Detecting Discrimination in the German Child Care Market – A Correspondence Study

We will send out fictitious email inquiries to child care centers in Germany. The emails include implicit signals of parents’ SES that we vary across emails. In addition, a subset of the randomized emails also includes an email signature that informs about the sender's educational background. Hence, we randomize over four treatments (high vs. low SES signal; education marker yes vs. no). We then analyze whether response rates, the response time, and the quality of the answers differ by senders' background.
In order to contact child care centers, we use a commercial dataset that entails basic information (e.g., the email addresses) of more than 90 percent of all child care institutions in Germany. We will impose certain sample restrictions (e.g., we focus on child care centers open for children below the age of three years), which leaves us with an experimental sample of about 22,000 child care centers. We will then randomly assign child care centers to one of our four treatment conditions and send the respective email.

The emails that we plan to send out are worded as follows [note that the names are just examples, see below for details]:
• Email 1 (no migration background and no education marker; translated from German to English):
Subject: Request for a child care slot
Dear Sir/Madam,
We are looking for a child care slot for our [son/daughter] starting in January 2022. [He/she] is now 1 year and 5 months old.
Do you still have a place available? How can we apply for a slot?
Thank you!
Sincerely,
Stefanie Müller

• Email 2 (migration background and no education marker; translated from German to English):
Subject: Request for a child care slot
Dear Sir/Madam,
We are looking for a child care slot for our [son/daughter] starting in January 2022. [He/she] is now 1 year and 5 months old.
Do you still have a place available? How can we apply for a slot?
Thank you!
Sincerely,
Ömer Yildirim
• Email 3 (no migration background and education marker; translated from German to English):
Subject: Request for a child care slot
Dear Sir/Madam,
We are looking for a child care slot for our [son/daughter] starting in January 2022. He/she is now 1 year and 5 months old.
Do you still have a place available? How can we apply for a slot?
Thank you!
Sincerely,
Sebastian Schmidt
----------------------------------------------------------------------
Sebastian Schmidt, Bachelor of Arts (FH)
[email protected]
----------------------------------------------------------------------
• Email 4 (migration background and education marker; translated from German to English):
Subject: Request for a child care slot
Dear Sir/Madam,
We are looking for a child care slot for our [son/daughter] starting in January 2022. [He/she] is now 1 year and 5 months old.
Do you still have a place available? How can we apply for a slot?
Thank you!
Sincerely,
Eyül Ötztürk
----------------------------------------------------------------------
Eyül Ötztürk, Bachelor of Arts (FH)
Eyül.Ötztü[email protected]
----------------------------------------------------------------------

To balance any effects of child gender or sender characteristics (other than migration background), we additionally randomize (separately for migration and non-migration background) the gender of the child (two alternatives), the gender of the sender (two alternatives) as well as the first name (two male, two female) and the surname (two alternatives) of the sender.
To indicate migration background we use two variations of the most common first names and surnames of people with Turkish migration background living in Germany, and people without migration background (e.g., Stefanie Müller, Sebastian Schmidt, Ömer Yildirim, Eyül Ötztürk). The names in the emails (and email addresses) were chosen based on a survey that asked respondents to associate different names with different migration backgrounds.
Taken together, we send four different versions of the email inquiry from 16 email accounts. We also envisage sending out a second wave of emails to the same child care centers to facilitate within-institution comparisons.

2021-03-22

2021-05-31

• Response rate
• Response time
• Response quality

Heterogeneity by provider type, location, size of the child care center, and share of migrants in a given region.

We send out email inquiries to child care centers which are drafted according to real parental requests. They vary by whether they signal high or low SES of the sender, and whether they contain an email signature informing about the sender's educational background.

• Stratified randomization
• Randomization and stratification carried out by the stratarand command in Stata

child care center level

No

Clusters on individual level (see number of observations)

18,000 child care centers (expected; contingent on email bounce rate; see sample size)

In order to contact child care centers, we use a commercial dataset that entails basic information (e.g., the email addresses) of more than 90 percent of all child care institutions in Germany. We will impose certain sample restrictions (e.g., we focus on child care centers open for children below the age of three years), which leaves us with an experimental sample of about 22,000 child care centers. We will then randomly assign child care centers to one of our four treatment conditions and send the respective email. We expect an email bounce rate of about 20% which leaves us with an expected final sample size of approximately 18,000.