**3. Results**

We started by geographically localizing the relevance of support for secession using the o fficial results of the last regional elections (21 December 2017), and adding the votes obtained by the three main current secessionist parties: JxCat (right), ERC (left) and the CUP (far left). These regional elections are the event where secessionist support is more clearly revealed, as majorities at the Autonomous Parliament and the Government depend on them.

As in Figure 1, when secession support at each municipality was greater than 50% of the electoral census, they were indicated in red. The highest values were found at the smallest municipalities, while they progressively dropped at the largest municipalities of Barcelona and Tarragona conurbations. Barcelona city showed intermediate scores. The main trend is exemplified with the following two results: Santa Cecília de Voltregà: Electoral census 163 citizens, secession support: 89.0%; Badía del Vallés: Electoral census 10,560 citizens, secession support: 14.5%. Figure 2 presents an image complementary to the previous one, showing percentages of support for unionist forces at each municipality with di fferent shades of green.

The summary of the results for Figures 1 and 2 appears in Table 1 (from the electoral data of 21–12–2017, the last regional elections):

Secessionists are majoritarian in three quarters of the territory, but less than 25% of the electoral census live in those parts of the region, while in the remaining part, more than 75% of the census is concentrated.

We developed also an alternative method to visualize the di fferential distribution of the pro-secession vote, presented in Figure 3. The municipalities are ordered in terms of an index, developed from a factorial analysis obtained from electoral data in general and regional elections throughout the period 2008–2012. This index, which shall be referred to hereafter as the unionism index, measures the degree of support for unionist forces at each municipality, from low to high degrees. The accumulated census and percentages of support for secessionist forces are represented for each municipality. The percentage of secession support is represented through a smooth curve, based on moving averages.

#### *3.1. Longitudinal Changes of Household Net Incomes on the Main Family Language Segments*

Based on Household net income levels among citizens with family/mother language Catalan and those with family/mother language Spanish, we estimated the medians of household net incomes for each linguistic segmen<sup>t</sup> throughout the period 2006–20194. Citizens with family language Catalan exhibited higher household incomes throughout the entire period compared with citizens with family language Spanish. The trend line suggests that the economic crisis increased the magnitude of such inequality (Figure 4).

**Figure 4.** Evolution of median estimates of household net incomes among citizens with family/mother language Catalan vs. those with family/mother language Spanish. Observe the di fferences between these groups and how they increase along the economic crisis. Primary source: CEO barometers 2006–2019. FML: Family/mother language (FML Spanish: 56% of Catalonian citizenry; FML Catalan: 36%; FML both Catalan and Spanish 6%; source: EULP2018-Enquesta Usos lingüístics de la població, Institut Estadístic Catalunya; https://www.idescat.cat/pub/?id=eulp).

Figure 5 displays the percentages of "Yes" in a (hypothetical) referendum of self-determination in population segments determined by family/mother language and the binary variable "household net incomes ≥3000 €/month" with two values "high" (greater than 3000 €/month) or "low" (lower than 3000 €/month). The di fference between these idiomatic segments was huge throughout the 2015–2019 period. There were also di fferences depending on income: Wealthier respondents were more in favour

<sup>4</sup> Until summer 2011 the survey question explicitly asked for 'family language', and after that for 'childhood language in the family'. This change resulted in a decrease in the percentage of people who answered 'both languages' and an increase in the Spanish-mother language segment, but it did not significantly affect other variables within the surveys immediately before and after that change.

of secession in both idiomatic segments. Finally, a slight tendency to increase support of secession appeared among Spanish speaking citizens with lower incomes.

**Figure 5.** Covariation between household net incomes and family/mother language segments on support for secession in a (hypothetical) referendum of self-determination. Attached to each label comes respondent percentages, for each segment, at a July 2019 CEO survey; rests up to 100% were DK/NA responses. 1M = one thousand euros/month. FML: Family/mother language.

## *3.2. Longitudinal Variations on National Identity Feelings by Household Net Incomes and Family*/*Mother Language Segments*

To obtain a more detailed picture of the potential relevance of economic variables in influencing the division between secessionism and unionism we analysed the evolution of "national identity" feelings in the main two linguistic segments taking into account the variable "household net incomes ≥ 3000 €/month", with two values "high" (greater than 3000 €/months) or "low" (lower than 3000 €/month). First, we directed focus (Figure 6) to the national identity "only Catalan" for Catalan speakers (as defined by family/mother language) and for Spanish speakers (also defined by family/mother language).

In both idiomatic segments, wealthier strata moved towards an increase on the national identity "only Catalan", although the jump was much more substantial on the fraction that had Catalan as family language. Observe the distance between the biggest group, family/mother language Spanish, with household net income not greater than 3000 €/month, in green (38.2% of respondents, at that survey), and the family/mother language Catalan group with household net income greater than 3000 €/month (7.2% of respondents, at that survey), in red. Notice also that trend lines overlap or almost overlap at the beginning of the period and we may conjecture that Catalan subgroups (red and purple) should have had similar figures in the recent past, around the start of the present century. The sizes of the purple and orange subgroups were 20.7% and 6.7%, respectively, at the last survey.

157

**Figure 6.** Evolution of "only Catalan" national identity depending on family/mother language and household net income (2006–2019). The size of this identification type went from 14.2% of the entire Catalonian population in a March 2006 survey, up to 23.5% in July 2019. Attached to each label are size percentages, for each group, at a July CEO 2019 survey. FML: Family/mother language.

These abrupt variations on the national identity "only Catalan" were clearly dependent on family language and on household net incomes ≥3000 €/month. Their respective relevance was examined through a standard analysis of covariance. The dependent variables were percentages of "only Catalan" national identity, considering a factor group (four levels), and time (year) as a covariate, with an adjusted R-squared of 0.95282 with a very significant global p-value < 2.2 × 10−16. The group factor and the interactions were highly significant (*p*-values < 1.282 × <sup>10</sup>−15). Furthermore, comparing the slopes of the regression lines, we found that almost all contrasts showed highly significant differences (p-value, obtained by the Tukey method, were equal to 0.0001), with the exception of the slopes among lines within family/mother language Catalan (p-value, 0.8051). Contrasts between lines within family/mother language Spanish groups were slightly significantly different (p-value, 0.0425). Other results from this analysis were also obtained and can be supplied by the authors to the interested readers on demand

We repeated the same approach to the national identity "as Catalan as Spanish" (Figure 7). Variations were smaller and there was a big degree of overlap at the start of the period in Spanish speaking segments. The same analysis of covariance was repeated considering the percentages of the "as Catalan as Spanish" national identity as a dependent variable, with the same factor group and time as covariates, as before, with an adjusted R-squared of 0.9397 with a very significant global *p*-value < 2.2 × 10−16. The effects of group factor and the interactions were highly significant. Moreover, comparing the slope of the regression lines, only those within family/mother language Spanish were significantly different (p-value obtained by Tukey method were equal to 0.0003). The slopes of household net incomes greater than 3000 €/month segments, comparing family/mother language Catalan versus the corresponding Spanish were also different (*p*-values obtained by Tukey method, equal to 0.009). Other results from this analysis were also obtained and can be supplied by the authors to interested readers on demand.

*Genealogy* **2020**, *4*, 19

**Figure 7.** Evolution of "as Catalan as Spanish" national identity depending on family/mother language and household net income (2006–2019). The size of this identification type went from 42.5% at a March 2006 survey, to 39.2% in July 2019. Attached to each label are size percentages, for each group, at a July CEO 2019 survey. FML: Family/mother language.

The same analysis of covariance was finally applied to the percentages of "only Spanish" national identity as dependent variable (Figure 8), with the same factor group and time as covariates, as before, with an adjusted R-squared of 0.7906, a worse model fit than the previous cases, but still with a very significant global *p*-value <2.2 × 10−16. The effects of group factor and the interactions were highly significant. In this case most of the comparisons between the slopes of regression lines were not significantly different, with the exception of the contrast between family language Catalan–household net income greater or equal 3000 €/month compared with family language Spanish–household net income less than 3000 €/month (p-values obtained by Tukey equal to 0.0284). Observe in Figure 8 that wealthier households of Spanish language had less polarized national identity (only Spanish), just the opposite of the wealthier Catalan family language segment. Observe also that scores for national identity "only Spanish" attained total levels that were 15 percent lower than those of national identity "only Catalan". The apparently large oscillations in the figure are due to the tiny sizes of subgroups of "only Spanish" national identity.

**Figure 8.** Evolution of "only Spanish" national identity depending on family language and Household net income (2006–2019). The size of this identification was pretty stable across the whole period, with levels between 6% and 7%. Attached to each label re size percentages, for each group, at a July 2019 survey. FML: Family/mother language.

To complete the analysis, we considered both dependent variables simultaneously (MANCOVA), obtaining that group factor, the covariate Time and the interactions were highly significant. Again, other standard results for this analysis were obtained and can be supplied by the authors to interested readers on demand.

## *3.3. Other Economic Measures*

Finally, we compared the degree of secession support among different groups obtained considering their reported economic resistance limits (in months), in case of an economic breakdown (Figure 9). Results showed that secessionism significantly increased with higher economic resistance and endurance: People with higher financial resources were much more in favour of secession. A very similar trend appeared when the measure was perception of the own economy during last year. When that perception improved, support for secession was higher as well. In all, these findings consistently indicated that the recent secessionist wave in Catalonia has been sustained by those society segments that enjoy better economic resources and higher well-being.

**Figure 9.** Support for secession and Economic wellbeing. **Left:** Secession was much more popular among those with higher economic resistance limits (in months), in case of an economic breakdown. Percentages are means from CEO barometers 2016–2017. **Right:** It was also more popular among those with a good perception of the evolution of their economies during the last year. Percentages derived from CEO barometer July 2019.
