x

NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Colluding to Keep Wages High: Evidence from Casual Daily Labor Markets
Last registered on December 27, 2016

Pre-Trial

Trial Information
General Information
Title
Colluding to Keep Wages High: Evidence from Casual Daily Labor Markets
RCT ID
AEARCTR-0001290
Initial registration date
December 27, 2016
Last updated
December 27, 2016 5:20 PM EST
Location(s)
Region
Primary Investigator
Affiliation
Department of Economics, Harvard University
Other Primary Investigator(s)
PI Affiliation
Columbia University, Department of Economics
PI Affiliation
Columbia Business School
PI Affiliation
University of California, Berkeley, Department of Economics
Additional Trial Information
Status
On going
Start date
2015-08-01
End date
2018-06-30
Secondary IDs
None
Abstract
This project seeks to establish and provide a micro-foundation for the presence of wage floors in village agricultural labor markets. Specifically, we posit that during times of high seasonal unemployment, some workers find it privately optimal to take up jobs at wages lower than the prevailing wage, but do not do so because this would result in sanctions from co-villagers. The study uses a field experiment in rural Indian villages to test for evidence that village laborers act as informal unions and are willing to pay a cost to enforce the wage floor. The design is implemented in partnership with local landowners to offer employment to randomly-selected workers during the agricultural lean season. It varies, at the village-level, both the wages for the jobs (either at the prevailing wage or at 10% below the prevailing wage) and the observability of the offers (in public in the village, or in private, within the worker’s home). If communities behave as an informal union, then (some) workers will be willing to accept work at a wage cut in private, but there will be a large decline in take-up when the job is offered in public. In contrast, we predict that the observability of workers’ take-up decisions will not affect labor supply when jobs are offered at the prevailing wage. In addition, we seek to provide positive evidence for the idea that workers impose sanctions on co-villagers if they accept a job below the prevailing wage. In a costly punishment game, laborers are given anonymized information about the real take-up decisions of other laborers in the same village. They are asked if they wish to give away money from their own endowments to reduce the amount that other individual receives.
External Link(s)
Registration Citation
Citation
Breza, Emily et al. 2016. "Colluding to Keep Wages High: Evidence from Casual Daily Labor Markets." AEA RCT Registry. December 27. https://doi.org/10.1257/rct.1290-1.0.
Former Citation
Breza, Emily et al. 2016. "Colluding to Keep Wages High: Evidence from Casual Daily Labor Markets." AEA RCT Registry. December 27. http://www.socialscienceregistry.org/trials/1290/history/12877.
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)
CONTEXT

The study takes place rural Odisha, one of India’s most underdeveloped states. Markets for casual daily wage labor are extremely active, and provide the primary source of wage labor earnings for residents in the area. A large portion of workers in construction, unskilled manufacturing, and other factories are hired through these labor markets (Breza, Kaur, and Shamdasani 2016).

In the data collected by Breza et al. (2016), there appears to be a wage floor, even during the agricultural lean season. This floor coincides with the prevailing wage in agriculture. Concretely, denote as W the prevailing wage in agriculture in a village. When workers from that village work in the non-agricultural sector—largely in jobs that take place outside the village—the wage they earn is at or above W in 98% of cases. This is despite the fact that unemployment levels appear high. Employment rates (in terms of total worker-days across all sectors) are below 50%. A striking 80% of workers report being involuntarily unemployed at least one day in the past two weeks. Understanding the source of the wage floor inside the village therefore has potential bearing on understanding determinants of the wage in the labor market as whole.

RESEARCH HYPOTHESES

This project aims to test whether worker collusion contributes to maintaining a wage floor in rural labor markets in the agricultural lean season. In essence, do workers act an as informal union—rejecting job offers that are below the wage floor? This behavior could be maintained through two potential forces. The first is that a worker faces explicit sanctions from co-villagers if he accepts a wage cut. The second is that there is a “fairness norm” against wage cuts that workers internalize that lower their willingness to accept wage cuts (Kahneman et al. 1986, Kaur 2015). Both forces could work in tandem—arising from the same microfoundation—to produce the same results (e.g. MacLeod 2007). Our primary goal is to provide explicit evidence for worker collusion. Our secondary goal is to test for suggestive evidence for these two forces.

Denote the prevailing village wage as W. If worker collusion contributes to downward wage rigidity at W, then we hypothesize that during the agricultural lean season:

H1) The true private opportunity cost of working for a subset of individuals is less than W. I.e., workers will be privately willing to accept work at wages below W.

H2) When other workers can observe an individual’s job take-up decision, workers will be less likely to accept work below W.

H3) Workers will be willing to pay to sanction others who have accepted work below W.

To test these hypotheses, we work with employers in villages in several districts in Odisha to offer lean season jobs to laborers on their land for common agricultural tasks like weeding, field maintenance, and compost crushing (which are paid at fixed daily-wages for men. Through an agreement with the employer, we pay part (approximately 75%) of the labor cost for the task in exchange for being able to control to whom we offer the job and under what wage and circumstances.

For each village in which we work, we hire laborers from the closest associated laborer colony that the village typically hires from. We first create a list of all households that provide agricultural labor in the colony. We then randomly offer jobs to two or three individuals in the village, depending on the size of the labor colony. We offer jobs to male daily-wage agricultural laborers (with the employer present in the vicinity). Randomization is at the village level, so all laborers within a village receive the same treatment.

Our treatments vary the level of publicity of the job offers and the wage rate. We cross-cut three treatment arms with two different wage rates:

WAGE RATES:

W1) Prevailing wage in the village for that task
W2) Rs. 20 (Approximately 10%) below the prevailing wage for that task

TREATMENT ARMS:

T1) Fully public, in which the job offers are made in the street outside the home of the participant, with the potential for people around to hear the terms of the job offer
T2) Partially private with the employer present, in which the job offers are made within earshot of the employer, but inside the home of the participant out of earshot of others in the village
T3) Fully private, in which job offers are made inside the home of the participant out of earshot of others in the village. The employer waits outside of the home and does not overhear the wage offer.

The cross-cutting of various treatment arms and wage rates leads to six distinct treatment arms. We list the treatment arms in detail and our hypotheses in comparing these treatment arms in the 'experimental design' section of this filing.

We record the participant's take-up of the job. We also conduct an endline survey that includes a ten-day recall of employment with everyone offered the job and five control individuals who were never offered the job. We conduct a survey with the employer to elicit general details about the employment process in the village and the rating of the quality of the participant's work.

Lastly, in a separate set of villages, we play incentivized lab games to illustrate that people are willing to engage in costly punishment of those to violate a village norm.
Intervention Start Date
2015-08-01
Intervention End Date
2017-12-31
Primary Outcomes
Primary Outcomes (end points)
Job take-up is the most important outcome

We will also collect survey measures from the employer:
Number of hours worked by study workers
Worker effort as rated by the employer
Any other transfers to the workers either cash or in-kind


Responses of study participants to survey questions:
Previous work experience

Survey responses of non-study participants:
Knowledge of the wage
Primary Outcomes (explanation)
The primary outcome variables are the job take-up decisions.
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
EXPERIMENT SET-UP

The project will be conducted in several districts of Odisha during the agricultural lean season. To implement our experiment, we partner with employers in the village, who typically hire daily-wage laborers for tasks like weeding or house maintenance. We subsidize the cost of the labor for the employers, in exchange for the ability to randomize who is offered work and the level of the wage that is offered. This enables us to hire workers for real jobs with a real employer, while varying wage levels and conditions of the offer. (Note that because we care about the labor supply side only, internal validity is not affected by the fact that the employer is being compensated for his cooperation). In all cases, the employer accompanies our staff when job offers are made, to prove the veracity of the offer. (However, in some treatments the employer is not able to overhear the exact wage offer given to the worker.)

PUBLIC VS. PRIVATE TREATMENTS

As an initial test of hypotheses H1, we randomize wage offers to be at W versus W-10% (i.e. a 10% wage cut below the prevailing wage). To test hypothesis H2, we vary the degree to which others in the village observe the worker’s job acceptance decision. This gives rise to a 3x2 design:

(i) Full public wage cut: wage offer of W-10% on the street (outside the worker’s home) where other workers and employer can observe the answer.

(ii) Partial private wage cut: Wage offer of W-10% in private inside the worker’s home, with the employer present.

(iii) Full private wage cut: Wage offer of W-10% in private inside the worker’s home, with no one else from village present (the employer waits outside, and remains blind to the wage level).

(iv) Full public prevailing wage: public straight offer at W.

(v) Partial private prevailing wage: Wage offer of W in private, with the employer present.

(vi) Full private prevailing wage: Wage offer of W in private, with no one else from village present (the employer waits outside, and remains blind to the offer).

Randomization is at the village level—only one treatment is implemented in a village. We offer the job to five people in each village, and record take-up rates for the job. The laborers who take up work for one day for the employer at the offered wage. The employer provides supervision and in-kind benefits as usual.

H1 and H2 predict that take-up of work under a wage cut will be substantially lower under Treatment (i) than under Treatment (iii). (Treatment (ii) allows us to partial out the extent to which employer observability drives effects).

In contrast, in treatments (iv), (v), and (vi), no community norms against wage cuts are being violated. Consequently, we expect observability to have no impact—providing a useful flasification test. We predict no substantial difference in take-up across these latter 3 treatments. We do expect (iv) and (v) to be weakly higher than (vi) if there are positive peer effects of taking up a job in public, or if there are negative signals associated with turning down a job at W in public or in the presence of an employer.

EVIDENCE ON SANCTIONS AND NORMS

We use two supplementary pieces of evidence to provide positive support that workers face sanctions for accepting work below the prevailing wage.

First, we plan to conduct 10 endline surveys in each village with those approached with a job offer, and a random sampling of control workers who were never approached. We test the spread of information throughout the village and elicit fairness norms (as in Kaur 2015). In addition, we elicit self-reports on what sanctions a worker would face if he took a job.

Second, we will use a set of incentivized lab-based activities with employers and workers, in similar villages to those in our field experiment. We will present lab respondents with the true decisions of anonymous subjects in our field experiments.

In the costly punishment game, we will give the lab respondent an endowment of Rs. 100. He will then be presented with the decision of an anonymized field subject, who will also be given an endowment of Rs. 100. The lab respondent will then be given the opportunity to levy a costly punishment on the field subject. Paying Rs. 1 will result in a decrease in the field subject’s endowment of Rs. 5. The maximal punishment, which costs Rs. 20, leaves the field subject with a final payoff of Rs. 0. Here, we will again vary the treatment and decision of the field subject. Again, we hypothesize that individuals who accept work at Rs. 180 will experience the largest punishments by lab respondents.

Upon completion of the laboratory exercise we will deliver the earnings from these games to the field subjects referenced in the lab activities. The field subjects are not expecting any additional transfer, so receiving any cash will be a positive surprise to them.
Experimental Design Details
Randomization Method
Randomization is done in the office by one of the PIs or a Research Assistant or graduate student on the computer.
Randomization Unit
We randomize treatments at the level of the village. Within a village, we randomize the households we approach and the order in which we approach them. We also randomly select three control households in the village household to approach for the endline control survey.
Was the treatment clustered?
Yes
Experiment Characteristics
Sample size: planned number of clusters
We cluster individual responses to the job offer at the village level. The complete experiment is targeting 300 villages. We completed some initial data collection in the lean seasons of 2015 and 2016. The bulk of our data will be collected in the lean seasons of 2017, during which we will collect as much data as possible. However, there is a chance that we will not reach our target of 300, depending on the productivity of our field teams.
Sample size: planned number of observations
Approximately 800 individuals. Most villages will have 3 treated individuals. However, in smaller villages we will limit this to 2 households.
Sample size (or number of clusters) by treatment arms
Fully Public (10% below the prevailing wage rate): 60
Fully Public (at prevailing wage rate): 40
Private with employer present (10% below the prevailing wage rate): 60
Private with employer present (at the prevailing wage rate): 40
Fully Private (10% below the prevailing wage rate): 60
Fully Private (at the prevailing wage rate): 40
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
Columbia University Institutional Review Board
IRB Approval Date
2014-10-09
IRB Approval Number
AAAO2500
IRB Name
Institute for Financial Management and Research (IFMR) Institutional Review Board
IRB Approval Date
2014-10-22
IRB Approval Number
IRB00007107
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