs and at least one of the three key characteristics manipulated and randomly ordered. This leaves us with 1000 observations in total.

Attention checks. We want to make sure our sample pays attention to the instructions. We use a simple attention check hidden among a set of belief questions that asks respondents to assign a specific probability to a specific response on a likert scale. A similar attention check is implemented also just before the belief elicitation and we alert respondents to an incorrect response, to increase the chance of attention paid in subsequent tasks. Our main analysis will use the attentive sample, while in secondary analyses we study the entire sample.

Comprehension check. After reading the instructions and before answering the questions, respondents get prompted with a comprehension check, asking three questions about probability distributions. Those who fail will see the correct answers highlighted in red. Our main analysis will use the sample of individuals with perfect comprehension (answering all three questions correctly), while in secondary analyses we study the entire sample.

Outcomes.
We generate a standardized index across the following belief measures: trustworthiness, professionality, flexibility, trustworthiness in working alone, timeliness, experience, fluency. We construct the index as follows: First, we assume that higher values are considered better (e.g., more flexibility is better than less flexibility). Second, we calculate weights with a likert scale centered around 0 at the neutral option (for experience, we take the 3-5 years option as equal to 0) according to elicited beliefs for each option for a given variable. Finally, we combine all the variables in a single standardized index. Higher values imply more positive beliefs.

Hypotheses.
Primary hypotheses (standardized beliefs index as outcome):
(1) Given price and quality, a native operator is more positively evaluated than a migrant operator.
1a) As a secondary hypothesis, we also test hypothesis (1) separately by different prices.
(2) Given quality, the size of the gap in positive beliefs between native and migrant operators increases with price (difference-in-differences).
Multiple hypothesis testing:
To address multiple hypothesis testing, we correct p-values using the method developed by Barsbai et al. (2020), which extends the procedure of List et al. (2019) by allowing for corrections in multivariate regression models.

Secondary hypotheses:
We examine the beliefs for each of the eight measures separately by answering the same questions as in the primary hypotheses.
We also study full distributions to understand if it is only means that change by treatment or if it is also variance that changes.

Heterogeneity by Socio-Demographic Characteristics:
We test whether the results vary based on socio-demographic characteristics such as gender, education, age, and income. For continuous variables (e.g., age, income), the sample is split at the median to facilitate subgroup analysis.

Sanity check:
Given name and price, a higher quality operator is more positively evaluated than lower quality operator.

Controls and Robustness Checks:
In all regressions, we control for demographic characteristics. Robustness checks include sensitivity analyses for including or excluding covariates. We also implement a two-step LASSO procedure to select control variables.

Standard Errors:
Standard errors are clustered at the individual level.
Detailed instructions: see attached script
Experimental Design Details
Randomization Method
software-based randomization
Randomization Unit
Respondent level
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
500
Sample size: planned number of observations
1000 flyer evaluations by 500 respondents
Sample size (or number of clusters) by treatment arms
Number of observations by treatment group (remember, 2 per individual):
Native / 140Kr / High quality - N=125
Native / 185Kr / High quality - N=125
Native / 140Kr / Low quality - N=125
Native / 185Kr / Low quality - N=125
Migrant / 140Kr / High quality - N=125
Migrant / 185Kr / High quality - N=125
Migrant / 140Kr / Low quality - N=125
Migrant / 185Kr / Low quality - N=125
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
To detect an effect of 0.25 (a small-sized effect) standard deviations difference in the standardized index of beliefs between native and migrant operator at a given price level (alpha = 0.05 and beta = 0.8), the size of a treatment arm should be around 250 (note that we pool across the two quality conditions in our primary hypothesis). To detect a medium-sized effect of 0.5, a treatment arm of 64 observations is sufficient.
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
Research Ethics Committee at the Department of Economics at University of Copenhagen
IRB Approval Date
2024-10-22
IRB Approval Number
N/A

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