Bridging the Digital Divide: training, equipment or both?

The objective of the project is to address and reduce the digital skills gap and the low employability of the Minimum Living Income and the Canary Insertion Benefit (PCI) recipients, through training itineraries adapted to their socio-educational characteristics and digital skills.

2023-01-01

2023-06-30

1. Labor supply: Employment/Nonemployment/Hours of work
2. Independence of the social welfare system: Unemployment/Social Security recipient
3. Job stability: Job retention/ Number of contracts (past 6 months)

1. Employment/nonemployment: Indicator variable equal to 1 if the participants is employed/nonemployed at endline (or in later periods we observe in the administrative data) and 0 otherwise.

2. Unemployment/Social Security recipient: Sum of months or amounts of social security benefits (IMV or PCI) in a given year

3. Job retention: Indicator variable that is equal to 1 if the participant found a job after the study started and is still in that job by the time of the endline.

1. Human capital accumulation: number of on-the-job training/internships in the past 6 months
2. Sorting to more IT-oriented Sectors/Occupations
3. Wellbeing and social inclusion

2. IT-oriented Sectors/Occupations measures will be constructed using text analysis on the ESCO task description
3. Weelbeing -- self-reported on a scale between 0-100
Social inclusion -- self-reported on a scale between 0-100

We randomize two treatment arms:
T1: receiving a tablet
T2: tablet and training
CG: Control group

The tablets for participants assigned to treatment groups T1 and T2 will be given at the beginning of the intervention. These tablets do not have to be returned to the implementer at the end of the project. Access to the internet is paid for 12 months since the delivery of the tablet.

The digital skills training consists of a total of 100 hours of teaching, 80 of them in person and 20 virtual, over a period of 10 weeks, followed by a 3-week unpaid internship at a company or nonprofit organization to apply the skills acquired. At the end of both the instruction period and the internship, there will be an evaluation conducted by the program’s implementers to assess the degree to which the new skills have been learned.

Regarding incentives for participation, all study participants—both in control and treatment groups—received a €30 supermarket voucher for participating in the first phone interview and will receive another €50 supermarket voucher after completing the follow-up and the endline surveys, respectively.

the SGOPIPS unit at the Ministry assigned, within each set of three groups, a random number between 0 and 1 using the random variable RANUNI in SAS. This random number was used to sort the three groups in each set from lowest to highest. Then, the first group was assigned to Control, the second to T1, and the third to T2.

Potential participants are first grouped into sets of 15 people by geographic proximity, as this is the desired class size for the digital skills course. The treatment is randomized at this group level.

No

Due to the different characteristics of each of the islands, we will stratify the random assignment by island. In addition, on each island the groups of 15 will be assembled into sets of 3, so that in each of these sets one group is assigned to T1, another to T2 and the last one to the control group. This will be done by the Ministry using a random number generator as explained below.

The number of groups of 15 people in each island will be attempted to be a multiple of three. In cases where this is not feasible, the "leftover" sets will be grouped into additional sets of three across islands and they will enter the randomization.

This process has already been implemented, as shown in Table 1 above. The number of groups was a multiple of three in two islands (Fuerteventura, with 9, and La Palma, with 6) but not in the others. In Lanzarote, with 7 groups, it was decided that one of the groups would be assigned to T2 and excluded from the RCT, because it was located far from the other groups geographically. In the two largest islands there were 98 groups (Tenerife) and 76 groups (Gran Canaria). The two “leftover” groups from Tenerife were grouped into a set with the one “leftover” group from Gran Canaria, and the treatment assignment rule was applied to that set of groups.

The sample of 3,010 people will be divided into three homogeneous groups: 1000 correspond to Treatment Group 1 (T1, receiving a tablet) 1000 to Treatment Group 2 (T2, tablet, and training) 1000 to the Control Group (CG).

A breakdown of each treatment arm by Island is in Table 1 of the attached PAP. In a high level:
- T1: 65 groups
- T2: 66 groups
- Control: 65 groups

We have made power calculations using the software Optimal Design and taking as a reference a question on the frequency of internet use from a survey conducted by the National Statistics Institute (INE). The variable has a mean of 1.17 and standard deviation of 0.478, from a sample of 382 individuals. For our setting, we make the following additional assumptions: a) A sample size of 3,000, with 1,000 in each treatment arm. b) Significance level = 0.05. c) Statistical power = 0.80. d) The regressors have a minimal (𝑅2=0) or moderate (𝑅2=0.3) explanatory power. The results of this calculation show that the minimum detectable effect is between 0.11 standard deviations (when assuming 𝑅2=0.3) and 0.13 standard deviations (under the 𝑅2=0 assumption).