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.
Do men earn more or report earning more?
Initial registration date
February 25, 2020
May 17, 2020 10:20 AM EDT
This section is unavailable to the public. Use the button below
to request access to this information.
Other Primary Investigator(s)
Additional Trial Information
The gender gap in earnings and in estimated returns to capital has been documented in many contexts worldwide. However, the vast majority of studies rely on self-reported measures of profits and revenues, which are liable to considerable measurement error, especially in developing countries where few entrepreneurs keep financial records. It can be especially problematic to estimate the gender gap in profits, if men and women report earnings differently. How much of the estimated gender profit gap is due to differential reporting by gender? In our survey of vegetable vendors in India, we find suggestive evidence that women tend to under-report revenues and men tend to over-report revenues. We propose a short experiment to shed light on gender differences between counted (enumerator-based) and reported (owner-based) revenues. We explore various mechanisms, including overconfidence and social norms around modesty. In the presence of differential mis-reporting by gender, the magnitude of the gender profit gap could be considerably biased.
Intervention Start Date
Intervention End Date
Primary Outcomes (end points)
Self-reported cash on hand on the day of the survey, self-reported revenue on the day of the survey, self-reported revenue the day prior to the survey, self-reported average revenue (daily, weekly, monthly), self-reported profits the day prior to the survey, self-reported average profits (daily, weekly, monthly), incentivized self-reported cash-on-hand today, true count of cash-on-hand today, estimate of revenue today given reported expenses and cash-on-hand count. We are also interested in the difference between the true count and self-report estimate of cash on hand, and the difference between the true count and the incentivized self-report of cash on hand.
Primary Outcomes (explanation)
In the evening, vendors will be asked how much cash they think they have on hand before we count the cash. This measure is the self-reported measure (SR). We will then announce to the vendor that if they are correct within Rs 100, they will get a bonus of Rs 50. We ask them if they want to revise their estimate: this measure is the incentivized self-report (ISR). We count the cash: get the “True count” (TC)
Secondary Outcomes (end points)
Overconfidence and modesty, as measured by the games outlined in the Experimental Design section. For overconfidence, the dependent variables are self-reported performance in the game in the public version, and the difference in self-reported performance and actual performance in the public version of game. For modesty, the dependent variables are: self-reported performance in the private version, the difference in self-reported and actual performance in the private version, willingness to publicize good performance, and measures of adherence to norms around modesty. Additionally, we have other measures of overconfidence and modesty, outlined in the section below in more detail. Finally, we elicit a private report of revenues.
Secondary Outcomes (explanation)
1. Similar to Sarsons and Xu (2016) (“Confidence Men? Gender and Confidence: Evidence among Top Economists”)
Ask questions related to the work of a market vendor, e.g. “To what extent do you agree with the following statement:
tomatoes are the most commonly sold vegetable in this market today. 1=Strongly Agree, ..., 5=Strongly Disagree.”
Ask degree of confidence on a scale of 1 (unconfident) to 10 (very confident).
Confidence can be measured in three ways:
# of people who decline to answer because they don’t know
# of extreme answers (Strongly Agree/Strongly Disagree)
Degree of self-reported confidence (1-10).
2. Similar to Niederle and Vesterlund (QJE, 2007): Suppose I choose 10 vendors at random in this market, including yourself. How do you think your business ranks in terms of profitability? 1= most profitable, 10 = least profitable. Then we ask vendors the same question but for their own gender (for women: out of 10 female vendors, for men: out of 10 male vendors). 3. Similar to Bursztyn et al. 2018 on misperceived social norms in Saudi Arabia around FLFP:
We also want to ask questions to measure social norms around modesty.
(“First order” beliefs) “To what extent do you agree with the following statement: In my opinion, women should not brag
about their accomplishments.” 1=Strongly Agree, ..., 5=Strongly Disagree. (We will also ask this about men)
(“Second order” beliefs) “We are asking this question to 10 other vegetable vendors in Jaipur, both male and female. How
many do you think agree or strongly agree with the statement?”
Private report of revenues: "How much did you make two days ago? Please write down your answer and put it in this envelope." (we will ask this for both profits and revenues). We will explain to the respondent that the surveyor will never see their answer. We note that women's literacy may be a significant constraint in this exercise, so we may have to disregard the data if it appears to be the case.
The experimental design consists in counting the cash the vendors have on hand ("true count") and offering an incentive to respondents to compare the "self-report" to the "incentivized self-report". We will compare these two measures to the "true count".
Additionally, respondents will play various games with us to help us understand mechanisms.
Experimental Design Details
This isn't a study that tests an intervention, but a study to expose a potential reporting gap when respondents report profits and revenues, and understand mechanisms. Additionally, we incorporate a few games to test for mechanisms, Each respondent plays both a private and a public version of the games, and the order of the games and public/private condition are randomized. Randomization done in office by a computer, incorporated to the survey on SurveyCTO or ODK.
The main variable of interest is gender differences in self-reporting. For that reason, we will sample 50% men and 50% women. We will be able to compare the unincentivized self-report to the incentivized self-report for every respondent.
Was the treatment clustered?
Sample size: planned number of clusters
Our sample size is 500 vendors in Jaipur.
Sample size: planned number of observations
For every seller, we will have one self-reported, one incentivized self-reported and one true count.
Sample size (or number of clusters) by treatment arms
Our sample size is 250 male and 250 female vendors in Jaipur.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Approval Date
IRB Approval Number