• Aucun résultat trouvé

2. C ONTEXTUALIZATION , D ATA AND D ESCRIPTIVE A NALYSIS

2.3 C ROSS -S ECTIONAL A NALYSIS

What percentage of Swiss wage-earners are union members? Does this proportion change over time? What is the average profile of the unionized worker? Does the evolution of union density follow the same pattern in all segments of the Swiss population? These are the questions we answer in this section. Since the analyses are made on a purely descriptive cross-sectional level, the results we present are weighted using individual cross-sectional weights that inflate the sample size to the size of the Swiss population.

2.3.1OVERALL EVOLUTION

In the first sub-section we describe the evolution of union mem-bership for the overall population of wage-earners in Switzerland.

In order to ascertain the quality of the SHP data, we compare our results with those given by other sources.

Figure 2.1 represents the evolution of the number of wage-earners in Switzerland according to the data of Swiss Household Panel (SHP), the Swiss Labour Force Survey (SLFS) and the OECD Labour Force Statistics.

Figure 2.1: Evolution of the number of wage-earners in Switzerland

Sources: Swiss Household Panel (SHP); Swiss Labour Force Survey (SLFS);

Organisation for Economic Co-operation and Development (OECD)

The definition of wage-earner used in the SHP data and in the SLFS survey is the same one: it includes all people having worked as wage-earners for at least one hour or being wage-earners on temporary leave (sick leave, vacation, maternity leave, military ser-vice,...) during the week prior to the survey interview (Swiss Fed-eral Statistical Office 2012). According to both surveys, the active population in Switzerland increases from around 3'000'000 indi-viduals in 1999 to 3'500'000 wage-earners in 2011. The trend is in line with an overall augmentation of the resident population in Switzerland. Although this evolution is not of primary interest for our purposes, it is good to note how the results given by our source, the SHP data, are almost equivalent to those given by the SLFS survey. This is very reassuring regarding the reliability of the

SHP data, since the SLFS survey is by far the most reliable source on the subject, based on approximately 40'000 yearly respondents (Swiss Federal Statistical Office 2012). Regarding the OECD data, they are originally based on the SLFS survey, but include also self-employed people. This explains why the number of individuals in-dicated is higher than what the other two sources point out.

Figure 2.2 shows the evolution of the number of union mem-bers in Switzerland according to the SHP data, to the data pro-vided by the Swiss Federal Statistical Office (FSO) and to the OECD database on Institutional Characteristics of Trade Unions.

Figure 2.2: Evolution of the number of trade union members in Switzerland

Sources: Swiss Household Panel (SHP); Swiss Federal Statistical Office (FSO);

Organisation for Economic Co-operation and Development (OECD)

The data provided by the FSO are based on administrative sources collected by the country's largest union confederation (SGB) (cf.

sub-section 2.1.1). They include all registered union members, ir-respective of their working status. Unemployed, inactive or retired members are also counted among them. Hence, it is not surprising that the number of members according to this administrative source is always above the one we computed using the SHP data.

It is also useful to note that the FSO data for 2000 and 2001, where

there is a sudden drop in the number of members, are very prob-ably affected by some error in the collection of the data. The OECD data are based on the same source of the FSO, but they have been adapted in order to take into account only employed members. Finally, using the SHP data, we considered only union members being at the same time wage-earners, since they repre-sent our population of interest. We see that, until 2008, the mem-bers indicated by the OECD data are above those computed through the SHP data. This is not surprising since, although the OECD tries to adapt the administrative data in order to account only for employed members, the unions that provide the data usu-ally overestimate their real members. Concerning the evolution, according to our source, we see that the number of union mem-bers grows from around 650'000 memmem-bers in 1999 to 700'000 in 2011. The two administrative sources, on the contrary, indicate a slow decline of the number of members during the period under examination. It is not easy to interpret this trend, since we do not know if it concerns wage-earners or if it represents only a diminu-tion of non-employed union members.

The evolution of the number of union members is usually not so interesting in itself. A more appealing indicator is the union density in a given year, computed as:

𝒖𝒏𝒊𝒐𝒏 𝒅𝒆𝒏𝒔𝒊𝒕𝒚 = 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒖𝒏𝒊𝒐𝒏 𝒎𝒆𝒎𝒃𝒆𝒓𝒔

𝒕𝒐𝒕𝒂𝒍 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 (𝟐. 𝟏)

Figure 2.3, on the next page, shows the evolution of trade union density in Switzerland using the SHP data, a combination of the OFS and SLFS data cited before and the aforementioned OECD database on trade unions.

Figure 2.3: Evolution of trade union density in Switzerland

Sources: Swiss Household Panel (SHP); Swiss Federal Statistical Office (FSO); Swiss Labour Force Survey (SLFS); Organisation for Economic Co-operation and Development (OECD)

Our computations with the SHP data are based on the definition of union density given in equation 2.1. The OFS-SLFS data are computed as a ratio of the administrative records on union mem-bers represented in figure 2.2 and the number of wage-earners de-scribed in figure 2.1 according to the SLFS survey. The same com-putational procedure is adopted in the database of the Interna-tional Labour Organization (ILO). It is not surprising to see that the union densities we get with the SHP data are lower than the ones obtained through the OFS-SLFS data, since they include in the numerator also non-employed union members. The densities provided by the OECD are computed as the ratio between the number of union members and the number of wage- and salary-earners described previously for the two OECD databases. Since we do not know exactly how the number of union members has been computed and since the set of wage- and salary-earners con-sidered by the OECD includes more individuals than the defini-tion given in equadefini-tion 2.1, it is difficult to interpret the union den-sities provided by the OECD database. However, we can see that the three databases describe the same declining trend of union membership in Switzerland. Referring to the SHP data, which

seems the most reliable source, union density decreases from around 22 % in 1999 to 18% in 2011.

Before turning to the description of the cross-sectional evolu-tion of union membership across different segments of the Swiss population, we provide an interesting analysis (figure 2.4) of the evolution of the proportion of union members that declare them-selves as active and passive members.

Figure 2.4: Evolution of the proportion of union members declaring themselves as “active members”

Source: Swiss Household Panel (SHP)

The representation shows a linear drop of the proportion of mem-bers declaring themselves as active union memmem-bers. In 1999, around 50% of union members declare themselves as “active members”. The same proportion is halved in 2011, counting only 25% “active members”. This seems to show a clear evolution to-wards a change in the form of union membership, members be-coming more and more personally detached of their unions. How-ever, since the definition of active and passive membership is completely subjective, it is complicated to give a clear interpreta-tion to this trend. In the causal analyses of the following chapters, we will however show that an active membership is constantly as-sociated with a higher attitudinal effect than a passive one.

2.3.2EVOLUTION ACROSS SPECIFIC SEGMENTS OF

WAGE-EARNERS

After having described the evolution of union membership for the whole set of wage-earners in Switzerland, in this sub-section we repeat the same analysis on specific segments of the population.

These are the same sub-populations we will use in the regression models of the next chapters as interaction terms to observe how the effect of union membership varies depending on the profile of the individuals concerned. The decline of union density we have pointed out for the population as a whole may not follow the same trend in different sub-populations and it is useful to examine the variations between them. Studying the evolution of union membership for different groups of the population involves the analysis of the trends concerning each group and, more im-portantly, the comparison of the evolutions between the groups taken into account. In order to carry out these analyses, we use four indicators that describe different dimensions of the union membership evolution in each sub-population and the relation-ship with the trends observed in other sub-populations. Although in most cases the four indicators lead to similar conclusions, each of them is better suited to highlight some aspects than the other ones. We briefly describe each of them in the next paragraphs.

Some of the indicators may seem quite abstract at first glance, but they will become clearer when we exploit them subsequently in concrete analyses.

The first indicator is represented by the evolution of the num-ber of memnum-bers in each segment of the population considered. It constitutes a measure well adapted to observe short-term changes such as sudden unionization waves of particular segments of the population or a drop of membership in other ones. Moreover, the number of members in a given group determines its practical rel-evance and potential to influence unions' strategies. A group com-posed of 1'000 union members may certainly not be “heard” the same way as a group of 100'000 members. Also, mobilizing 1'000 or 100'000 individuals implies the use of different organizational dynamics.

The second indicator we use is the proportion of union mem-bers in each category of the population taken into account. For a category i, the proportion of union members in a given year is given by:

𝒑𝒓𝒐𝒑𝒐𝒓𝒕𝒊𝒐𝒏 𝒐𝒇 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊

= 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊 𝒕𝒐𝒕𝒂𝒍 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒎𝒆𝒎𝒃𝒆𝒓𝒔

(𝟐. 𝟐)

where all the quantities are taken from the same given year. It rep-resents an indicator quite similar to the number of members, but it highlights more explicitly the relative importance of each sub-population compared with the others. For example, if a propor-tion of 80% members belonged to a specific group, it is very likely that unions' strategies would be highly oriented towards the needs of that particular group. Also, the number of members and the proportion of members of a particular segment of the population can follow divergent evolutions. As an illustration, the number of members of a category of individuals may remain constant over time, while its proportion relative to other categories may increase if the number of members in other categories decreases.

The first two measures we described are usually highly depend-ent on the relative importance of each category in the population of wage-earners. If a large proportion of wage-earners belong to a particular group, this group is likely to be well represented in terms of number of union members even though only a small fraction of them joins unions. Conversely, a group composed only of a small number of wage-earners is not likely to represent a high pro-portion of members even though most of the individuals that be-long to it are members. In order to account for these dispropor-tions between groups related to their relative importance in the labor market, we compute (on the next page) trade union densities within each category:

𝒕𝒓𝒂𝒅𝒆 𝒖𝒏𝒊𝒐𝒏 𝒅𝒆𝒏𝒔𝒊𝒕𝒚 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊

= 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊

(𝟐. 𝟑)

This is the same union density concept we presented for the whole population of wage-earners, but computed within each category i, considered as a distinct sub-population of wage-earners. It repre-sents a measure of the probability of being a union member within each category, the propensity to join a union within each category.

Some groups may not have an important weight in terms of pro-portion of members they represent and yet have high union den-sities.

The first three indicators are well suited to allow us to compare the union membership dynamics between different categories of union members. As a fourth a measure, we would like to have a quantity that gives us the possibility to compare the relative im-portance of a given characteristic or of a given sub-population be-tween members and non-members. If the characteristic or the var-iable that defines different segments of a population is a numeric one, such as age for example, we take the mean of the character-istic among members and non-members. If the variable that we analyze is not expressed in a numeric scale, such as the level of education for example, we cannot compute a mean since the dif-ferent values of it have not a quantitative meaning. In that case, instead of a mean, we consider (on the next page) the ratio of the proportions in the population of union members and in the pop-ulation of non-members for each category of the variable taken account:

𝒓𝒂𝒕𝒊𝒐 𝒐𝒇 𝒕𝒉𝒆 𝒑𝒓𝒐𝒑𝒐𝒓𝒕𝒊𝒐𝒏𝒔 𝒃𝒆𝒕𝒘𝒆𝒆𝒏 𝒖𝒏𝒊𝒐𝒏 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒂𝒏𝒅 𝒏𝒐𝒏 − 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒇𝒐𝒓 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊

= 𝒑𝒓𝒐𝒑𝒐𝒓𝒕𝒊𝒐𝒏 𝒐𝒇 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊 𝒑𝒓𝒐𝒑𝒐𝒓𝒕𝒊𝒐𝒏 𝒐𝒇 𝒏𝒐𝒏 − 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊

=

𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒎𝒆𝒎𝒃𝒆𝒓𝒔

𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒏𝒐𝒏 − 𝒎𝒆𝒎𝒃𝒆𝒓𝒔 𝒊𝒏 𝒄𝒂𝒕𝒆𝒈𝒐𝒓𝒚 𝒊 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒘𝒂𝒈𝒆 − 𝒆𝒂𝒓𝒏𝒆𝒓𝒔 𝒏𝒐𝒏 − 𝒎𝒆𝒎𝒃𝒆𝒓𝒔

(𝟐. 𝟒)

This corresponds to the ratio of the proportions introduced in equation 2.3 computed separately for members and non-mem-bers. Each of the two proportions can be seen as the probability of drawing an individual belonging to category i by considering separately members and non-members. If the ratio is greater than 1, individuals of category i are overrepresented among union members. Conversely, a ratio less than 1 implies that the chances to belong to that particular category are lower for union members than for non-members. The principle behind it is the same as the comparison of the means of a numeric variable between members and non-members. As an example, if we consider as category the individuals that vote for a given party, the ratio of the proportions indicates whether the propensity to vote for that party is higher for union members or non-members. If a proportion of 20% of union members vote for that party, while among non-members the same percentage is represented by 10% of individuals, the ratio of the proportions would correspond to a value of 2, showing that a union member has twice more chances of being a voter of that party. The indicator is not of fundamental importance in this chapter, but it will be much more relevant in the descriptive results that precede the causal analyses provided in the next chapters.

This same indicator can also be computed for numeric variables, after having defined specific categories decomposing the numeric parameter. However, we prefer to compute the mean because of its higher synthetic power (we compare only one value between union members and non-members instead of one for each cate-gory).

The rest of the sub-section is dedicated to the presentation of the evolution of the four aforementioned indicators for different sub-populations of union members between 1999 and 2011. If

used to segment the population and then comment the four plots related to the four indicators. It is to note that we created the cat-egories of each variable in light of the causal analyses we turn to in the next chapters. When we merged the original categories into new ones for some of the variables, two main criteria have guided our choices. First, the merged categories had to make sense on a substantive level. There has to be some homogeneity between the units composing each category that gives a coherence to it, a uni-tary structure. This feature is needed in order to hope to make the heterogeneous attitudinal effects of union membership more ho-mogeneous within each category than in the population as a whole. Second, on a more practical level, we pay attention to cre-ate ccre-ategories that are not composed of too few individuals. In fact, since some of the estimators we use in the next chapters highly rely on asymptotic assumptions, we need to have a minimal number of individuals in order to carry out our analyses. Descrip-tive statistics on these sub-populations can be found in table 8.1, section 8.1 in the appendix chapter (the columns on cross-sec-tional data are those that concern this chapter).

As first variable, we consider the type of occupation (full- or part-time) of a wage-earner. Figure 2.5 (on the next page) repre-sents the four indicators by occupation. The first two plots show that, among union members, full-time workers are clearly more numerous than part-time ones. However, the trend is character-ized by an obvious decrease of the gap between the two categories over the period (the difference between the two proportions is around 50% in 1999 and drops to 20% in 2011). A similar obser-vation is drawn from the third plot, where the union density of full-time wage-earners is higher than the one of part-time ones, but with a gap that becomes narrower and narrower (the densities of around 24% for full-time workers and 17% for part-time work-ers at the beginning of the period approach both 17% in 2011).

Hence, the union density of full-time workers is declining, while the one of part-time ones remains more or less stable. The last graph leads to very similar conclusions when comparing the pro-pensity to work full- or part-time for union members with non-members.

Figure 2.5: Cross-sectional descriptive analysis of union membership by occupation

Source: Swiss Household Panel (SHP)

The sex of an individual (figure 2.6, on the next page) is the second variable we use to describe the evolution of union mem-bership across different sub-groups. The evolution we observe is very similar to the one of the previous case since the dichotomy full/part-time work is directly related to a dichotomy between the two sexes, women representing the majority of part-time workers.

In the first two plots of figure 2.6, we see that, in terms of number of members and of their proportion relative to men, women are less represented than men as union members. That is a foreseeable outcome since it is well known that, despite the progresses made in the last decades, women still occupy a less important position than men in the labor market. As a consequence, they have a lower probability of becoming union members. This is particularly true for Switzerland, where the movement towards the equality be-tween women and men regarding the access to the labor market has been slower than in other occidental countries (cf. sub-section 2.1.3). However, if we look at the trend, we see that the gap is being constantly reduced, the number of female union members increasing every year, while the male members remain more or less constant over the period (the proportion of female members rises from approximately 32% to around 40% in 2011). The third graph

confirms this aspect and shows how, when we account for the disproportion of the number female and male wage-earners (as it is done in the computation of union densities), the gap between

confirms this aspect and shows how, when we account for the disproportion of the number female and male wage-earners (as it is done in the computation of union densities), the gap between