Is There a New Service Proletariat? Post-industrial Employment Growth and Social Inequality in Spain.

In Spain the absolute number of employed persons has increased from about 12,300,000 persons in 1994 to 19,300,000 at the end of 2005. In the same period, the number of immigrants has increased from about 500,000 to more than 4 million. The aim of this paper is to analyse the implications of these changes for social inequality. In particular, we investigate whether a new type of unskilled service class is likely to emerge as a distinct social class. We address this issue by means of three more specific research questions. The first one refers to the pattern of changes in the employment structure by occupational class: has the mentioned employment growth implied an expansion of the swelling service proletariat? The other two questions refer to the issue of demographic class formation: which is the composition of the class structure by gender and country of origin? And, are unskilled service occupations stop-gaps springboards towards better positions, or are they long-term traps? We answer these questions by means of a dynamic analysis of the panel data of the Spanish Labour Force Surveys (SLFS). We study trends over time in the class structure and, then, analyse upward mobility chances and the risk of falling into unemployment from unskilled occupations from year t to year t + 1.

Introduction

Although sociologists often tend to over-emphasize social transformations and stress the novelty and uniqueness of the times they happen to live in, it seems undeniable that the last decade has been a period of impressive change in the Spanish labour market and society. First, in the second-half of the 1990s and first years of the 21st century, Spain has experienced a remarkable growth in employment, much above the European average. The absolute number of employed persons in Spain has increased from about 12,300,000 persons in 1994 to 19,300,000 at the end of 2005, with a net employment creation of more than 7 million persons. The largest share of this growth in employment has occurred in the service sector. Second, from being a country with a modest proportion of foreign workers, Spain has become an immigration country. The most recent estimates show that the number of immigrants in Spain has increased from about 500,000 in 1996 to more than 4 million in 2005 (i.e. from about one per cent of the population to 8 per cent of the total population).

In this paper, we will study the implications of these changes in the labour market and immigration for social inequality. As the title suggests, we are interested in investigating whether a new type of unskilled service class is likely to emerge as a distinct social class. This type of research question was put forward in the early 1990s by Esping-Andersen (1993) in a comparative study on social stratification in six post-industrial societies. However, the empirical analysis presented in Esping-Andersen (Ibid.) mostly referred to the 1970s and 1980s. In other words, the period investigated precedes to a certain extent the full advent of post-industrial transformations. Some years later, Carabaña (1996) replicated Esping-Andersen's study for Spain, but his analysis is also based on data for the 1980s. It seems, therefore, timely to rescue the research agenda on an emergent service proletariat in advanced societies for more recent years. We do so by addressing the problem of contemporary class formation for the Spanish case by means of three more specific research questions.

The first one is at the macro level and refers to the patterns of change in the employment structure by sector and occupational class: has the mentioned growth in employment implied an expansion of the swelling service proletariat? The other two questions analysed address the issue of contemporary class formation: what is the composition of the unskilled service occupations by gender and, given the importance of the new phenomenon of immigration, by country of origin? Also, are these occupations stop-gap springboards towards better positions or are they long-term traps? In order to answer these questions we have studied trends over time in the class structure and, then, concentrated on the chances for upward mobility and the risk of falling into unemployment from unskilled occupations.

The structure of the paper is the following: In the next section, we will present the theoretical framework of our study. In the third section, we will describe the data, methods and variables that we have used in our empirical analysis. Then, we will present the main results and in the last sections we will draw some tentative conclusions.

Theoretical Framework

Post-industrial theory has dealt with two interrelated issues of social stratification: the overall transformation of the occupational structure and the possibility of the emergence and formation of new post-industrial social classes (Esping-Andersen, 1993). We will discuss these two issues separately.

Post-industrial Changes in the Occupational Structure

Post-industrial theory has produced a full set of predictions on the evolution of the occupational structure in advanced societies that range between the most optimistic and the pessimistic scenarios. One might perhaps find the most authoritative voice of the optimistic band in Bell's (1973) classic book on the coming of the post-industrial society. By focusing on the incipient electronic industry in the early 1970s, Bell concludes that science and theoretical knowledge are more and more crucial for technological innovations in production. In this scenario post-industrial society is characterized by the rise of professional and technical occupations, while automation reduces the traditional blue collar occupations in industry. Thus, Bell forecasts that a knowledge-biased shift from industry to service brings about an overall skill up-grading in the occupational structure.

The critiques to this ‘rosy picture’ stress that Bell (1973) overlooks the bottom of the occupational structure. Given that an automatic up-grading and flux from unskilled employment in industry to professional and technical employment in the service sector is unlikely, the question is whether and to what extent low-skilled occupations in the service sector might also expand. All the pessimistic scenarios derive from Baumol's (1967) model of unbalanced growth. The core of Baumol's argument is that economic activities can be divided into technologically progressive industries (that include manufacturing and agriculture), with constant increase in labour productivity, and non-progressive ones (that include services), where labour productivity can be assumed to be almost constant over time. One possible outcome of this model is that, if wages in the service sector follow those in manufacturing, service labour might ultimately out-price itself, since the increase in wages does not correspond to a real increase in productivity. The cost disease of service labour might imply a dying-out of service employment or, what is equivalent, mass unemployment. Much in the same vein, Gershuny (1978) predicted that costs differentials between services and final goods would encourage household self-servicing instead of the purchase of services in the market. In other words, households would tend to substitute highly expensive services in the market with relatively cheap durable goods that facilitate the self-production of equivalent services. If, on the other hand, labour prices in the service sector are allowed to adjust to productivity differentials, service wages would remain relatively low. This implies that household demand of market services will increase and (low wage) service employment would grow.

The productivity differentials and the mechanism of wage formation across sectors highlighted by Baumol's model implies that post-industrial societies face the unpleasant trade-off between the expansion of unskilled low-wage occupations or a high level of unemployment. There is, however, a third way out of this dilemma: the expansion of subsidized welfare state jobs (Esping-Andersen, 1999). This solution characterizes Scandinavian countries where bottom-level service jobs in the public sector have expanded. These low-skilled welfare state jobs, that are largely held by women, offer wages that are not so low, relatively secure contracts and, most important, some opportunities for ‘on-the-job promotion’. One might argue that this third alternative falls in between the more optimistic and the pessimistic predictions. On one hand, it is pessimistic in that it recognizes that the solution to the problem of mass unemployment must include the expansion of low-skilled service jobs in the public sector. But at the same time it offers some light, to the extent that, in the public sector, there are institutionally supported chances of upward mobility out of disadvantaged occupations.

Comparative research in the 1990s has shown that the emergence of one of the three alternative routes of post-industrial change in the occupational structure (low-wage jobs, mass unemployment or unskilled welfare state jobs) depends on the institutional configuration of a given country (Myles and Turegun, 1994). Structural changes in the production market are interpreted as the underlying engine of the post-industrial shift, but its direction and quality is historically and socially dependent. The literature on production and welfare regimes in the 1990s has argued that the social bases of post-industrial employment rest crucially on the wage-setting mechanisms and labour market regulation (State-market nexus), the public supply of care services (State-family nexus) and the emergence of new household types (double earners, single parents, and singles) that suffer from high time constraints and are less likely to rely on self-servicing solutions (Esping-Andersen, 1999; Soskice, 1999). At the beginning of the 1990s, the analysis of the industrial relation systems and of the welfare-market-family triad leads to very gloomy predictions for employment in Spain. According to Calmfors and Driffill's hypothesis of a hump-shaped relationship between unemployment and the level of coordination in the wage bargaining system, no substantial job growth was to be expected in countries such as Spain, that combined strong union power and an intermediate level of centralization and coordination (Soskice, 1990). At the same time, the welfare state did not promote a large expansion of social services and the traditional one-earner family was still relatively common and favoured self-serving options. Thus, southern European countries were pointed out as the worst-case scenario of the post-industrial shift, with limited possibilities of job creation and a deepening gulf between the insiders with secure jobs in a ‘frozen Fordist employment structure’ and the outsiders, largely marginalized from the labour market (Esping-Andersen, 1996a,b).

Post-industrial Class Formation

A second important issue raised in Esping-Andersen (1993) is whether post-industrial changes in the occupational structure bring about the emergence of new social classes and, in particular, whether low-skill service occupations are becoming the site for the formation of a new service proletariat. According to neo-Weberian class theorists, a set of occupations coagulates to form a social class when there is a connection enduring over time between the occupations and the individual biographies of the subjects who fill them. A crucial prerequisite for class formation is that its potential members remain in the positions that define the social class over time (Goldthorpe, 1983). In this respect, social closure is a necessary precondition for the transformation of a set of occupations into a specific social collectivity. This type of process is defined as ‘demographic class formation’ to distinguish it from the ‘socio-political class formation’ that implies some sort of collective action and class consciousness (Crompton, 1993). In this paper we are interested in investigating the first type of process, that is, demographic class formation of the unskilled service occupations.

The problem of class emergence can also be framed in terms of Schumpeter's famous metaphor according to which classes are like an omnibus that is always full, but always of different people. One might thus argue that demographic class formation occurs when the same passengers remain on the same bus from the departure to the end of the trip, that is, throughout their life course. Following this approach, in order to address the issue of class formation, one has to analyse the career trajectories and flux of mobility between given class situations. In the case of the unskilled service occupations, there are two possible scenarios. In the pessimistic one, unskilled service occupations are dead-end career roads that offer little chance of upward mobility and a high risk of carousel in-and-out from unemployment. In this scenario, bottom occupations in the service sectors become permanent traps characterized by strong social closure and the biographies of their incumbents end up being pretty similar in terms of past, current, and future life chances. In this case, we would have the condition for the demographic formation of a new type of service proletariat. In the more optimistic scenario, unskilled service occupations are stop-gap jobs that function as entry doors into the labour market for youth and immigrants or as re-entry doors for women who had previously interrupted their careers. In this situation unskilled occupations are springboards towards better positions in the occupational structure and do not have the potential to coagulate into a new social class.

Data, Variables, and Methods

The data used in this paper come from the Spanish Labour Force surveys (SLFS). This survey is carried out four times a year and in each quarter about 175,000 persons are interviewed. The SLFS sample is composed of six rotation groups and in each quarter one-sixth of the sample is replaced. To put it differently, about one-third of the original sample in the quarter k of year t is still part of the sample in the quarter k of year t + 1. Therefore, the SLFS provides a 1-year panel design that enables us to follow one-third of the sample in year t (about 65,000 persons) up to year t + 1.

We used data from the SLFS for different years to study change in class structure and in its composition by gender and country of origin. Then, we drew on the 1-year panel design of the SLFS to study outflows from unskilled occupations. We chose the 2003–2004 panel of the SLFS, the most recent one available when we started the empirical analysis. In order to analyse changes over time in patterns of class mobility, we also considered the SLFS panel for the years 1994–1995.

We chose to use the SLFS panel data rather than other longitudinal data sources for various reasons. First, the SLFS data are available for current years, while the only truly retrospective survey for Spain (the Socio-Demographic survey) that would probably be best suited for our research question was carried out in 1991 and is therefore rather outdated. Second, when compared to the data of European Household Panel survey, the SLFS sample is much larger. This feature is important since we are studying mobility patterns out of the unskilled service occupations and the number of observations falls rapidly when one focuses on a single occupational class. Third, the SLFS includes a specific sample design for foreigners who are largely under-represented in other surveys. The main weakness of the SLFS is that it allows us to investigate only 1-year time intervals and it is, then, problematic to infer longer term mobility rates from the observed 1-year rates. In order to tackle this limitation, we have used a series of cross-sectional SLFS to generate synthetic cohorts of entry into the labour market. We have, thus, defined synthetic entry cohorts by considering the year in which the individual left the educational system assuming that it correlates with his/her entry into the labour market. For this analysis we have considered only Spanish-born people. Immigrants were excluded because synthetic cohorts are based on subsamples of a given population which is assumed to be relatively stable over time. Given the rapid increase of immigration stocks in Spain in the last decade, assuming stability in the population of reference does not hold. We have, then, followed the employment patterns of Spanish-born synthetic cohorts from 2000 to 2006. In this way, we are able to put together information of different samples of the same entry cohort in different years. Consider, for instance, those who have left the educational system and, under our assumption, entered into the labour market in 1999. One can track the class distribution of this same entry cohort for each year between 2000 and 2006. Imagine that, for them, the percentages of those employed in the unskilled service class were 12 per cent in 2000. If this percentage sharply declines over time and, accordingly, the percentage of skilled occupations increases, we could then argue that for them the unskilled service occupations have served as stop-gaps springboards towards better positions. If, on the contrary, this percentage is stable we would have an indicator of certain class closure, even if we can not exclude a possible carousel effect, i.e. that people move forward and backwards from unskilled to skilled occupations.

With regard to the occupational structure, we have employed an adapted version of Esping-Andersen's (1993) class scheme. Given the specificity of the Spanish employment structure, we have not omitted occupations in the primary sector. Moreover, we have considered self-employed as a distinct class and we have further separated unskilled manual workers in construction and manufacture. We have used ISCO equivalent codes with three digits to distinguish the following major 10 occupational classes: unskilled workers in agriculture, agricultural self-employed, unskilled manual workers in manufacture, unskilled manual workers in construction, skilled manual workers, sales and clerical non-manual workers, urban self-employed, managers and proprietors, unskilled service workers, skilled service workers, technicians, and professionals.

In order to study the issue of class formation, we focus on unskilled occupations in year t and define six different destination states in year t + 1: (1) upward moves, meaning occupational moves into skilled occupations (including both short- and long-distance moves), (2) lateral moves, meaning occupational moves into other unskilled occupations (for instance, from unskilled service worker to unskilled worker in manufacture), (3) stability, (4) transition into unemployment, (5) transition to housekeeping, and (6) transition out of the labour market. We study yearly outflows from unskilled occupations towards these six possible destinations by means of a multinomial logit model.

The other independent variables of the multivariate analysis include gender (a dummy variable taking on the value one if the individual is a woman) and level of education (coded with five values: 1 = primary or no education, 2 = compulsory education, 3 = secondary education, 4 = vocational education and 5 = university education) and the sector of employment (that is a dummy variable with value ‘1’ if the occupation is in the public sector and ‘0’ otherwise). Previous studies on the determinants of mobility out of the unskilled occupations have regularly shown that women and those with little education are most likely to remain trapped in unskilled service occupations (Esping-Andersen, 1993). Moreover, there is evidence for Denmark that those who are employed in the public sector are also more likely to move upward (Esping-Andersen et al., 1994). In interpreting our results of these determinants of the mobility rates in Spain we will, therefore, be able to highlight differences and similarities with other post-industrial countries. Moreover, given the abrupt increase in immigration in the last 10 years, we have considered a dummy variable that has a value of ‘1’ for those who were born in Spain, in the European Community prior to unification or in other advanced economies and ‘0’ otherwise. This variable is available only for the period between 2003 and 2004. Finally, we have analysed whether the likelihood of a move depends on the time already spent in the unskilled occupation by means of a set of dummy variables that identify durations in the unskilled occupation which are shorter than 1 year, between 1 and 5 years and longer than 5 years. This variable is also available only for the period between 2003 and 2004.

The Results of the Empirical Analyses

The empirical analysis is divided in two parts. First, we will study changes in the occupational structure over the last three decades and, then, we will address the issue of the possible emergence of a new service proletariat.

Occupational Employment Shift

Table 1 shows the overall change in the occupational structure that has occurred in the last three decades in Spain. This table reproduces for the Spanish case a table with a comparison of the occupational structure in Germany, Sweden, and US presented in Esping-Andersen (1999). Three results seem notable. First of all, at the end of the 1970s, one Spanish worker in five was still employed in agricultural occupations. As noted, de-ruralization takes place in Spain in the years when the Fordist model of industrial mass production was already entering into crisis in the United States and other European countries (Marimon and Zilibotti, 1998). This implies that Spain has experienced an almost direct shift in the occupational structure from pre-Fordist agricultural occupations to post-Fordist service occupations.

Table 1

Post-industrial changes in the employment structure: occupations, sectors, and total amount of employment in Spain between 1977 and 2005

	1977	1994	2005
Occupations
Agrarian occupations	20.3	8.7	5.0
Self-employed in agriculture	12.2	5.9	2.6
Agricultural workers	8.1	2.8	2.3
‘Fordist occupations’	51.5	47.6	49.2
Employers and managers	5.1	6.3	6.7
Non-manual workers	13.8	16.7	17.0
Manual worker	32.5	24.6	25.5
Skilled		11.7	11.4
Unskilled in manufacture		8.7	8.5
Unskilled in construction		4.2	5.6
Urban self-employed	9.3	11.9	7.7
‘Post-industrial occupations’	18.5	31.3	37.9
Professionals and technicians	6.2	14.7	19.4
Skilled services	3.0	4.5	4.3
Unskilled services	9.2	12.1	14.2
Missing cases	0.4	0.5	0.2
Total	100.0	100.0	100.0
Sectors
Traditional economy	76.4	59.5	54.1
Primary	21.6	9.6	5.7
Manufacture	26.9	20.4	17.4
Construction	10.2	9.5	12.6
Distribution/Sales	17.7	20.0	18.5
Service economy	20.9	34.4	40.0
Producer services	3.3	6.7	9.1
Cleaning in producer services	0.2	1.2	1.9
Personal services	10.5	13.2	14.6
Social services	6.9	13.2	14.4
Public Administration	2.7	6.1	5.9
Total	100.0	100.0	100.0
Employment
% not employed (aged 25–60 years)	40.9	43.8	29.5
Total employment	12,477,863	12,309,718	19,314,344

Source: Spanish Labour Force Surveys; authors’ calculations.

Second, the proportion of not employed people has markedly declined since 1994. One should note that this spectacular employment growth of about 7 million jobs was to a large extent unforeseen by scholars studying post-industrial society. As we have already mentioned, the employment predictions in the mid-1990s for Continental and Southern Europe were, actually, rather gloomy (Esping-Andersen, 1996a,b). Labour shedding via early retirement, strong labour market rigidities and familism in social policy were pointed out as factors conspiring to inhibit sustained post-industrial employment growth. The Spanish deviation from this prediction, thus, stands out as an interesting case for macro socio-economic explanations. Although in this paper we are interested in the consequences of the growth in employment more than in its causes, we can mention that the loosening of employment rigidities for new entrants into the labour market since the mid-1980s cannot by itself account for the job growth of the last decade. The share of temporary contracts with respect to total employment is still comparatively high, but it has actually declined slightly in the last years. It seems clear, on the other hand, that women's steady incorporation into the labour market and strong immigration inflows have fuelled the engine of employment growth. But putting together supply and demand factors to provide a full explanation of the Spanish employment record falls beyond the purpose of this paper.

The third interesting result of Table 1 is that the employment growth between 1994 and 2005 has come about with an expansion both at the top and at the bottom of the occupational structure. The expansion has been larger among professionals and technicians (+ 4.7), but at the same time a notable growth has also occurred among unskilled service workers (+2.1) and unskilled construction workers (+1.4). It is interesting to note that Wright and Dwyer (2003) found a similar pattern of employment expansion in the United States in the 1990s. They describe it as ‘asymmetrical polarization’: very strong growth in the top tier of the employment structure, moderately strong at the bottom and extremely weak in the middle. If one further compares the contemporary post-industrial hierarchy in Spain with that of the countries studied in Esping-Andersen (1993; 1999), Spain's profile is rather different from the German profile and surprisingly similar to the American one at the end of the 1980s. In Spain in 2005, professionals and technicians accounted for 19.4 per cent of employed people and unskilled service workers for 14.2 per cent. At the end of the 1980s, in the USA professionals and technicians accounted for 18.1 per cent and unskilled service workers for 11.7 per cent, while in Germany in 1985 unskilled service workers were only 4.5 per cent of employed people. A more updated comparison with data from the European Social Survey (ESS) confirms that Spain has one of the highest proportions (14 per cent according to the ESS in 2002/03) of unskilled service workers in Europe, and only Norway and Sweden (and Portugal) rank higher (Leiulfsrud et al., 2005).

Moreover, if one compares the distribution of sectorial employment, Spain also differs from Scandinavian countries. The key dissimilarity is found in the incidence of employment in social service, about 14 per cent (and notably stable between 1994 and 2005) in Spain, while it was about 25 per cent in Norway and Sweden in the mid-1980s (Esping-Andersen, 1993: 38). One might be tempted to conclude that, with a 20-year time lag, Spain is following the American post-industrial trajectory more closely than the Continental European or Scandinavian ones. There are, however, also notable differences due to the incidence of self-employment, the non-negligible proportion of employment in agriculture and the high incidence of construction that add a southern European ‘touch’ to the Spanish occupational structure. In any case, the expansion of unskilled service workers, which currently accounts for more than 14 per cent of employed persons, raises the question of whether they have the potential to emerge as a new social class. We turn to this issue in the next section.

A New Service Proletariat?

Table 2 presents the composition of the occupational structure by gender and country of origin in 2005. Unskilled service workers turn out to be rather heterogeneous. One in four is an immigrant and one in two is a Spanish women. Immigrants also account for about one-fourth of the unskilled workers in construction and agriculture. While immigrant men are over-represented in construction and agriculture, immigrant women are over-represented in the unskilled services. On the other hand, Spanish men are disproportionately located in ‘Fordist occupations’, with the exception of non-manual occupations. This first aggregate picture of the occupational structure shows that gender and ethnicity differences are strongest among unskilled workers and in particular among unskilled service workers. Subjects that share the same disadvantaged position are, therefore, likely to differ with regard to socialization, previous life experiences and expectations for the future. This internal differentiation is the first factor that plays against the possibility of class formation. But whether unskilled service occupations might become the site for the emergence of a new social class also depends on their incumbents chances for the mobility towards better positions. In other words, the question is whether unskilled service occupations are stop-gap jobs or long-term dead-end road where persons are trapped all throughout their employment career. In this last case, persistence in a common disadvantaged position and exposure to the same living conditions might favour class emergence.

Table 2

Class composition in 2005 by gender and country of origin (row %)

	Spanish & Eu		Others
	Men	Women	Men	Women	Total
Agrarian occupations
Self-employed in agriculture	72.0	27.3	0.4	0.2	100.0
Agricultural workers	53.8	19.1	22.0	5.1	100.0
Fordist occupations
Employers and managers	71.7	25.2	2.0	1.1	100.0
Non-manual workers	33.1	63.4	1.3	2.2	100.0
Skilled manual workers	82.2	8.3	8.7	0.8	100.0
Unskilled manufacture	69.7	20.7	7.6	2.0	100.0
Unskilled construction	73.1	0.4	26.4	0.1	100.0
Urban self-employed	62.5	32.6	3.3	1.6	100.0
Post-industrial occupations
Professionals and technicians	53.9	43.6	1.6	0.8	100.0
Skilled services	77.8	18.4	3.0	0.7	100.0
Unskilled services	22.5	54.3	4.4	18.8	100.0
Total	54.9	35.9	5.3	3.8	100.0

Source: Spanish Labour Force Surveys; authors’ calculations.

In the next table, we compare the exit rates out of unskilled occupations for the years 1994–1995 and 2003–2004 (Table 3). The comparison of these two snapshots of mobility rates should not be overtly biased by period effects since in both time intervals overall employment has been growing: the first period (1994–1995) corresponds to the beginning of the growth of employment in the past 11 years, while the second one (2003–2004) corresponds to a phase of ongoing growth of employment. Six possible destinations are considered: an upward move, a lateral move, stability, unemployment, housework, and a move to inactivity. The first conclusion from Table 3 is that the prospects for jumping the skill barriers are slightly smaller for unskilled service workers (5–6 per cent) than for other unskilled workers (7–9 per cent). Moreover, the yearly upward exit rates have not changed much between 1994 and 1995 and 10 years later. Thus, an increase in their chances of upward mobility has not come about with the expansion of unskilled service workers. On the other hand, the risk of a transition to unemployment was higher on average in the mid-1990s than between 2003 and 2004. This is particularly true in the case of unskilled workers in the construction sector: between 1994 and 1995, 16.4 per cent of them moved to unemployment, while only 6.2 did so between 2003 and 2004.

Table 3

Yearly exit rates from unskilled occupations by type of move and type of occupations between 1994 and 1995 and between 2003 and 2004

		Destination in 2004
Origin in 2003	Upward	Lateral	Stable	Unempl.	House-work	Other inactivity	Total
Agricultural workers	8.1	7.8	56.4	16.4	3.8	7.5	100.0
Unskilled manufacture	8.9	2.7	78.0	4.9	2.0	3.6	100.0
Unskilled construction	7.1	4.0	78.9	6.2	0.9	2.9	100.0
Unskilled service workers	5.2	2.0	80.0	5.5	4.4	2.9	100.0

Source: Spanish Labour Force Surveys; authors’ calculations.

If one deepens the comparison among the unskilled occupations, unskilled workers in construction and agriculture show the highest yearly rates of transition to unemployment. Unskilled agricultural workers also have the highest yearly rate of transition out of the labour market. This makes for an extremely low level of stability: from one year to the next, about 40 per cent of those employed in unskilled occupations in agriculture have moved to a different state.

Still, these contrasts over time and among unskilled occupations leave the question of whether a yearly 5.2 per cent upward rate from unskilled service occupations is a large or small outflow in substantive terms. If one considers a period of 10 years, a 5.2 per cent yearly rate would allow about 38 per cent of the unskilled workers to move up in the occupational structure. But the other 62 per cent would remain immobile or moving between unskilled occupations, unemployment, and non-employment.

Do these figures make the case for the social closure or for the fluidity of unskilled service occupations in Spain? Giving a straightforward answer to this question is complicated. The only directly comparable data available suggest that the likelihood of an upward move from the Spanish unskilled service occupations is lower than in the United States and Denmark. The annual upward rate in the United States between 1987 and 1988 was 7.8 (Jacob, 1993: 214), while it was 7.9 in Denmark between 1986 and 1987 (Esping-Andersen et al., 1994: 133). This result is in line with Carabaña's (1996) findings which show a relatively very high level of social closure in the unskilled service occupations in Spain. Carabaña (1996) compared intragenerational mobility rates in Spain with those presented by Blossfeld et al. (1993) for Germany. One should note that Germany is the country with the lowest mobility out of the unskilled service occupations among those studied in Esping-Andersen (1993). The result of the comparison is that the stability in the unskilled service occupations was much higher in Spain than in Germany at the end of the 1980s.

In order to overcome the limitation imposed by the use of yearly exit rates, we have used the method of synthetic cohorts as described in Section 3. Using cross-sectional data we have, thus, followed the employment patterns of the Spanish-born cohorts who have entered into the labour market between 1994 and 2000, from the year 2000 up to the year 2006. This research design allows us to study the changes in the occupational distribution of the same entry cohort in the labour market over a 6-year time interval. We use a collapsed class scheme which distinguishes between the unskilled service occupations, other unskilled occupations (including unskilled occupations in agriculture, manufacture and construction), all other skilled occupations (including self-employment), and those who are not employed. The results of this analysis are shown in Table 4. The main finding for the purpose of this paper is that the proportion of unskilled service occupations did not sharply decline in the period under observation. For instance, among the 2000 entry cohort, 9.5 per cent were employed in unskilled service occupations in the year 2000. Six years after still 8.2 per cent were employed in the same type of unskilled occupations. With the caveat that the immigrants are not included in this analysis and that one can not exclude the possibility of an (unlikely) carousel effect, these results also point to a substantial stability in the unskilled service occupations.

Table 4

Changes in the occupational distribution between 2000 and 2006 by Spanish-born synthetic cohort of entry into the labour market

	Skilled occupations		Unskilled service occupations		Other unskilled occupations		Not employed
Cohort of entry	2000	2006	2000	2006	2000	2006	2000	2006
1994	58.8	63.7	11.9	9.1	15.6	12.3	13.7	15.0
1995	57.8	65.7	10.9	8.4	15.7	12.5	15.6	13.5
1996	57.9	66.3	10.3	7.8	14.7	11.8	17.2	14.1
1997	57.6	67.4	10.1	8.1	13.3	11.4	19.1	13.1
1998	56.1	69.0	10.2	8.3	11.1	9.7	22.6	13.0
1999	50.9	70.4	10.3	8.4	8.9	8.4	30.0	12.8
2000	45.3	70.7	9.5	8.2	6.6	7.9	38.6	13.2

Source: Spanish Labour Force Surveys; authors’ calculations.

If one goes back to Schumpeter's metaphor of social classes as an omnibus, it is tempting to conclude that the exit door from the unskilled service occupations towards a better class seems to be rather narrow. If this is the case, which ‘passengers of the unskilled service omnibus’ make it through the narrow door that leads to better occupations? Table 5 presents exit rates from unskilled service occupation between 2003 and 2004 and provides a first answer to this question.

Table 5

Exit from unskilled service occupations between 2003 and 2004 by type of move and sex, education, country of origin, age, duration in the job and sector

	Upward	Lateral	Stable	Unemployed	House-work	Other inactivity	Total
Gender
Men	8.9	4.1	80.1	3.6	0.5	2.9	100.0
Women	3.6	1.2	80.4	6.3	5.9	2.7	100.0
Education
Primary or less	2.5	1.3	83.2	5.1	5.8	2.0	100.0
Lower secondary	5.6	1.7	79.8	5.1	5.2	2.7	100.0
Vocational	5.5	1.9	81.2	6.2	2.1	3.1	100.0
High secondary	6.5	3.0	78.7	5.1	2.5	4.3	100.0
University	10.9	5.5	68.8	10.0	2.9	2.0	100.0
Country of origin
Spanish and EU	5.1	1.3	80.2	5.8	4.8	2.8	100.0
Others	4.7	5.4	81.0	4.1	2.4	2.5	100.0
Age
16–31	9.2	2.8	70.1	8.6	3.3	5.9	100.0
32–47	4.2	2.3	82.7	5.4	4.5	0.9	100.0
48–64	2.2	0.6	86.9	2.4	5.3	2.7	100.0
Duration in the job
0–12	7.6	4.9	67.3	9.2	6.9	4.2	100.0
13–60	5.3	1.5	80.9	5.7	4.2	2.5	100.0
61–	2.9	0.4	89.5	2.6	2.8	1.9	100.0
Sector
Public	2.8	1.9	86.0	4.2	2.1	3.0	100.0
Private	5.6	2.0	78.8	5.9	5.0	2.7	100.0
Total	6.7	3.1	77.5	6.4	3.0	3.3	100.0

Source: Spanish Labour Force Surveys; authors’ calculations.

Men are two-and-half times more likely than women to climb the skill barrier, while their risk of falling into unemployment is about half as large as women's. People with university degrees are also more likely to exit the unskilled service occupations either with an upward move or with a transition to unemployment. The same is true for young people, who are about four times more likely to move upward and about three times more likely to be unemployed in 2004 when compared to older workers. Duration in the job is another key factor that affects the risk of entrapment in the unskilled service occupations: the longer the duration in the job, the less likely it is that there will be any type of move. Thus, 90 per cent of those employed for more than 5 years in an unskilled service occupation do not move from one year to the other, while the same is true for only 67 per cent of those employed for <1 year. In contrast to the findings in Denmark (Esping-Andersen et al., 1994), public employees in unskilled service occupations have less chance for upward mobility than those employed in the private sector. Finally, the country of origin does not seem to affect the likelihood of an upward move. If anything, immigrants employed in unskilled service occupations in 2003 were slightly less likely to be unemployed in 2004 when compared with Spanish and EU workers.

Table 6 presents the results of a pooled mobility analysis of the types of exit from unskilled occupations for the 1994–1995 period and the 2003–2004 period. These results expand the bivariate results discussed so far using a multivariate analysis. The design of this pooled cross-section multinomial logit allows us to test whether the overall rate of exit changed between the two time periods and to examine whether the effects of independent variables changed over time and among the type of unskilled occupations. We present only the estimates for upward moves and transitions to unemployment. The results are presented in terms of relative risk ratios.

Table 6

Relative risk ratios of upward mobility and transition to unemployment from unskilled occupations; years 1994–95 and 2003–04 (multinomial logistic regression; the reference category is immobility)

	Model 1a		Model 2b		Model 3b		Model 4c
	Upward	Unempl.	Upward	Unempl.	Upward	Unempl.	Upward	Unempl.
Occupational class
Unskilled manufactory (ref.)	1.0	1.0	1.0	1.0
Unskilled services	0.7**	0.9	0.9	0.8	0.9	0.8
Unskilled construction		2.2**	0.9	1.7**	0.8	1.4
Gender
Men (ref.)	1.0	1.0	1.0	1.0
Women	0.6**	1.9**	0.5**	2.3**	0.5**	2.0**	0.4**	1.8**
Education
Primary or less (ref.)	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0
Lower secondary	1.1	1.0	1.1	0.9	1.1	0.9	1.5	0.7
Vocational	1.5*	0.9	1.5	0.8	1.6	0.8	1.8	0.8
High secondary	1.1	0.9	1.2	1.0	1.2	1.0	1.6	0.7
University	2.3**	1.8*	2.5**	1.6	2.6**	1.6	3.9**	1.6
Age
16–31	2.8**	3.6**	2.9**	2.5**	2.3**	1.5	2.8**	3.3**
32–47	1.6**	1.8**	1.3	1.7**	1.2	1.4	1.5	2.1**
48–64 (ref.)	1.0	1.0	1.0	1.0
Sector
Public (ref.)	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0
Private	1.1	1.2	1.2	1.1	1.3	1.1	1.9**	1.2
Period
1994–1995 (ref.)	1.0	1.0
2003–2004	0.9	0.5**
Duration in the job
0–12					1.0	1.0	1.0	1.0
13–60					0.4**	0.4**	0.6*	0.5**
61–					0.5**	0.2**	0.4**	0.3**
Country of origin
Spanish and EU					1.0	1.0	1.0	1.0
Others					0.7	0.7	0.6	0.4*
Number of observations	9469		5323		5323		2614

*P < 0.05, **P < 0.01.

aall sample.

bperiod 2003–2004.

cperiod 2003–2004 and to exit from the unskilled service occupations only.

Notes: The complete results with the relative risks for the other types of moves (lateral move, housekeeping, inactivity) are available upon request.

Source: Spanish Labour Force Surveys; authors’ estimations.

The findings of Model 1 are based on the pooled data for the two periods considered and refer to exit from all unskilled occupations. In general, all the results of this modelling exercise are in line with those of the bivariate analysis. If one considers the effects of the dummy variable for the period 2003–2004, these results indicate that the likelihood of upward mobility from unskilled occupations did not change much (if anything, it has decreased slightly) when compared with the 1994–1995 period, while the risk of falling into unemployment was halved (the risk ratio is 0.5).

If one focuses on unskilled service workers, they are less likely to move upward when compared to unskilled manual workers in industry. On the other hand, their relative risk of falling into unemployment does not differ much (the ratio is almost equal to 1), while it is two times higher for unskilled workers in construction. Overall, these results suggest that there is not a higher grade of social fluidity among unskilled service workers than among other unskilled occupations. In this respect, Spain is more similar to Germany than to the Anglo-Saxon and Scandinavian countries studied in Esping-Andersen (1993). With regard to the other characteristics that affect the exit from the bottom of the occupational structure, the relative risk of an upward move is higher for men, for those with a university degree and for young people, while no difference is found among those employed in the public and the private sectors.

In Model 2 we restrict the analysis to the period 2003–2004. The main conclusion is that there is no remarkable variation in the pattern of results. This means that the mobility regime out of unskilled occupations has not changed much in more recent years. The advantage of men, young people and those with university degrees in climbing the skill barrier has slightly increased. No differences are observed among unskilled occupations.

In Model 3 we deepen the analysis of the factors underlying entrapment in unskilled occupations and consider the country of origin and the previous duration in unskilled occupations. People not born in Spain, in the EU (previous to enlargement) or in other advanced economies are less likely to move out of unskilled occupations, although neither the coefficient for an upward move nor a transition to unemployment is statistically significant. The likelihood of a move also declines abruptly with the duration in the job: the relative risk of an upward move is more than two times higher in the first year than thereafter. This result suggests that either one manages to move up quickly or one runs a higher risk of entrapment at the bottom of the occupational structure. Time dependency is even stronger in the case of the transition to unemployment. In this case, if one considers workers with tenures shorter than 1 year, their risk of unemployment is three time higher than those with a tenure between 1 and 5 years and four times higher than those with a tenure longer than 5 years. These results are in line with the fact that the greater the worker's seniority, the higher the severance pay that employers have to pay in case of unfair dismissal (Toharia and Malo, 2000).

Finally, in Model 4, we focus on unskilled service occupations. This last model allows us to test whether the determinants of mobility differ in the specific case of unskilled service occupations. The results suggest that the correlates of upward mobility and the risk of unemployment are quite similar to the general pattern discussed above. One might, however, note that those characteristics positively associated with upward mobility from unskilled work (Model 3) are even more decisive for mobility out of unskilled service occupations. For instance, the relative chance of escaping from an unskilled service occupation with an upward move is about four times higher (3.9) for people with university degrees than for people with primary education or less. In the general model without distinguishing among unskilled occupations, the same ratio was equal to 2.6 (Model 3). The impact of gender and age also increases. The most remarkable result is the positive (and statistically significant) effect of being employed in the private sector on the relative chance of an upward move. To put it negatively, the risk of entrapment in unskilled service occupations is higher in the public sector than in the private one independent of all the other characteristics of the worker. The opposite result was found by Esping-Andersen et al. (1994) in Denmark, where employment in the public sector offers channels of upward mobility out of unskilled service occupations.

Conclusions

In line with Merton's (1987) recommendation, we have tried to establish some ‘social facts’ of the post-industrial transformation in Spain in this paper. This descriptive and analytical approach has implied a rather long and, in some cases, tortuous journey through tables and comparative figures. At the end of this journey, we can try to answer to the questions put forward in the introduction. Has the impressive job growth in Spain since 1994 implied an increase in unskilled service workers? The answer is yes, but at the same time one should also note that in the last decade there has been a strong growth in professional and technical occupations. Still, unskilled service workers’ share of employment is currently about 14 per cent. Directly comparable figures for other countries suggest that Spain is characterized by one of the largest segments of this type of unskilled occupations (Esping-Andersen, 1999; Leiulfsrud et al., 2005).

The other questions that inspired our research addressed the key issue of whether unskilled service workers have the potential to emerge and coagulate into a new social class. The answer in this case is less straightforward. On one hand, we find that the composition of unskilled service workers is rather heterogenous. While the traditional industrial working class is still largely homogenous with respect to gender and ethnicity (being predominantly male and Spanish born), the unskilled service occupations present the strongest differences. One in four of the people in this employment category is an immigrant and one in two is female. This internal differentiation might be interpreted as a factor that plays against the possibility of class formation. On the other hand, when we examine mobility patterns out of unskilled service occupations, we find that the yearly exit rates towards better positions are relatively low when compared to similar rates for the United States and Denmark. The results of the analysis with synthetic entry cohorts also points to a noticeable degree of social closure within the post-industrial bottom-level jobs. In contrast to the majority of countries studied in Esping-Andersen (1993), in Spain unskilled service jobs are not like transitional stop-gap jobs or spring-boards to move upwards in the occupational structure.

With regard to the characteristics of the ‘movers’, we have found that there are higher chances of escaping from unskilled service jobs for young people, men and those with tertiary education. The likelihood of an upward move is also time-dependent insofar, as it sharply declines after the first year in the job. This means that either one manages to move out very fast or the risk of entrapment increases abruptly. Finally, and contrary to what seems to happen in the Scandinavian countries (Esping-Andersen et al., 1994), the public sector does not promote upward mobility of unskilled service workers, which is more likely in the private sector.

If one brings all the results together in light of the comparative evidence available, one might conclude that the service proletariat in Spain is relatively large (possibly as large as in the United States) and stable (possibly more so than in Germany). Future research should try to explain which factors underline the (unexpected) Spanish post-industrial employment growth and how its trajectory fits in the welfare and production regime framework of comparative analysis (Esping-Andersen, 1993; Soskice, 1999). If we turn to the issue of class formation and consider Schumpeter's omnibus once more, it seems that in Spain one finds a small group of passengers (male, young, with high-education, employed in the private sector) who manage to get off the omnibus very quickly, while others are likely to remain trapped on it. This relatively high level of social closure speaks in favour of an ongoing demographic class formation of unskilled service workers. But the other crucial characteristic of the ‘service proletariat omnibus’ is that the passengers who stay on are rather different among themselves. Nowadays, ethnic and gender cleavages segment unskilled service workers in groups that have very heterogenous life-courses in terms of past experiences and, probably, expectations for the future. In the light of these differences and despite the high level of immobility, we are tempted to believe that the demographic class formation of the unskilled service workers is rather unlikely.

Finally, this paper could be expanded in two directions. First, it would be interesting to address the issue of social class formation in Spain over a longer time span. Second, one could replicate the same analysis for different countries. This would imply a full and timely revitalization of the research agenda on social stratification in post-industrial societies. We hope this paper might be a first contribution to such an enterprise.