Nudge theory to reduce workplace accidents in the construction sector

Last registered on February 04, 2026
View Trial History

Pre-Trial

Trial Information

General Information

Title
Nudge theory to reduce workplace accidents in the construction sector
RCT ID
AEARCTR-0017543
Initial registration date
December 22, 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
February 04, 2026, 9:33 AM EST

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
Françoise Delmez
Affiliation
Université Libre de Bruxelles
Contact Primary Investigator

Other Primary Investigator(s)

PI Name
Ilan Tojerow
PI Affiliation
Université Libre de Bruxelles
Contact Investigator
PI Name
Justine Carlier
PI Affiliation
Université Libre de Bruxelles
Contact Investigator
PI Name
Magali Verdonck
PI Affiliation
Université Libre de Bruxelles
Contact Investigator

Additional Trial Information

Status
On going
Start date
2024-04-30
End date
2026-12-31
Keywords
Behavior, Health, Labor
Additional Keywords
Workplace safety, Occupational accidents, informational intervention
JEL code(s)
J81: Working conditions; J28 – Safety; Job Satisfaction; Related Public Policywhat; C93 – Field Experiments; D91 – Intertemporal Choice; Behavioural Aspects
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Workplace accidents are disproportionately frequent in the construction sector, with falls from height representing a major cause of fatal and permanently disabling injuries in Belgium. This paper evaluates whether low-cost behavioural interventions targeting employers can reduce the incidence and severity of workplace accidents. We conduct a large-scale randomised controlled trial among 19,372 Belgian construction firms, testing informational nudges that (i) remind employers of their legal obligations regarding fall-prevention equipment, (ii) warn them about the costs of non-compliance, and (iii) combine both messages. The experiment also varies the communication format, comparing text-based messages with short videos.

Companies are randomly assigned to treatment by researchers based on company unique identifier, with randomisation stratified by firm size. Treatments are delivered through repeated emails and an official digital mailbox (e-Box). Using linked administrative social security data and comprehensive workplace accident records, we estimate intention-to-treat effects on accident rates per full-time equivalent, accident severity, and duration of work incapacity. Secondary analyses examine potential mechanisms, including changes in workforce composition and accident types. The study provides causal evidence on the effectiveness of reminder and warning messages in improving occupational safety and informs the design of behavioural policies in high-risk sectors.
External Link(s)

Registration Citation

Citation
Carlier, Justine et al. 2026. "Nudge theory to reduce workplace accidents in the construction sector." AEA RCT Registry. February 04. https://doi.org/10.1257/rct.17543-1.0
Sponsors & Partners
Experimental Details

Interventions

Intervention(s)
A randomised controlled trial (RCT) in the construction sector will be conducted to assess the effectiveness of two nudges and a combination of the two nudges aimed at fostering employers to implement protective equipment to avoid fall on construction sites. Companies were randomly assigned to one of the treatment arms or to the control group by researchers, using RandTreat command from Stata based on compnay unique identifier. The various treatment arms were designed to allow disentangling effect linked to compliance to legal obligations only from internalisation of information on the private cost of non-compliance. In addition, specific treatment arms were designed to test whether the communication format is a relevant factor, comparing the effect of short videos versus text.
Intervention Start Date
2025-09-22
Intervention End Date
2025-11-03

Primary Outcomes

Primary Outcomes (end points)
- The number of accidents per company adjusted by the size of the company (number of accidents/number of full-time equivalents) (“Outcome 1”). The probability of having at least one accident, as an alternative way to test the impact on the number of accidents.
- The number of severe accidents adjusted by the total number of accidents in the company (number of accidents followed by death, a permanent disability or a temporary incapacity of more than 180 days/number of accidents). (“Outcome 2”)
- The average duration (in days) of the temporary work incapacity by accident, as an alternative way to test the impact on severity. (“Outcome 3”)
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
- Effects through workforce composition of firms. Accidents tend to be more frequent among younger workers relative to their share in the workforce. The analysis will hence test whether companies adapt their workforce composition towards workers relatively less exposed to risk of accidents. The analysis will also test whether companies react by changing the type of contracts offered to workers (temporary, interim, or indefinite duration).
- Effects through the distribution of accidents. The analysis will test whether bringing focus (through the nudges) on “fall from heights” lowers the focus and increases accidents resulting from other causes, or if in general a redistribution of accidents from those with severe consequences to accidents with less severe consequences.
- Effects by treatment arm. The analysis will test whether the three nudges have a statistically equal effect on accidents and their severity, hence testing the relative importance for behavioural changes of recall of legal obligations as opposed to warning on costs, or a combination of both.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The study implements a large-scale randomised controlled trial to assess whether behavioural nudges directed at employers reduce workplace accidents in the Belgian construction sector. The population of interest consists of construction firms with legal entity status operating in Belgium. The experimental sample includes 19,372 firms that have a registered email address and an active official digital mailbox (e-Box).

Firms are randomly assigned to one of six treatment groups or to a control group, randomisation is done by the researchers using the RandTreat command in Stata. Companies do not know that they belong to a treatment or control group. Randomisation is stratified by firm size using five strata defined by the number of full-time equivalent workers. The control group receives no communication. Treated firms receive informational messages aimed at encouraging compliance with fall-prevention regulations. The interventions test three types of nudges: (i) a reminder of legal obligations to provide protective equipment against falls from height, (ii) a warning highlighting the economic and legal costs of non-compliance, and (iii) a combination of both messages. In addition, the experiment varies the communication format, comparing text- and image-based messages with short video messages.

Each treated firm receives the same nudge three times over the intervention period—two messages via email and one via the official e-Box. Messages are delivered in the firm’s official language (French or Dutch). A reminder is sent if the e-Box message is not opened within two weeks. The intervention targets employers rather than individual workers.

Outcomes are measured using linked administrative social security data and exhaustive workplace accident records. The primary outcomes include the number of accidents per full-time equivalent, measures of accident severity, and the duration of work incapacity following accidents. Effects are estimated using an intention-to-treat framework over multiple post-intervention horizons, with additional analyses exploiting variation in message opening to estimate local average treatment effects.
Experimental Design Details
Not available
Randomization Method
Stata randtreat command implemented by researchers.
Randomization Unit
company, within clusters by company size.
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
5 based on company size
Sample size: planned number of observations
19,372 companies
Sample size (or number of clusters) by treatment arms
(Cluster) (Number of companies) (Number of full-time equivalents)
Stratum n°1 (0-5 FTEs) 14,049 25,381.5
Stratum n°2 (6-10 FTEs) 2,562 18,140.9
Stratum n°3 (11-25 FTEs)1,822 27,801.2
Stratum n°4 (26-100 FTEs) 768 33,371.6
Stratum n°5 (More than 100 FTEs) 171 40,906.4
Total 19,372 145,602.0
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
See pre-analysis plan attached.
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
Comité d’éthique de la recherche de l’ULB
IRB Approval Date
2024-12-16
IRB Approval Number
n/a
Analysis Plan

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

Request Information