Fighting teacher sorting with behavioral interventions

The RCT will be conducted through an online platform that teachers use countrywide to apply for teaching vacancies. Teachers can apply for up to 5 teaching vacancies within their specialty and must rank these vacancies according to their preference. The treatment arms are designed to encourage teachers to apply for jobs at disadvantaged schools. To determine schools as “disadvantaged”, it was created an index that combined information about a) student performance, b) student socioeconomic background, c) proxies indicating whether schools are hard-to-staff and/or have shortage of teachers.

The RCT has three arms:

1. Control
• Teachers see the list of vacancies available with an icon indicating whether the vacancy is from a disadvantaged school. The list of vacancies is organized in random order.
• Teachers also see an explanation of the icon that says: “Schools that need committed teachers like you and where your ability to generate social impact is greater.”

2. Treatment 1
• Teachers see the list of vacancies available with an icon indicating whether the vacancy is from a disadvantaged school. The list of vacancies is organized in random order.
• Teachers also see an explanation of the icon that says: “Schools that need committed teachers like you and where your ability to generate social impact is greater.”
• Before teachers see the list of vacancies, they are prompted with the following text and question: "Some schools need more committed teachers. In these schools, your work could generate greater social change. Would you be interested in working in these schools where you could have more social impact?"

3. Treatment 2
• Teachers see the list of vacancies available with an icon indicating whether the vacancy is from a disadvantaged school. The list of vacancies is organized in a way so that disadvantaged schools are placed first.
• Teachers also see an explanation of the icon that says: “Schools that need committed teachers like you and where your ability to generate social impact is greater.”

2019-11-20

2020-03-01

Our primary outcomes are related to school choice:

1) The number of "high impact"/"disadvantaged" schools in the school choice set.
2) The proportion of "high impact"/"disadvantaged" schools in the school choice set.
3) A dummy indicating if there is at least one "high impact"/"disadvantaged" school in the first N choices.
4) A dummy indicating if the teacher was allocated to a "high impact"/"disadvantaged" school.
5) The average quality of teachers (as measured by the grade in the teachers' evaluation) allocated to "high impact"/"disadvantaged" schools per group.
6) The average quality of teachers (as measured by the grade in the teachers' evaluation) that included a "high impact"/"disadvantaged" school in their choice set per group.

We may also include similar outcomes related to choice. We are also planning to implement an ex-post survey to measure preferences and beliefs, but this is still not confirmed. In the same survey, we plan to measure some traits that could help us understand potential mechanisms (for instance, altruism).

We plan to conduct heterogeneity analysis by gender, age, etc. We would expect our treatment to be more effective among more altruistic people, among people that are more willing to travel farther away.

The RCT will be conducted through an online platform that teachers use countrywide to apply for teaching vacancies. Teachers can apply for up to 5 teaching vacancies within their specialty and must rank these vacancies according to their preference. The treatment arms are designed to encourage teachers to apply for jobs at disadvantaged schools. To determine schools as “disadvantaged”, it was created an index that combined information about a) student performance, b) student socioeconomic background, c) proxies indicating whether schools are hard-to-staff and/or have shortage of teachers.

The RCT has three arms:

1. Control
• Teachers see the list of vacancies available with an icon indicating whether the vacancy is from a disadvantaged school. The list of vacancies is organized in random order.
• Teachers also see an explanation of the icon that says: “Schools that need committed teachers like you and where your ability to generate social impact is greater.”

2. Treatment 1
• Teachers see the list of vacancies available with an icon indicating whether the vacancy is from a disadvantaged school. The list of vacancies is organized in random order.
• Teachers also see an explanation of the icon that says: “Schools that need committed teachers like you and where your ability to generate social impact is greater.”
• Before teachers see the list of vacancies, they are prompted with the following text and question: "Some schools need more committed teachers. In these schools, your work could generate greater social change. Would you be interested in working in these schools where you could have more social impact?"

3. Treatment 2
• Teachers see the list of vacancies available with an icon indicating whether the vacancy is from a disadvantaged school. The list of vacancies is organized in a way so that disadvantaged schools are placed first.
• Teachers also see an explanation of the icon that says: “Schools that need committed teachers like you and where your ability to generate social impact is greater.”

The Ministry shared an anonymized dataset with the research team including all the teachers that passed the test to become teachers and apply for a job. We stratified the samples by "Cantons" (second-level administrative units, there are 221 in the country). Within each Canton, we assigned a number 0,1 or 2 (control, treatment 1 or treatment 2) consecutively. We then tested balance among groups using observable characteristics (gender, teaching test scores, age, ethnicity).

The randomization unit is the individual (teachers). There are no clusters.

No

25,000 teachers (no clusters)

25,000 teachers in total

Our same size is of 25,000 teachers for Control and Two treatment arms.

We used "% of teachers that selected disadvantaged schools (at least 1)" as the main outcome for power calculation. Assuming 20% of teachers in the control would select these schools (based on conversations with the government), we need 12,606 teachers per arm to detect an effect of 2.5pp (1 arm). We have two arms and approximately 25,000 observations.