Analyzing data is, who are you going to do, one of the most enjoyable things in the world in my humble opinion. And when he had, then, begun to do so-pleasurable task with the data of my thesis, will not be more -even if very preliminary – show some early results.
Data and Methodology.
The reader may well skip this section if you want to go direct to the results. But there is that describe from where they come out of these things: The source of the data is the Social Protection Survey, which allows you to work with work stories (beginning from 1980). Now well as the work histories in the EPS are collected by memory, and we know very well that the memory of the interviewees cannot be required too, we thought that 1999 (3 years of the first wave, conducted in 2002) may be a good starting point. It completes the 2008 way of having a full decade -that allows ‘to happen’ different things in the work history-and because it is the last year for which we have complete information on the EPS, (EPS, 2009, the last published, has no data of all the cases for all months in 2009).
We used only cases with complete information (i.and without gaps, and which would respond to all the rides), and is used -given that we are interested in own-account workers – persons cases that have at least one of their working situations as a worker of this type. This results in a total of 3.282 cases for analysis.
The analysis was sequence analysis (used, in particular, the package TraMineR for R –manual here on-line). The sequences were generated through the combination of employment status (working, unemployed, inactive) and occupational category (employer, own-account, wage-earner, the other -for example, Unpaid Family). They were simplified for the analysis 4:
- Employer
- Own Account
- Dependent (includes paid and other)
- Without Work (unemployed and inactive).
Others had less than 1 month average per person-which in general was better -in order to simplify the analysis – to reunite with wage-earners (all the others were, in one way or another, dependent). And though the difference between the unoccupied and inactive is very important, for our purposes what is crucial is simply that none of them works).
To generate segments -that is one of the common uses of this type of analysis – was used as a measure of distance, Optimal Matching (there are other alternatives, but OM has more time and some better you know, even when it has been discussed quite limited in the literature), and the groups were generated using a hierarchical cluster (using Ward as mechanisms of agglutination).
And now to the data.
General Results.
Before describing the segments, it is of interest to -if at all – to determine some general characteristics of the stories: how much time is spent on our people in each state? Why many work situations spend people? What type of sequences of states are the most common?
The results of all these questions lead to a clear conclusion: own-account workers are not, finally, a set of people apart and clearly separated from the rest.
In the first place, because among those who have had this experience, the percentage of time (a full decade) that are as own-account workers is less than 50%. Specifically, 46.2%.
Table 1. Times Means by Condition.
| Variable | Entrepreneur | Own Account | Dependent | Without A Job |
|---|---|---|---|---|
| N° Months (10 years) | 5,5 | 55,4 | 29,8 | 29,3 |
| Percentage | 4,6% | 46.2 per cent | 24,8% | The 24.4% |
In the second place, and clearly associated to the above, because you only have experience to be your own account, is relatively low among these workers: Only 11% of the trajectories are the trajectories of people who have only been self-employed. The graph also shows that in our data there are a variety of sequences (the first 10 add up to only 44.2% of them)
Figure 1. 10 Types of Sequences are more Frequent (in Percentage)
This diversity of sequences (and we’re seeing just in terms of steps, not taking into account the duration) then also ask, how many situations happens to someone who has been self-employed in a decade? The numbers on the matter are quite clear, in the average 3,98 situations (that is, the duration for the situation is around 2.5 years on average). The median is 3, and quartile 1 occurs in 2 situations and quartile 3 has 5 situations (that is, 50% of the cases have between 2 and 5 different situations in 10 years). The maximum in the sample is of 28 different situations (which is almost 3 per year). If we take into account that the extent of change in the situation is just the change of job category and does not include, for example, change work being both wage earner, we can observe the instability of these trajectories.
The diversity of trajectories sent immediately to the following question: What types of paths are there?
3 types of trajectories between own-account workers.
A hierarchical classification has the advantage that it allows, by observing the dendrogram in full, to visually determine the best number of groups, and also allows you to set how they relate to each other. The dendrogram generated from this analysis shows that 3 is the number most appropriate to classify (as shown clearly in figure 2). In fact, do some analysis with another number of groups (6, which corresponded to the point at which each of the main divided), but did not return major differences. It also shows us the dendrogram that within these 3 core groups there are two that are more close to each other (the right group is the first difference). In the charts and tables presented below, then Groups 1 and 3 are closer to each other than what they are in relation to the Group 2.
Figure 2. Dendrogram of Work Stories Category.
A first way of determining what they consist of these segments consists in observing the evolution of states over time, which is what is shown in the following graph (the scale is month, so that y229 refers to the month 229 from the start of the period of the EPS, which is January of 1980, for publication will be corrected these problems)
Figure 3. Evolution of the United Labor between Segments
The results show us that:
- The G1 (40.2% of the population) is clearly a group of high predominance of own-account workers (even when this is somewhat lower later)
- The G2 (36,1%) is a group of wage workers who live with periods of independence (and be without a job)
- And finally the G3 (23,7) is a segment of people who do not have work, but when they do they are own-account workers.
One of the segments that appeared in the analysis with 6 groups was one that combined being an employee and then passed to be own account worker (and was the only one that had a personality of its own when broken down in greater detail the groups), this group is subsumed in the G1 in the current classification.
These results are also shown with clarity when one examines the average time of each group in each of the states as is done in the following table:
Table 2. Average times for State in each Segment (Percentage)
| State | G1 | G2 | G3 |
|---|---|---|---|
| Entrepreneur | 7,7% | 3,1% | 1,7% |
| Own Account | 76,5% | 24,6% | 27,7% |
| Dependent | 6,4% | 57,1% | 6,8% |
| Without a job | 9,3% | 15,2% | 63.8 per cent |
Actually the G1 is a group centrally of own-account workers. Now, even in them most has only been self-employed: 27% of the segment has that kind of trajectory. It is also the group that has the largest presence of business situations (in fact it is so common as to be dependent to them).
The G2 is a group of employees who also have experiences of self-employed: 57% of the time as a wage-earner, and 25% as own account. It is interesting that this group has a higher non-permanence work that the G1 (of 15.2% against 9.3%). It is also a very diverse group in their types of trajectories: the 10 most common add up to 35.8% of the group total.
The G3 is defined by the fact of being inactive (64%), and that when it works it tends to be self-employed (28%), more dependent. In fact, the 4 most common sequences in the group are only movements between being employed and being out of work.
The references to the diversity of paths also makes us wonder about the number of work situations. It is relevant to mention that only in G1, it is possible to have a single state (i.and the state of being only own stay in that group, and given that the universe has this condition, it is not possible for a single state in the other groups). The graphics Boxplot shown below are an indication, in any case, the difference between the groups in this respect:
Figure 4. Boxplot of Number of business Situations by Segment
The results show that clearly the group 2 is the largest dispersion (the only one where the interquartile range exceeds the 5 situations, and where more extended dispersion upwards-even when it is the G3 which has the greater dispersion absolute, a product of some outliers. On the other hand, the G1 has a scattering rather more low -Quartile 3 is located level with or below the level of Quartile 1 of the other segments. In other words, despite the fact that be your own account, is a relatively unstable, and tends to combine with other alternatives, those who focus on be own-account workers have greater stability (and therefore fewer changes of state).
Between the instability and the stability, the diversity and the combinations of state there is a lot to say. And that we have not taken into account until now none of its social location (we have not said who they are), nor of their other job characteristics (for example, the types of trades or sectors of the economy work). But all of those things are for other entries on the topic.