In order to answer the research questions, we present the results in the same order that we posed them. Table 1 shows the demographics (gender, age and level of education) of the individuals per MOOC. Although the subpopulation of interest is n = 951, the number of respondents differ in the variable occupational setting due to no-response of some individuals.
Overall, the percentage of women who participated in the MOOCs included in the study is around 56 %. This is considerably higher than the 2:1 male to female rate found in some U.S studies (Glass et al. 2016). This finding could be explained by the fact that the MOOCs in our sample did not include the usual IT or engineering courses. The male–female ratio varies according to the topic of the MOOC. More women tend to participate in Hands-on ICT and test anxiety courses (over 70 %), whereas fewer participate in business intelligence and entrepreneurship courses (38 and 42 % respectively).
The mean age of men and women in our population is similar (M = 42.9 SD = 9 and M = 41.6 SD = 9.5 respectively), the median age is 43. These figures are higher than reported in other studies. For example, Ho et al. (2015) reported a median age of less than 30 years in Harvardx and MITx courses and Zhenghao et al. (2015) report a median age of 41 for Coursera MOOCs. Although longitudinal research suggested that the average age of MOOC learners is increasing, (Glass et al. 2016), the high mean age in our data may be caused by the fact that we were selecting only learners who were currently working or looking for a job, and excluding younger people such as students.
Unsurprisingly, the vast majority of our sample (81 %) have completed tertiary education and 65 % have completed a second stage of tertiary education. (70 % of women and 58 % of men). This is in line with the high educational levels reported in previous research on MOOCs (Ho et al. 2015).
Around one-third of our sample of interest (35 %) was unemployed but looking for a job and the rest (65 %) was employed. Among those individuals who were currently working only one-third (30 %) received employer support for professional development activities. The proportion of unemployed/employed for wages in our sample differed from the data available for the US context (Christensen et al. 2013) where focusing only on the subsample of learners who were either unemployed or workers for wages, the proportion is 11 % unemployed and 89 % employed..
Figure 4 shows the means for each of the digital competence items. Unsurprisingly, the data showed that our survey respondent’s digital competence level was high according to the criteria usually applied for measuring digital competence in the wider population.
The results of PCA carried out to reduce the information of the six items into two scales are shown in Table 2. The 2-factor solution fitted well with the measurement of our theoretical concepts (factor 1 = information skills and factor 2 = interaction skills) and met the basic statistical requirements: eigenvalues of the factors are higher than 1 and the two factors together explain 80 % of the variance of the six original variables. These two factors will be used as variables in regression models which aim to answer aimed to respond to research questions 2, 3 and 4.
Finally, Table 3 depicts the average number of MOOCs participants were enrolled in the past and the average number of MOOCs they completed. For comparability and presentation reasons we distinguished between MOOC takers with or without previous experience in MOOCs. The majority of our respondents had followed one or several MOOCS. Only 19 % were participating in a MOOC for the first time. The mean number of MOOCs taken by the learners with previous experience with MOOCs was around 6 and most of them had taken more than one MOOC in the past. Participants enrolled on a mean of around 6 MOOCs and completed a mean of more than 4.
Influence of occupational setting on MOOC participation
Our findings (see Table 4) showed that MOOCs were an important part of non-formal learning for individuals who were facing difficulties in the labour market. The unemployed in our sample tended to participate in MOOCs more than the employed, and indeed the estimates showed that this is one of the most important variables for predicting the number of MOOCs a learner will enroll in.
When we focus on workers, the regression models showed how those who have their employer support for professional development activities participated less in MOOCs than those who do not have this support. This was true in the models where all learners were included. In order to determine what the main cause for this was, we compared the level of participation in non-MOOC learning activities of the two groups. We focused on participation in the last month and last week. The results showed that the number of hours devoted to non-MOOC learning was higher for learners with employer support than the hours devoted by learners without support (mean = 5.9 h/week and 4.9/week, t = −1.7321 p = 0.04 and 25 and 19 h/month t = −2.12 9 p = 0.016 respectively). This result supported the hypothesis that workers with employer support participate in MOOCs less often than those who have employer support because they participate more frequently in other training activities.
Influence of digital competence on MOOC participation
The level of digital competence in the areas measured played a key role in the decision to enrol on MOOCs. The results presented in Table 4 indicated that both information and interaction skills were important. However, our estimates for interactions skills were higher and more stable than those for information skills, revealing interaction as a key area of digital competence for MOOC participation. This result was consistent with previous research which showed how interaction skills were more important than information skills for taking advantage of and being successful in online courses (Bernard et al. 2009; Castaño-Muñoz et al. 2014).
Impact of digital competence in different occupational settings
When analyzing the interactions proposed in the model (models 2, 4, 6 and 8 in Table 4), the fact that occupational setting plays a moderator role of occupational setting in the effect of interaction skills was clear. The fact of being unemployed or employed with employer support to professional development maximized the impact of having good digital interaction skills. Therefore, unsupportive settings were hindering the possibilities of using communication competences for MOOC participation. This was especially relevant because as seen before communication skills were important determinants of participation in MOOCs.
On the other hand, the moderator role played by occupational setting in the effect of digital information competence was not stable among the proposed models. In the models where the whole population was included it seemed being occupied with support of the employer increased the effect of information competences. However, this relationship disappeared when we analyzed the regressions without including people who declared they had enrolled on more than 30 MOOCs. Indeed, in the regressions where we left out participants who were taking their first MOOC (models 7 and 8) the effect was reversed and negative. So our data did not allow to confirm that occupational setting plays a moderator role for information skills.
Impact of other variables
According to our estimates in Table 4, age played a positive role in MOOC enrolment. In addition, in the models where interactions were included a quadratic relationship was found which indicates that the positive effect on MOOC enrolment declined for older students. This relationship makes sense because young people tend to have more “fresh” knowledge and skills and need less up- or re-skilling. On the other hand, the quadratic relationship could be interpreted as a sign of older workers participating less in training activities simply because they had less career-time to benefit from the new skills acquired.
In all the proposed models women tended to participate in MOOCs less often than men. However, this trend was smaller in the models where extremes (people with more than 30 MOOCS) were eliminated, indicating that there were fewer women in the extreme end of the distribution. Finally, those respondents with the highest levels of education (second stage tertiary education) enrolled more often than those with first stage tertiary education.