x

Please fill out this short user survey of only 3 questions in order to help us improve the site. We appreciate your feedback!
The Hidden Cost of Training
Last registered on December 31, 2018

Pre-Trial

Trial Information
General Information
Title
The Hidden Cost of Training
RCT ID
AEARCTR-0002711
Initial registration date
February 02, 2018
Last updated
December 31, 2018 11:35 AM EST
Location(s)
Region
Primary Investigator
Affiliation
University of Cologne
Other Primary Investigator(s)
Additional Trial Information
Status
Completed
Start date
2018-02-05
End date
2018-12-30
Secondary IDs
Abstract
In this project, I study the effect of training participation on the reference point with respect to a fair wage and how the subsequent productivity gains generated through training depend on the actual post-training wages. I consider a principal-agent relationship with four agents being matched to one principal. In each of two working phases the agents work for 10 minutes on a real effort task. I vary whether the respective agent receives training or can enjoy free time between the two working phases, and whether he receives a wage increase in the second working phase. Both before and after the training phase/free time, I elicit the social norm with respect to the fair wage for the second working phase using a mechanism similar to the one introduced by Krupka and Weber (2013). The results of my research will shed light on a behavioral mechanism potentially affecting the size of gains generated through training and thus have important implications for firms and institutions.
External Link(s)
Registration Citation
Citation
Petters, Lea. 2018. "The Hidden Cost of Training." AEA RCT Registry. December 31. https://doi.org/10.1257/rct.2711-4.0.
Former Citation
Petters, Lea. 2018. "The Hidden Cost of Training." AEA RCT Registry. December 31. http://www.socialscienceregistry.org/trials/2711/history/39753.
Sponsors & Partners

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

Request Information
Experimental Details
Interventions
Intervention(s)
Intervention Start Date
2018-02-05
Intervention End Date
2018-06-18
Primary Outcomes
Primary Outcomes (end points)
Fair wage norm, number of decoded words
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Time needed per decoding task, time worked on the task, revenue generated
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
In my experiment, I consider a principal-agent relationship with four agents being matched to one principal. In each of two working phases the agents work for 10 minutes on a real effort task, namely decoding of sequences of numbers into words with the help of a decoding table (comp. Erkal, Gangadharan, and Nikiforakis 2011). For each working phase the agents receive a fixed wage, whereas the principal is paid a fixed amount for each correctly decoded word by her agents. Between the two working phases, agents either participate in a training phase or enjoy free time and surf the internet. In the training phase, which also lasts for 10 minutes, agents are shown a short animation with the decoding of the numbers into letters and thereafter asked to type in the solution themselves. Since every second decoding task in the second working phase consists of a task already practiced in the training phase, the training enables the trained agents to decode more words compared to their untrained counterparts. Both before and after the training phase/free time, I elicit the social norm with respect to the fair wage for the second working phase using a mechanism similar to the one introduced by Krupka and Weber (2013). In the second working phase, I vary whether or not the respective agent receives a wage increase. While both the training decision as well as the decision of the wage level for the second working phase is determined by the principal, the treatments are completely exogenous. This is because the principal has no information about the performance or any other characterics of her agents and can only choose between two (for her) arbitrary options. This ensures that my observations are independent from each other and provides me with a perfectly balanced data sample. After the experiment, the agents are paid their respective wages for both working phases, while only one working phase is randomly drawn to determine the principals’ payoff.
The experiment will be conducted at the Cologne Laboratory for Economic Research (CLER) and a total of 480 subjects (96 principals + 384 agents) in 16 experimental sessions will be recruited through the online recruitment software ORSEE (Greiner 2015). The experiment ist programmed using Java and oTree software (Chen, Schonger, and Wickens 2016).

Treatments:

No Training & No Wage Increase (n=96)

No Training & Wage Increase (n=96)

Training & No Wage Increase (n=96)

Training & Wage Increase (n=96)

Experimental Design Details
Randomization Method
Treatment randomization by random seating as subjects draw a cabin number when entering the lab.
Randomization Unit
Individual randomization into one of the four treatments within each session.
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
16 sessions with a total of 480 subjects (96 principals + 384 agents)
Sample size: planned number of observations
384 subjects (=observations from agents)
Sample size (or number of clusters) by treatment arms
Approx. 96 subjects per treatment arm
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Since, to the best of my knowledge, there exist no other studies which use a similar experimental setting to test the effects of training, I conducted a pilot session (n=22 ) to gain an idea of the potential effect sizes. Using the means and standard deviations of the fair wage level norm for untrained and trained agents from the pilot session to test the effect on the reference point (No Training: mean 10.70, std. dev. 2.5; Training: mean 13.18, std. dev. 2.14), power calculations predict a power of 95% for a sample of 24 observations per treatment group. To determine the power with respect to the performance effect using the pilot data, however, is more challenging since it includes four different groups and requires a regression approach with an interaction term. Therefore, I ran Monte Carlo simulations with 1000 iterations limiting the number of observations to the financially feasible 36 independent observations per treatment arm. I drew the agents’ simulated delta in performance from a normal distribution again using the values from the pilot session (No Training & No Wage Increase: mean 1.6, std. dev. 8.50; No Training & Wage Increase: mean 4, std. dev. 2.45; Training & No Wage Increase: mean 0.83, std. dev. 5.88; Training & Wage Increase: mean 6.17, std. dev. 5.34). The results of these simulations, indicate that I am slightly underpowered for my hypotheses regarding the performance (p-value below 0.1 in <50% of cases). Nevertheless, knowing that in my actual sessions I will be able to control for the pre-treatment reference point (this was not included in the pilot session), which will increase my power, and given my budget constraints, I run the experimental session with a total number of 180 subjects, which amounts to 36 independent observations per treatment arm.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan

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

Request Information
Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
No
Is 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