A factorial survey experiment on the duration of parental leave

Our goal was to identify, through a Factor Survey Experiment, the causal effect of some aspects of the leave system, workplace and family environment on the decisions of Spanish sample of employees aged between 25 and 45 about the total number of weeks they would be on parental leave.

We use Sen’s capability approach to identify factors able to increase the effective freedom of working parents when deciding the length of their parental leave. The experimental dimensions (or factors) considered were: Statutory length of paid leave (17, 22 or 26 weeks), replacement rate (60% or 100%of current salary), job security (indefinite contract or temporary contract), earnings compared wiht the partner (income less or higher than your partner's), harm to colleagues (leave does not harm or you will har your collegues by taking leave), harm to promotion (by taken a parental leave you do not or you will harm promotion or your career advancement), peers role model ( inside firm it is very frequent thah men/women who have a baby are on leave for long periods, or it is very rare), and easy application for parental leave (from an organizational and legal point of view, the procedure to be able to use the leaves is very simple or is not simple because you have to negociate with your bosses on dates and duration).

We conducted a factorial survey experiment (FSE) with a Spanish sample of 946 employees aged between 25 and 45. Respondents were asked to imagine that they were first time parents and several descriptions of hypothetical situations or vignettes (where aspects of the leave system, workplace and family environment were randomly varied) were presented to them. Then they are asked to inform about the total number of weeks they would be on parental leave under each hypothetical situation

2021-04-12

2021-05-20

A longer statutory duration and a higher replacement rate (of the paid part of the leave), having job security, and a workplace with friendly practices and cultures, were found to have a significant positive impact on the duration of parental leave

Factor survey experiment design. In a first stage, the total sample of respondents was randomly divided into three parts and assigned to one level of the first dimension "Statutory length of paid leave" (17, 22 or 26 weeks).

For each of these three subsamples, in a second stage, for each third of the sample, the remaining seven dimensions were assigned through the vignettes. Since they are seven dimensions of two levels each, the vignette population is composed by 128 (2x2x2x2x2x2x2) possible hypothetical situations. We used the SAS macro “%MktBlock” to allocate the 128 vignettes to 16 different questionnaire versions (decks), each consisting of 8 vignettes, providing a 100% D-efficient design. Once these 16 decks had been designed, each respondent was randomly presented with only one deck, so each respondent had to make the decision on how many weeks of parental leave to take in only 8 hypothetical situations.

Data for this study factor survey experiment were obtained from an email survey of 946 respondents in May 2021 through the Cint Panel (www.cint.com) and by using the web platform tickStat (www.tickstat.com) that allows the automatic management of complex quota sampling configurations. Our target population was individuals residing in Spain who were wage earners and aged between 25 and 45 years. We have used a quota sampling among the panelists belonging to the target population according to the Economically Active Population Survey (EPA), which is a representative labour force survey produced in Spain by the National Institute of Statistics (INE 2021). A total of 42 sampling quotas for the target panelists (wage earners aged between 25 and 45 years and residing in Spain) who participated in the FSE have been established based on the combination of seven regions (NUTS1) of residence, two genders and three educational levels.

Randomization was done in office by a computer when each panelist was assigned to each branch of "Statutory length of paid leave" (17, 22 or 26 weeks) and to one of the 16 predefinned decks (composed of 8 vignettes each deck).

Randomization Unit was individual or panelist. Each panelist wass randomly assigned to each deck, so each respondent had to make the decision on how many weeks of parental leave to take in only 8 hypothetical situations.

We use a sample quota configuration of wage earner aged between 25 and 45 years residing in Spain) according to three characteristic of the panelist: Spanish Regions (seven NUTS1), gender, and three educational level (42 sampling quotas in total). ). When the desired sample of participants belonging to each of these quotas was reached, no more responses from panelists with these characteristics (belonging to that sample quota) were admitted. On the contrary as long as the quota of panelists for each quota was not reached, those panelists were accepted and randomly assigned to each deck.

Yes

Planned 962 panelists. Final number of panelist 946

Planned 962 panelist * 8 vignettes = 7696 vignettes. Final number of Observationes 946*8= 7568 observations

Sample size or number of clusters by treatment arms
Each respondent must make decision on how many weeks of parental leave to take in only 8 hypothetical situations (vignettes) that include a total of 8 experimental factors or 8 experimental dimensions.

In total 946*8=7568 observations for each experimental dimension is obtained.

- Statutory length of paid leave dimension had three levels (“arms”) with the following sample for each level 17 weeks: 2472, 22 weeks: 2568; and 26 weeks: 2528). 7568 in total. Each panelist responds only to one of these branches.

- Each Deck (composed by 8 hypothetical situations or vignettes) was designed to incorporate 4 control arms and 4 treatment arms for each of the other 7 experimental dimensions. These other 7 dimension have only two levels (arms) with 3784 sample each level (7568 in total). This 7 dimension are: High replacement rate, Job security, earning Lower income than my partner, leave does not harm colleagues, leave does not harm promotion, peers role model inside firm in using parental leaves, and easy application for parental leave. Each panelist responds to the same number of arms of these 7 experimental dimensions.