Social Comparisons in the work place: A field experiment in Kolkata, India.

We conduct a field experiment in Kolkata,
India, using a sample of student assistants hired to help with a data entry task.
In the experiment, recruited assistants worked in teams of two for two successive
working sessions in return for a fixed wage per hour. Workers were randomly
assigned to one of the following treatment groups (1) A Control Treatment:
not subjected to wage cuts; (2) A Unilateral Wage Cut Treatment: experiencing
unilateral wage cut, gender identity of co-worker not revealed (3) and the
Identity Effect Treatment: experiencing unilateral wage cut, gender identity of
co-worker revealed.

2017-03-01

2017-05-01

We use regression analysis to estimate the effect of wage cuts on
workers’ effort supply. Workers could exert effort in two dimensions: quantity
(number of entries) and quality (correctness of information entered) of work
produced.

On the first working session, workers were assigned desks and computers in
office-like rooms on campus in addition to unique identification numbers. Participants
were instructed to sign an attendance sheet at the beginning of each
working session to indicate their presence.The attendance sheets are an important
feature of our experiment, they are used to make certain information about
the co-worker salient to the participant. The attendance sheet was attached to
the first page of each booklet and had two rows of information, each resembling
one of the two workers assigned to work on the booklet. Participants were then
asked to sign next to the row resembling their identification number.The information
presented on the attendance sheet were: worker’s identification number, hourly wage, and the gender identity which was revealed only for participants
assigned to treatment three (The Identity Effect). The gender identity was made
salient by adding the first name of each team member next to their identification
number, we abstain from including the full name given that family names
in India can be used to deduce information on the caste identity.
Workers were left to work unmonitored, the coordinator only interfered to
announce the end of the working sessions and to pay the hourly wages.To reduce
any potential bias, both the workers and the session coordinators were blind to
the research question and to the fact that this was an experiment.
On the second working session (post-intervention phase), workers returned
to their desktops and were again asked to sign the attendance sheet before they
continued working on their assigned booklets.This time, however, workers in
treatments two (Unilateral Wage Cut) and three (Identity Effect) experienced
a unilateral wage cut where their new hourly wage was reduced to 210 rupees
per hour as indicated next to their identification numbers. Following the work
by Cohn et al. (2014) the session coordinators were instructed not to give any
further explanations or rationalizations for the cut, instead, workers had the
full freedom to leave the session following the announcement of the new pay. It
was clearly announced that workers who chose to leave the session would still
be fully compensated for their efforts in the first working session, that is, they
would still receive the 350-rupee hourly wage for the work they had done earlier.
None of the workers chose to leave the session.
At the end of the second working session the coordinator returned to the
room to announce the end of the task and to pay the participants their wages.The
coordinator distributed short feedback forms that were filled in by all workers
prior to payment. The feedback form incorporated questions on job satisfaction,
difficulty of task, and clarity of instructions, in addition to questions on beliefs
regarding own performance and the performance of others working on the same
booklet.

Randomisation was done in the office using a computer.

Unit of randomization is at the individual level.

No

We have only 3 universities. Before the trial the randomisation was undertaken within each university. However, we do not conduct a cluster level analysis.

We managed to recruit 312 student assistants to complete the work.

We use data from 312 student assistants hired from three public universities in
Kolkata: Jadavpur University, University of Kolkata, and Presidency University.
To conduct our analysis, we merge three forms of data sets: data extracted
from application forms submitted before the experiment, experimental data,
and finally data extracted from feedback surveys conducted at the end of the
experiment.
The application forms provide us with data on the basic demographic characteristics
of the sample. We use the main characteristics such as gender, religion, caste, subject of study at the university to randomise the participants into the different treatment groups.