Increasing the proportion of female-led firms in government support programs: A randomized e-mail information experiment

The intervention is an information treatment whereby three different e-mails will be sent to a sample of 4847 Tunisian firms to motivate them to participate in an e-commerce training that intends to promote exports of SMEs. The sample is randomly split into three groups whereby two treatment groups will receive e-mails that, in addition to the general information about the program, contains information that shall encourage female business owners to participate. The first e-mail contains information about free-of-charge childcare services during the classroom trainings that will be offered to participants. The second email includes information and link to a video of a successful female Tunisian entrepreneur that talks about her experience as an exporter. The third e-mail contains just the information required for registration but no additional encouragement messages and will be sent to the control group. The sample consists of both female and male-led firms, as both the potential effect of female encouragement and a potential backlash (by male business owners) who may feel discouraged by the additional messages shall be investigated.

2021-11-01

2021-11-30

Primary Hypothesis 1: Receiving information about free-of-charge childcare services will lead to increase in program registration.
Primary Hypothesis 2: Receiving information about a female entrepreneur that suceeded in exporting through e-commerce will lead to an increase in program registration.

Our hypothesis are based on existing experimental evidence that finds a positive impact of female applications and no differences in male application when using female affirmative action messages in job announcement (Ibanez & Riener 2018)
Our primary hypotheses will be tested through binary outcome variable indicating whether the firm that received the email registered for the program (1) or not (0).

A firm that successfully received an email (no bounce back notification received) but did not sign up for the program will be coded as zero. A firm that successfully received an email (no bounce back notification received) and signed up for the program will be coded as one. String matching based on the email address and firm name will be used to identify registered firms in the original dataset.

We will estimate the gender-specific treatment effect and undertake subgroup-analyses by firm size, governorate (region), town and sectors. The gender-specific effect of the e-mail treatment on registration will be estimated through a gender variable, binary-coded based on first name of the director that was indicated in the data set, which will be interacted with the two treatment dummies (E-Mail 1, E-Mail 2).

Secondary Hypothesis 1: Receiving information about free-of-charge childcare services will lead to a larger increase in program registration among women-led firms than among men-led firms.
Secondary Hypothesis 2: Receiving information about a female entrepreneur that succeeded in exporting through e-commerce will lead to a larger increase in program registration among women-led firms than among men-led firms.

Secondary Hypothesis 1 and 2 are based on the same evidence cited above (Ibanez & Riener 2018)

Secondary Hypothesis 3: Receiving information about free-of-charge childcare services will increase the registration more for small firms than for larger firms. We expect this size-specific effect because larger firms usually have more possibilities to dispose of an employee for the training, whereas this may be more difficult for smaller firms where employees or the business owners have more conflicting duties. Hence, the childcare information will be more relevant for the latter group.

Secondary Hypothesis 4: Receiving information about a female entrepreneur that succeeded in exporting through e-commerce will increase the registration more among small firms than among larger firms. We expect that small firms perceive the act of exporting as more complex and hence the encouragement effect of seeing a successful example of an established (female) entrepreneur should have a larger effect on this group.

Secondary Hypothesis 5: Both treatment effects will vary across sectors. We expect the encouragement messages to have a larger impact in sectors where female underrepresentation is stronger.

The sample consists of 4847 firms. The firms were randomly assigned to treatment 1 (childcare message), treatment 2 (influencer video) or control group. Prior to the randomization, the sample was stratified based on the gender of the firms' director and the sector. Sector information was only obtained for 3723 out of the 4847 firms. For the remaining 1124 firms, sector was categorized as “undefined” and two additional strata were formed for female and male firms with undefined sector information. In total 17 strata were formed. In cases where the number of firms per strata was not divisible by 3, the remaining firms (misfits) were allocated randomly to treatment or control groups within each strata.

Computer randomisation, stratified by gender and sector (see above), using Stata.

Firm

No

4847 firms

4847 firms

Treatment Arm 1 (child-care message): 1615 firms
Treatment Arm 2 (female export entrepreneur encouragement message): 1617 firms
Control Group (neutral message): 1615 firms

see pre-Analysis plan (available from authors upon request).