Implicit Gender Bias: Evidence from 41 Countries

Survey participants are shown information about a pair of potential job candidates applying for a position in the country's health department and asked to indicate which annual salary they would give to both candidates.

Information about candidates includes the candidate’s age, education level, family status, work experience, and the the gender of the candidate. Educational background varies across countries and is adjusted such that it correspond to a 4-year college degree for each surveyed country.

Within each country, we randomize the genders of the job candidates such that respondents are asked to make a decision for either a male-male, female-female, male-female or female-male pair. We target 250 observations per treatment, per country, for a total of 1,000
targeted responses per country. Responses are constrained to lie between the (rounded) 15th and 97th percentiles of the country income distribution in order to minimize outliers due to typographical errors.

2022-01-01

2022-03-31

Respondents salary decisions.

All primary outcomes can be observed directly from the data.

No secondary outcomes considered at this time.

No secondary outcomes considered at this time.

We use the following text to elicit salary decisions of respondents.

"""
Suppose the [insert country-specific national health department name] in your country hires two new workers. Their responsibilities include gathering information from hospitals on patient health outcomes and writing reports.

Imagine you are responsible for deciding these two workers’ annual salaries. Annual salaries at the [ministry] can range between [15th percentile of country income distribution] [country currency] and [97th percentile] [country currency].

Here are some details about these two workers. Both have [a 4-year college degree or country equivalent] and have 3 years of relevant work experience. Both are married with 2 children.

The only difference is that worker A is a 32-year-old [man/woman] whereas worker B is a 34-year-old [man/woman].

What annual salary would you give worker A? Enter value in whole [currency].

What annual salary would you give worker B? Enter value in whole [currency]
"""

The name of the national health department name, the currency and salary ranges vary between countries. Within each country, we randomize the genders of the job candidates such that respondents are asked to make a decision for either a male-male, female-female, male-female or female-male pair.

Responses are constrained to lie between the (rounded) 15th and 97th percentiles of the country income distribution in order to minimize outliers due to typographical errors. We obtained these percentiles from the World Inequality Database.

The above questions was included as a module in a larger Gallup survey to be conducted in 41 countries. We could include 31 already pre-selected countries to be surveyed online. In addition, we could include 10 countries that were to be surveyed through face-to-face interviews.

Pseudorandom number generator managed by Gallup.

We sample units at the individual level.

No

We cluster at the individual level. The number of individuals that we target corresponds to 41 000 respondents across 41 countries.

In total we target 41 000 respondents across 41 countries.

For a total of 1,000 targeted responses per country, we collect:

-250 observations for a male-male job candidate pair.
-250 observations for a female-female job candidate pair.
-250 observations for a male-female job candidate pair.
-250 observations for a female-male job candidate pair.

Pre-analysis plan for: Implicit Gender Bias: Evidence from 41 Countries