B.: VÉANSE ASIMISMO LA RELACIÓN DE MATERIAL DE DE­ FENSA Y EL ARTÍCULO 1C111.

In the following, we document the main labor market statuses in Spain and describe how they evolved in the time period 2005-2017, thereby confirming our accuracy in constructing the dataset by showing that we are able to match key labor market facts as provided by official bodies like theOECDor the Spanish National Statistics Institution (Instituto Nacional de Estadística (INE)).

Labor Force. The composition of the labor force is plotted inFigure II·1. The largest part of the labor force is compounded by employed workers. At the beginning of the sample period this part is almost constant at 78% and declines due to the crisis from 2008 onward, until a share of approximately 70% is reached. This drop of 8 percentage points (p.p.) is absorbed in the unemployed individuals’ share which increases after the crisis by an equivalent amount. In contrast, the share of self-employed individuals stays roughly constant at 18%. A slight increase in the self-employment share is observable since 2013. Looking at the age distribution of the labor force,Figure II·2reveals that self-employment is more relevant for the older individuals (age groups over 40) than for younger individuals. The share of self-employed as percentage of the labor force is only around 10-15% for those younger than 40, whereas it ranges between 20-24% for those in the age groups above 40. A closer look at Spain’s labor force levels fromOECDdata reveals that a four percent sample should equal on average 913,000 individuals across the sample period (OECD 2018).18

Evolution of Labor Market Statuses. InFigure II·4Spain’s annual unemployment rate usingMCVLandOECDdata is illustrated for the sample period. The unemployment rates from both sources are based on individuals of working age and include all sectors, as well as all social security schemes, such that they are comparable. It is important to note that theOECDrestricts the working age population to individuals between 15 and 64 years old, whilst the INE’s Working Conditions Survey (Encuesta Nacional de Condiciones de Trabajo) focuses on individuals who are 16 years or older. We restrict our descriptive sample to individuals who are 18 years or older19. In spite of these differences, the computed unemployment rate usingMCVLdata is very similar to the quarterly unemployment rate reported byINE(left panel figure) and also matches OECD’s annual unemployment rate (right panel figure). Concerning the self-employment rate that is measured with respect to the total employment rate,Figure II·3reconfirms that our data cleaning process and the construction of our dataset from theMCVLdata enable us to match quarterly statistics fromINE (left-hand panel), as well as annual statistics fromOECDdata (right-hand panel). Specifically,Figure II·3in the Appendix shows that self-employment has been slowly rising until reaching a peak in 2014 at nearly 20% and then declining again.

18_{The Spanish average labor force level from 2005 until 2015 was about 22,817,000 individuals per year (OECD}

2018). Thus, a four percent sample should result in 0.04 · 22, 817, 000 ≈ 913, 000 individuals.

For completeness,Figure II·5illustrates part-time andFigure II·6temporary contract rates in Spain. Again, note that we compare our calculated data with both official statistics from Spain (quarterlyINEdata in the left-hand panels) andOECDdata (annual data in the right-hand panels). In particular, while the part-time rate has increased from around 10% to 15% during the whole sample period, the temporary contract rate reflects a U-shape evolution. This is in line with the observation that during an economic crisis temporary contracts are not renewed, and therefore this group of workers is among the first to be laid-off (as can be seen from the drop of around 30 to 23% during the crisis). In contrast, when the recovery started (in Spain around the end of 2013) temporary employment recovered first and in 2017 reached pre-crisis levels.

Labor Market Flows.Figure 2·1shows that the inflow into self-employment is consid- erably dominated by flows from unemployment. Thus, a relevant share of founders in Spain has been previously unemployed, and the inflow from unemployment into self- employment is important. It makes up 30-50% of all new self-employed every year.20 Moreover, the evolution of the composition of inflows into self-employment exhibits counter-cyclical patterns, especially from 2010 onward. The share of inflows from previously employed workers decreases, the one of previously unemployed individuals increases during a crisis. Although it is true that the outflows from unemployment to self-employment may only reflect 5% of the whole unemployment stock (Figure II·12), it should be noted that usually there exist job spillovers, i.e. most founders have employees. Figure 2·1:Composition of Inflows into Self-Employment excl. Stocks

0 10 20 30 40

Total of self-employed workers (1000)

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 OL E U 0 20 40 60 80 100 Total of inflows to SE (100%) 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 OL E U

Notes:These figures illustrate the yearly inflows to self-employment in Spain, in both absolute (left-hand side) and relative (right-hand side) terms. The sample consists of all individuals older than 18. We distinguish inflows of individuals from the relevant states: out of the labor force (OL), dependent employment (E), and unemployment (U). Source:Authors’ calculations based onMCVL2005-2017 data.

20_{In the Appendix, we show also the same figures including the stocks of self-employed. Looking at stocks} of self-employed individuals in our representative sample shows that about 80% of the self-employed remain in self-employment (SE) in the next year (less during the crisis period):Figure II·7shows the yearly inflows including theSEstock dimension,Figure II·8shows the same for outflows fromSE. The graphs confirm that new inflows into SEare mainly composed out of new self-employed individuals who were previously unemployed or employed. In particular, the share of new inflows toSEout of unemployment increases until around 2013.

Consequently, the economic significance of this flow is a multiple of the outflow statis- tics from unemployment to self-employment and quantitatively important. In other words, startups can be engines for economic growth. Note that the role of self-employment for the inflows into employment (Figure II·9) appears not to change much over time. This is also true for the outflows from employment to self-employment (Figure II·10), but different to the patterns when analyzing the outflows from unemployment. We ob- serve inFigure II·12that the share of individuals who transition from unemployment to self-employment remains quite stable during the years surrounding the reform, even though they are relatively larger than at the beginning of the sample period. However, the outflows from unemployment are clearly dominated by employment, especially during the years 2012 and 2013. Moreover,Figure II·11shows a similar pattern regarding the inflows into unemployment. The relative destruction of employment increases year by year until 2012/2013, then the economic recovery leads to a decreasing trend thereafter, by which the inflow into unemployment from employment declines.

Self-Employment Characteristics. Table II.2shows a comparison of the main char- acteristics of individuals in self-employment, compared to individuals in dependent employment. Regarding their socioeconomic features, we observe that there is a gen- der gap in the group of self-employed individuals: while 48% of the individuals in the sub-sample of dependent employment correspond to female workers, only 36% of individuals in the self-employment group are female. Moreover, the average age in the self-employment group is slightly larger than in the sub-sample of employees. We also notice how the distribution of education levels differs to some extent across groups: e.g. the share of highly educated workers is larger for self-employment than employment. This may be due to the fact that the service sector is more important for self-employment (39% of allSEspells).

Moreover,Figure II·13illustrates the composition of self-employment with respect to the sector in which the business has been started. We can confirm the fact that self- employment is indeed quite important in the construction sector. Figure II·13 shows that the share of workers in that sector increases until 2008, when it starts to decrease in favor of other sectors like transport, tourism and retail, but also professional, scientific, administrative and auxiliary services. Lastly, we obtain information on the average amount of experience and tenure in both groups, as well as concerning the spell duration and daily earnings. In this context, it becomes apparent that the average duration of the self-employment spells is remarkably larger than that of the employment spells (Table II.2). The self-employed individuals tend to remain longer in their position than the average employee, which indicates that many of the self-employed are successful, and that a majority of them are rather opportunity than necessity-driven entrepreneurs. Earnings. Figure II·16shows yearly mean earnings comparing our income data from either tax or from social security data. The mean annual labor income is about 20,000 Euro. It declined during the crisis and is recovering again since 2014, but still below the

2007 pre-crisis level. Figure II·17shows that monthly earnings translate into about 1,700 Euro, which corresponds to 60 Euro on a daily basis. This helps to ease the interpretation of the size of SpanishUIbenefit levels (relative to other countries). Figure II·15shows that the distribution of monthly earnings is skewed to the left with a large dispersion across top incomes, but that most Spanish citizens earn income that is below the mean.21