Keith Darden and Anna Gryzmala-Busse, “The Great Divide”

4
Social Protection and the
Formation of Skills: A Reinterpretation
of the Welfare State
Margarita Estevez-Abe, Torben Iversen,
and David Soskice
4.1 Introduction
Social protection does not always mean ‘politics against markets.’ In this
chapter we argue, as did Polanyi (1944), that social protection rescues
the market from itself by preventing market failures. More specifically,
we contend that social protection aids the market by helping economic
actors overcome market failures in skill formation. We show, in this
chapter, that different types of social protection are complementary to
different skill equilibria.1
The market failure problem in the provision of skills is generally attributed to employers’ reluctance to invest in transferable skills. Here we
pay more attention to workers’ reluctance to invest in specific skills.
Young people are less likely to invest in specific skills if the risk of loss
of employment opportunities that require those specific skills is high.
Employers who rely on specific skills to compete effectively in international markets therefore need to institutionalize some sort of guarantee
to insure workers against potential risks. Without implicit agreements for
long-term employment and real wage stability, their specific skills will
be under-supplied. Employers’ promises are not, however, sufficiently
credible by themselves. This is why social protection as governmental
policy becomes critical.
1111
211111
3456789
10111
123456789
20111
123456789
30111
123456789
40
1
Original version prepared for presentation at the 95th American Political Association
Meeting at the Atlanta Hilton and the Marriott Marquis, 2–5 Sept. 1999. We wish to thank
Robert Fannion for his excellent research assistance. Thanks also to Paul Pierson and the
participants at the European Political Economy Workshop at the Center for European
Studies, 5–6 Nov. 1999 for many helpful comments. Torben Iversen gratefully acknowledges financial support from the Hoover Institution while working on this paper.
1 For a discussion of complementarity, see the Introduction to this volume.
The upshot of this chapter is that the shape of social protection has
bearings on national competitive advantage in international markets and
choice of product market strategies. Relative abundance in certain skills
in a given country constitutes a comparative advantage for firms in that
country. But relative abundance in, for instance, a highly flexible multiskilled workforce does not simply come by as an accumulative result
of individual firms’ decisions. Firms’ product market choices are constrained by the availability of necessary skills. Availability of specific
skills, in turn, requires appropriate forms and levels of social protection.2
Institutional differences that safeguard returns on specific skills explain
why workers and employers invest more in specific skills. The absence
of such institutions, in countries such as the USA and UK, gives workers
a strong incentive to invest in transferable skills. In such an environment,
it then also makes more economic sense for firms to pursue product
market strategies that use these transferable skills intensely.
We refer to the set of product market strategies, employee skill trajectories, and social, economic, and political institutions that support them,
as welfare production regimes. A primary objective of this chapter is therefore to identify the main varieties of welfare production regimes and
their consequences for distribution and economic outcomes. In the
rapidly growing ‘varieties of capitalism’ literature, production regimes
are conceptualized as institutional complementarities that reinforce one
another and particular ways of producing and competing in international
markets (cf. Hollingsworth and Boyer 1997; Kitschelt et al. 1999b;
Hall and Soskice, this volume). We contend that the welfare state can
also be understood as a complement in national production systems. Our
chapter thus is part of the new efforts to understand the link between
models of capitalism and welfare states (Swenson 1997, forthcoming;
Mares 1998; Scharpf and Schmidt 2000; Estevez-Abe 1999a; Huber and
Stephens 2001).
The model of micro-level links between skills and social protection we
develop in this chapter has important policy implications. First, our
model predicts what types of political alliance are likely to emerge in support of a particular type of social protection. For example, in economies
146 Estevez-Abe, Iversen and Soskice
2 We are not arguing that social protection is the sole institution that makes a particular
skill formation possible. Other institutions are also necessary: for instance, in the case of
industry skills, strong employer associations are needed to develop agreed vocational
training standards; and in order for the social protection institutions to function effectively
strong employee representative organizations are required. Furthermore, different product
market strategies require access to different kinds of finance: the longer term the commitments companies have to make, the longer term the finance has to be. Thus a range of
complementary institutions need to be in place.
where companies engage in product market strategies that require a combination of firm- and industry-specific skills, and where a large number
of workers invest in such skills, a strong alliance between skilled workers and their employers in favor of social protection advantageous to
them is likely to emerge—even if this means reducing job opportunities
for low-skilled workers. By contrast, where business has no common
interest in the promotion of specific skills, it will have no interest in
defending any of the three components of social protection (cf. Mares, this
volume). Second, we show that different systems of social protection have
deeper ramifications for inequality than commonly assumed. Some skill
equilibria—sustained by different systems of social protection—produce
more inequalities based on the academic background of workers, while
others produce more inequalities based on gender.
That said, it should be emphasized here that our model is not intended
to explain the origin of specific social policies. Although our model helps
understand intra-class differences in policy preferences, we do not claim
to make a historical argument about origins.3 The path-breaking work
by Mares (1998) and Swenson (1997, forthcoming) shows the potential
insights that can be gained from studying questions about origins from
a political economy perspective of the sort we advocate here. But this is
not a study of origins.
The chapter is divided into four sections. Section 4.2 outlines the basic
argument. Section 4.3 ties the welfare–skill links to broader issues of
income inequality across different groups of citizens. Section 4.4 provides
empirical support for our argument. The final section concludes.
4.2 Product Market Strategies, Skill Types,
and the Welfare State
We explore the logical links between product market strategies and their
welfare implications in two steps. First, we identify three types of skills
and argue that different product market strategies are facilitated by a
workforce with particular skill profiles. Second, we then spell out how
different types of social protection influence the propensity of individuals to invest in particular skills, which, in turn, determines the skill
profile of an economy.
Social Protection and Skill Formation 147
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
3 There is a significant amount of work that evaluates the role of business in shaping
welfare policies, although authors disagrees in terms of the actual business influence of
economic interests (cf. Skocpol and Ikenberry 1983; Skocpol and Amenta 1985; Quadagno
1984; Martin 1995; Pierson 1995b, 1996a; Hacker and Pierson 2000; Mares 1998; EstevezAbe 1999b).
4.2.1. Skills and Product Market Strategies
This chapter distinguishes three types of skills associated with different
product market strategies: (i) firm-specific skills; (ii) industry-specific
skills; and (iii) general skills.4 These skills differ significantly in terms of
their asset specificity (i.e. portability). Firm-specific skills are acquired
though on-the-job training, and are least portable. They are valuable to
the employer who carried out the training but not to other employers.
Industry-specific skills are acquired through apprenticeship and vocational schools. These skills, especially when authoritatively certified, are
recognized by any employer within a specific trade. General skills, recognized by all employers, carry a value that is independent of the type of
firm or industry. Of course, any actual production system will involve
all three types of skills to some degree. Nonetheless, we can characterize
distinctive product market strategies based upon the ‘skill profile’ they
require.
A Fordist mass production of standardized goods does not require
highly trained workforce. Production work is broken into a narrow range
of standardized tasks that only require semi-skilled workers. Traditional
US manufacturing industries such as automobile and other consumer
durables fall into this category. There is, however, a variant of mass
production called diversified mass production (DMP). The DMP strategy,
in contrast, aims at producing a varied range of products in large
volumes. Japanese auto-makers and domestic electronic appliances industry are good examples. This production strategy depends on workers
capable of performing a wide range of tasks to enable frequent product
changes in the line (Koike 1981). Workers are also expected to solve problems that emerge in the production line themselves to minimize downtime (Shibata 1999). The tasks these workers perform involve high levels
of knowledge about their company products and machineries in use, and
hence are highly firm-specific.
There are product market strategies that do not mass produce. One
strategy is a high-quality product niche market strategy. It requires a
highly trained workforce with industry-specific craft skills. The prototype of this production strategy does not involve any scale merit, and
148 Estevez-Abe, Iversen and Soskice
4 Our framework builds upon Gary Becker’s distinction between general and specific
skills (1964: ch. 3). In Becker’s definition, firm-specific training increases productivity only
in the firm where training takes place. General training, in contrast, raises productivity
equally in all firms. In an analogous manner, industry-specific training can be defined as
training that raises productivity in all firms in the industry, but not in other industries.
Firm-, industry-, and general skills are skills acquired through firm-specific, industryspecific, and general training.
the process tends to involve highly craft-intensive workshops. Custommade clothing, jewelry, and fine porcelain may be examples of such
production. Another strategy is a hybrid. It pursues high-quality product
lines, but takes the production out of small-scale craft shops in order to
increase the volume of production. Streeck (1992b) calls this diversified
quality production. This production strategy requires firm-specific skills
in addition to high levels of craft skills. Germany is a prototype of this
type of production.
All the above strategies require firm-specific and industry-specific
skills to varying degrees. It is important, however, to note that relative
abundance of high levels of general skills (i.e. university and postgraduate qualifications) brings comparative advantages in radical product
innovation. Let us take the example of the USA to illustrate this point. For
example, start-up software companies in the USA take advantage of a
highly flexible labor market with university-educated people combining
excellent general skills with valuable knowledge about the industry
acquired from switching from one job to another. Another example would
be American financial institutions, which have taken advantage of an
abundant supply of math Ph.D.s to develop new products such as derivatives. Complex systems development (for e-commerce, for example),
biotechnology, segments of the telecommunications industry, and
advanced consulting services are other examples that fall into this class
of industries.
4.2.2 The Welfare–Skill Formation Nexus
We make the three following assumptions about workers’ economic
behavior:
(i) People calculate overall return to their educational/training investment
before deciding to commit themselves. (The investment cost of further training
and education can be conceptualized in terms of wages forgone during the period
of training and education, in addition to any tuition or training fees incurred.)
(ii) People choose to invest in those skills that generate higher expected
returns, provided that the riskiness of the investments is identical.
(iii) Ceteris paribus, people refrain from investing in skills that have more
uncertain future returns (i.e. people are risk averse).
From these assumptions, it follows that a rational worker must
consider three factors in making skill investment decisions: (i) the initial
Social Protection and Skill Formation 149
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
cost of acquiring the skills as, for instance, when a worker receives
a reduced wage during the period of training; (ii) the future wage
premium of specific skills; and (iii) the risks of losing the current job and
the associated wage premium.
The core skills required by an industry are critical for this analysis
because they vary in the degree to which they expose workers to the risk
of future income losses. Highly portable skills are less risky than highly
specific skills because in the former case the market value of the skill is
not tied to a particular firm or industry. Faced with future job insecurity,
a rational worker will not invest his or her time and money in skills that
have no remunerative value outside the firm or industry. In other words,
in the absence of institutional interventions into workers’ payoff structure, general rather than asset-specific skill acquisition represents the
utility-maximizing strategy.
Let us now examine what types of institutions are necessary in order
to protect investments in asset-specific skills. We can distinguish three
different types of protection, which might be called employment protection, unemployment protection, and wage protection. Employment protection
refers to institutionalized employment security. The higher the employment protection, the less likely that a worker will be laid off even during
economic downturns. Unemployment protection means protection from
income reduction due to unemployment, and can thus reduce the uncertainty over the wage level throughout one’s career. Wage protection, finally,
is an institutional mechanism that protects wage levels from market
fluctuations. In this section, we first contrast the significance of employment protection and unemployment protection for firm-specific and industryspecific skills. We will discuss wage protection in a separate section,
because it is generally not considered to be part of the welfare system.
Firm-specific skills are, ex hypothesi, worthless outside that specific
firm, and they therefore require a high level of employment protection in
order to convince workers to invest in such skills (Aoki 1988). Since
workers will only be paid the value of their non-firm-specific skills in
the external market, the greater their investment in specific skills the
greater the discrepancy between current wages and the wages they could
fetch in the external market. In order to invest heavily in firm-specific
skills, workers therefore need assurances that they can remain in the
company for a long enough period to reap the returns on such investments (see Lazear and Freeman 1996; Osterman 1987; Schettkat 1993). If
not, the expenditures of training must be commensurably lower, and/or
the premium on future wages higher. In either case, the cost of training
for the firm goes up, and it will offer less training.
150 Estevez-Abe, Iversen and Soskice
Because rational workers weigh higher expected income later in their
career against the risks of losing their current job, the only way to
encourage workers to carry a substantial part of the costs of firm-specific
training is to increase job security and/or reduce the insecurity of job
loss. Hence we can interpret institutionalized lifetime employment, or
subsidies to keep redundant workers within the firm, as safeguarding
mechanisms for firm-specific skill investment.
It should be noted here that this argument appears to conflict with
Becker’s (1964) famous proposition that companies will pay fully for
company-specific skills, since they need to pay only marginally above
the market wage for the employee’s marketable skills in order to retain
the employee, and can thus appropriate the full return on the investment.
Becker’s argument, however, makes two critical assumptions that are
not generally satisfied. The first is that the acquisition of companyspecific skills does not reduce the maximum present value of the
employee’s marketable skills, either marketable skills which the
employee currently possesses or marketable skills in which the employee
will invest in at some future point. This assumption is reasonable where
relatively minor specific skills are concerned—for example, the understanding of specific office routines. It is implausible where the companyspecific skills constitute a major part of the employee’s skill portfolio.
There are three broad reasons for this.
First, the greater the proportion of company-specific skills, the less
likely is the employee to use preexisting marketable skills; since most
skills are maintained by use, refreshment, and updating, the employee’s
marketable skills will deteriorate. Second, after full-time education, a
large proportion of marketable skills are acquired by most employees at
low cost during their employment by doing different jobs (often within
the same company) and being taught, for example, the different software
packages which each job requires. If a company, however, invests heavily
in an employee’s company-specific skills, the company is likely to want
the employee to focus on using those skills, hence not to move around.
Third, the greater the importance of company-specific skills across a
sector in general, the less valuable will marketable skills be to employees
in that sector. Workers will therefore only work for firm-specific skill
companies if they know the chances of job loss are very low, and this in
turn presupposes high employment protection.
A second critical assumption which Becker makes is that companyspecific skills confer no ‘hold-up’ power on employees. If they do, as is
often the case, then the ability of companies to appropriate the full return
Social Protection and Skill Formation 151
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
no longer holds, and then companies will seek credible guarantees from
their workforce. This typically involves cooperative arrangements with
unions, such as works councils, which in turn is made possible because
employees have made investments in company-specific assets. In order
to encourage employees to make these investments, employees require
some measure of credible protection.
For industry-specific skills, employment protection per se matters less.
If skills are truly specific only to the industry, not the firm, workers can
in principle move between firms without loss of income. Instead, what
becomes important for workers’ incentives to invest in industry-specific
skills is the protection of ‘skilled wages,’ regardless of employment
status. Unemployment protection achieves this in part by securing
earning-related benefits and also by helping to keep the skilled wages
high even when the supply of skills exceeds the demand for those skills.
In part, generous unemployment protection is also important in so far
as it allows workers to turn down job offers outside their previous
industry or occupation. If compelled to accept a job offer outside the
worker’s core competencies, either because of low benefits or a strict
requirement to accept almost any job offer, this undermines the worker’s
incentives to invest in industry-specific skills.
A high replacement ratio, especially when the unemployment benefits
are earnings-related, rewards the worker for his or her specific skill
investment even when the worker is out of work. A high replacement
ratio also eliminates the downward pressure on specific skilled wages,
as unemployed skilled workers do not have to take job offers at
discounted wages. Benefit duration and the administration of requirements to accept a ‘suitable job’ further reinforce this mechanism. A longer
benefit duration permits the unemployed industry-specific skill-holders
enough time to find another job that matches their skills, especially if
they are permitted to turn down jobs that are outside their core competencies. This ensures that their reemployment will generate the same
skilled wages as before, simultaneously reducing downward pressures
on the skilled wages. In short, these two components of unemployment
protection—a high replacement ratio and ‘secure’ benefits—guarantee
return on skill investment sufficient to compensate for economic fluctuations.
For firms pursuing product market strategies which depend heavily
on firm- and industry-specific skills, promise of employment and unemployment security can thus provide a cost-effective path to improving
the firms’ competitive position in international markets (cf. Ohashi and
Tachibanaki 1998; Koike 1994). Contrary to conventional neoclassical
152 Estevez-Abe, Iversen and Soskice
theory, which sees efforts to increase protection against job loss as an
interference with the efficient operation of labor markets, measures to
reduce future uncertainty over employment status M hence uncertainty
over future wage premiums M can significantly improve firms’ cost
effectiveness (Schettkat 1993). And the more successful these firms are,
the greater their demand for specific skills. We are in a specific skills
equilibrium.
If there is little protection built into either the employment or the
unemployment system, the best insurance against labor market risks for
the worker is to invest in general, or portable, skills that are highly
valued in the external labor market. If general skills are what firms need
for pursuing their product market strategies successfully, low employment protection can thus give these firms a competitive edge. Indeed, if
most firms are pursuing general skills strategies, then higher protection
will undermine workers’ incentives to invest in these skills, without
significantly increasing their appropriation of specific skills (because
there is little demand for such skills). In this general skills equilibrium
the neoclassical efficiency argument for little protection is more valid.5
The predictions of the argument are summarized in Fig. 4.1, which
identifies the four main welfare production regimes and gives an empirical example of each (discussed below). The empirical details will be
discussed in section 4.4.
4.2.3 Wage-Bargaining Institutions and Wage Protection
The previous section argued that the greater the uncertainty attached to
employment in the use of particular skills the greater the incentive to
invest in general skills which permit mobility. The proper calculation
relates to the uncertainty of income, namely employment multiplied by
earnings. If there is uncertainty in the earnings from particular skills, this
is as much of a disincentive to sink investments in those skills as uncertainty in employment.
We suggest that some wage determination systems provide what we
call wage protection. Wage protection reinforces the effects of employment
and unemployment protection by reducing the risk that the wage levels for
specific skills might drop radically in the future. In other words, wage
Social Protection and Skill Formation 153
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
5 Since the general skills are portable, there is no risk associated with separation from
current employer. See Gary Becker (1964). This does not mean that high turnover in countries with more general skills does not produce negative welfare consequences from the
economy-wide efficiency perspective. For an interesting elaboration on this issue, see
Chang and Wang (1995).
protection makes it possible for workers to calculate viable lifelong earnings if they were to invest in specific skills. This possibility is arguably
as important as employment and unemployment protection in making
sunk investments in human capital. In the absence of a credible lifetime
earnings, any major investment in specific skills is unlikely to result.
Thus the wage determination system should be seen by political economists not just in macroeconomic terms but as a critical linking institution between the welfare state and the production regime. Just as the
nature of employment and unemployment protection is contingent on
the type of skill investments, so too is wage protection.
At a rough first cut, then, it is necessary to provide wage protection
in order to induce front-end investments in company- or industryspecific skills. The natural interpretation of a wage protection system is
where wage determination provides a broadly stable proportionate
distribution of earnings across different occupations. This is the result
when there is a high degree of coordination in a system of wage determination. Thus institutionally we would expect to find coordinated
wage-bargaining systems in economies in which specific skills are important, and non-coordinated systems where they are not. And in terms of
outcomes we would expect to find stable distributions of earnings across
occupations in the first, but not necessarily the second case.
There is one complication to the above which is worth noting. A
distinction can be made between wage protection for the employed and
for the unemployed: for the employed wage protection means simply
that wages do not fall out of line with wages in other occupations. For
the unemployed, a natural definition of wage protection is that there is
some guarantee that the wage at which the unemployed person is rehired
is the same as wage at which he or she was previously employed. (The
154 Estevez-Abe, Iversen and Soskice
Low
High
Low
Unemployment
protection

High
Employment protection
Fig. 4.1 Social protection and predicted skill profiles
unemployment benefit, it will be remembered, is part of unemployment
protection.)
As far as the three cases of industry-specific, industry- and companyspecific, and general skills are concerned, employed wage protection and
unemployed wage protection are the same. In the first two cases, they
are both high; in the third case they are both low. But in the case of
company-specific skills employed wage protection is high but unemployed wage protection is low. This mirrors the (very) high employment
and low unemployment protection in the Japanese case. And it can be
explained in the same way: if an employee has a ‘lifetime’ employment
and employed wage protection guarantee, at least so long as he works
effectively, this constitutes an effective incentive for the potentially
responsible employee to invest appropriately in company-specific skills.
Low unemployment and unemployed wage protection serve as a disincentive to leave the company or to work in such a way as might lead
to dismissal.
The association between skills and wage-bargaining has implications
for the distribution of earnings. Wage-bargaining systems have consequences for the wage structure for three associated reasons. First, as
implied by our argument, intra-occupational compression of wages
serves as a complement to employment and unemployment protection
because it helps insure against a big drop in income if a worker loses
her job. Secondly, to the extent that collective bargaining systems are
designed to prevent poaching, they limit the ability of individual firms
to pay wages that are significantly above the negotiated rate. The third
reason has to do with the effect of collective bargaining arrangements on
the relative bargaining power of different income groups. Collective
bargaining at the industry or higher levels brings diverse income groups
into a collective decision-making process, and this affords low-income
groups opportunities to influence the distribution of wages that they lack
in more fragmented systems.
4.3 Self-reinforcing Inequalities and Political Preferences
So far the discussion has focused on the efficiency aspects of social
protection. In this section we extend the core argument to unravel two
sets of previously neglected logics by which welfare production regimes
perpetuate inequalities. First, we point out that general skill systems are
more likely to create a ‘poverty trap.’ Second, we cast light upon the
gender inequality consequences of different product market strategies.
Finally, we discuss how these distributive implications of different
Social Protection and Skill Formation 155
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
welfare state regimes are reproduced and perpetuated through distinct
patterns of political support for social protection.
4.3.1 Distribution, Poverty Traps, and Product
Market Strategies
Our argument has far-reaching implications for equality and labor
market stratification, some of which are poorly understood in the existing
welfare state literature. Product market strategies that rely on high levels
of industry-specific and firm-specific skills are likely to create more egalitarian societies than product market strategies based on general skills.
They therefore help us understand large and persistent cross-national
differences in the distribution of wages and incomes. The existing literature can only account for these differences in so far as they are caused
by redistributive state policies. This is far too narrow an approach. We
contend that most inequalities result from particular welfare production
regimes (i.e. combinations of product market strategies, skill profiles, and
the political-institutional framework that supports them).
The basic logic of our argument is straightforward. We argue that
different skill systems and accompanying training systems have important economic implications for those who are academically weak and
strong respectively. For the bottom one third, or so, of the academic
ability distribution, a highly developed vocational training system offers
the best opportunities for students to acquire skills that are valued by
employers. When entry into vocational training is competitive, these
students have an incentive to be as good as they can academically in
order to get into the best training programs with the most promising
career prospects (Soskice 1994b). Therefore, countries with well-developed (and competitive) vocational training systems provide a stable
economic future even to those students who are not academically strong.
General education systems, in contrast, offer these students relatively few
opportunities for improving their labor market value outside of the
school system. As a result, there are fewer incentives for them to work
hard inside the school system.
In firm-specific skill-training systems, employers develop strong stakes
in overseeing the quality of potential employees (i.e. trainees) and developing clear job entry patterns.6 Since employers are committed to make
156 Estevez-Abe, Iversen and Soskice
6 It is worth noting that monitoring the quality of the general education system becomes
important where a lot of human capital investment takes place beyond the general education system, because poor general education increases the cost of training workers in
industry-specific and firm-specific skills.
significant initial human capital investment in new job entrants, they will
be interested in monitoring the quality of the pool of the new school
leavers. As a result, they are likely to establish a working relationship
with various schools for systematic hiring of new school leavers. Since
employers in a firm-specific skill system carry out initial job training,
new school graduates have a chance of building careers as skilled
workers. This gives young schoolgoers a strong incentive to work hard
in school. The ‘from-school-to-work’ transition is likely to be more institutionalized (Dore and Sako 1989).
Similarly, in the case of industry-specific skills where employers are
involved, employers take an interest in ensuring the quality of vocational
training and the certification of skills (Finegold and Soskice 1988). In
these systems, education–work transition is also relatively institutionalized (Ni Cheallaigh 1995; Blossfeld and Mayer 1988).
In general skill regimes, in contrast, the ‘from-school-to-work’ transition is less institutionalized (see Allmendinger 1989). Hiring is more
flexible. Employers hire new job entrants with different educational backgrounds. Promotion and opportunities for further skill training are themselves contingent upon the job performance of the worker. There is not
so much initial human capital investment by employers as there is in
firm-specific skill systems. Because of the absence of a clear vocational
track, systems based on general skills therefore tend to disadvantage
those who are not academically inclined. Regardless of the presence or
absence of vocational schools and apprenticeship programs, for
employers who emphasize general skills a certificate from a vocational
school does not add much value to the worker. Potential workers therefore have to demonstrate their competence in terms of general scholarly
achievement, and getting a tertiary degree becomes an essential component. Because there is a hierarchy of post-secondary schools, if the
student thinks there is a possibility of making it into the tertiary educational system, he or she has a strong incentive to work hard. For those
who are not academically inclined, by contrast, the system produces the
unintended consequence of undermining the incentive to work hard in
school. In the absence of a specialized vocational track, unless a student
believes that he or she can make the cut into college, there is not much
gained by being a good student.
In short, in general skill systems, since the completion of elementary
and secondary school does not qualify them for a vocational certificate
that leads to secure jobs, academically weak students face lower returns
from their educational investment. Since the opportunity for vocational
training M both on the job and off the job M for these students will
Social Protection and Skill Formation 157
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
remain low, it creates an impoverished labor pool. In contrast, at the
top end of the ability distribution, a general education system offers
the largest returns to those with advanced graduate and postgraduate
degrees. These returns tend to be more modest in specific skills systems
because a large number of companies depend more on industry-specific
and firm-specific skills than professional degrees or broad academic qualifications. General skill systems, therefore, reward those students who
are academically talented in terms of labor market entry. Distribution of
academic aptitude thus translates into distribution of skills, and consequently into a very skewed distribution of earnings. As a consequence,
academically weak students in general skill regimes are worse off than
their counterparts elsewhere: they are more likely to be trapped in lowpaid unskilled jobs.
4.3.2 Gender Equality and Skill Types
Compared to men, women face an additional set of issues when making
skill investment choices (see Estevez-Abe 1999b). In addition to the probability of layoff, women have to take into consideration the likelihood of
career interruption due to their role as mothers (see Daly 1994; Rubery
et al. 1996). For a woman to invest in specific skills, she has to be assured
that potential career interruptions will not (i) lead to dismissal; or (ii)
reduce her wage level in the long run. A high probability of dismissal
reduces the incentives to acquire firm-specific skills. A high probability
of reduction in wages after becoming a mother M because of time off
due to childbirth and -rearing M reduces the incentives to invest in either
firm-specific or industry-specific skills.
For women, therefore, employment protection necessarily involves
two factors in addition to the employment and unemployment protection
discussed earlier. These two factors are (i) protection against dismissal,
such as maternity, parental, and family leave policies; and (ii) income
maintenance during leaves and guarantees of reinstatement to the same
job at the same wage level upon return to work.
As for industry-specific skill investments, leave programs and generous income maintenance during the leave function in the same way as
unemployment protection for male skilled workers. A higher wage
replacement ratio thus encourages specific skill investment. Firm-specific
and industry-specific skills again require slightly different institutional
guarantees. While income maintenance during leave is sufficient for
industry-specific skills, firm-specific skill investment by women faces
another issue. In firm-specific skill regimes, reinstatement to the original
158 Estevez-Abe, Iversen and Soskice
job after the leave means that women fall behind their male cohort in
skill formation and promotion. This means that despite generous income
replacement during the leave, time off due to childbirth and -rearing
reduces women’s overall earnings. The very fact that the child-rearing
years for women coincide with the critical early years of employment
compounds the problem. Therefore, for women to invest in firm-specific
skills, affordable childcare is more important than a family leave policy.
In short, compared to men, it takes more institutional support to
encourage women to make specific skill investments. This means that
employers’ incentives differ significantly from the earlier descriptions of
employment and unemployment protection. From the employers’
perspective, it costs more to provide incentives for women to invest in
specific skills than it does for men (Spence 1973). Not only do additional
income maintenance and childcare create a greater financial burden, but
they come with the organizational cost of hiring replacement workers
during regular workers’ maternal and childcare leaves. And not only is
it expensive to hire highly skilled workers as replacement workers, but
it is also very difficult to seek those skills in the external labor market—
especially in the case of firm-specific skills.
Given these additional financial and organizational costs, employers
are unlikely to support family leave or childcare programs except under
two circumstances: (i) when someone other than the employer covers the
program expenses; or (ii) when there is an acute shortage of men willing
to invest in the skills they need.
From a woman’s perspective, this means that it does not pay to invest
in skills for which there is an abundant supply of males. Even if a woman
invests to acquire a specific skill, as far as there is an abundant supply
of male skilled workers, her skill investment will not be protected to the
same degree as men’s. Given this situation, women are more likely than
men to invest in general skills. Furthermore, even women who are
willing to invest in skill training will rationally choose trades and professions where there are few men. Hence a vicious cycle of occupational
segregation of women arises. In countries where there is an established
vocational training system, women’s enrollment choices will reflect
women’s tendency to avoid ‘male jobs.’
In short, product market strategies that rely on firm-specific and
industry-specific skills are more gender segregating than product market
strategies based on general skills. As we argued, general skills provide
more flexibility without penalizing career interruptions, precisely
because they do not require any external guarantee and reinforcement.
We can thus predict that economies with a large presence of companies
Social Protection and Skill Formation 159
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
with specific skill strategies demonstrate high occupational gender segregation, while general skill systems are more gender neutral.
4.3.3 Employers, Core Workers, and Median Voters
The argument presented so far explains the complementarity between
different combinations of welfare programs and product market strategies, and their distributional implications. From this, it follows that
rational employers who pursue distinctive product market strategies can
benefit from welfare programs and policies that favor their production
strategy. Complementary welfare programs and policies reduce
employers’ cost of providing adequate rewards to persuade workers to
invest in the skill required for specific product market strategies. We can
assume, as in the case of workers, that employers are rational and that
they are aware of the incentive structure affecting workers’ skill investment decisions. In other words, a rational employer who is interested in
specific skills will support policies that ensure an adequate return for
workers who make investments in those skills. Given such benefits,
employers are likely to develop preferences for the ‘right’ sets of
programs and policies (Mares 1998, this volume; Swenson 1997; EstevezAbe 1999a). Similarly, rational workers who have made investments in
specific skills will prefer welfare programs and policies that reward and
protect these investments in the future (Iversen and Soskice 2000).
One of the most salient divisions in employers’ preferences over types
of social protection is firm size (Mares 1998). Small firms are more severely
affected by restrictions on their ability to hire and fire because they do
not have the same organizational capacity to adapt to the business cycle
as do large firms. Moreover, small firms with limited R&D capacity typically depend more on industry technologies and skills than do large
firms, which are often in a position to develop proprietary technologies
based on their own R&D effort. Depending on their particular product
market strategy, and hence skill needs, large firms are therefore more
likely to favor high employment protection than small firms, who tend
to view such protection as an unnecessary financial burden and excessive restriction on their manpower flexibility. For small firms a much
more important resource in developing a healthy supply of workers with
the appropriate industry-specific skills is generous and publicly financed
unemployment benefits. This allows small firms to ‘park’ some of their
skilled workers in the unemployment-benefit system during downturns,
without undermining the incentives of workers to invest in relevant
skills.
160 Estevez-Abe, Iversen and Soskice
Particular welfare programs and policies also have advocates among
workers. Workers who have invested in asset-specific skills have vested
interests in retaining institutions that protect the value of these assets.
Such support will tend to be much lower among workers with heavy
investments in general skills since these workers share equally in the
costs of social provision, yet are less likely to benefit from social protection. The presumed negative relationship between income and support
for social protection, as implied by a standard Meltzer–Richard median
voter model (Meltzer and Richards 1981), is thus mediated by the asset
specificity of workers’ skill investments (Iversen and Soskice 2000).
Consequently, the greater the significance of a specific skill system in
the overall national employment structure, the greater the number of
workers who possess the ‘key skills,’ and the more likely it is that the
median voter would be someone with an interest in supporting generous
social protection. Because the interests of these workers would be well
aligned with most employers, there would be a formidable political coalition in favor of retaining and strengthening existing institutions.7
We can find empirical examples of such cases in northern Europe and
Japan. For instance, employers and unions in Germany and the Nordic
countries collaborate in setting industry-specific skilled wages, and
unions intervene to maintain earnings of skilled workers by means of
combining welfare benefits and wages. In so far as industry-specific skills
are core skills in a particular economy, not only are the key private unions
likely to protect favorable institutions in place but the proportion of the
skilled workers in the overall voting population is likely to be significant. Those not possessing the core skills would be shot out of the most
attractive jobs, and in many specific skills countries this has meant a
secondary position for women in the labor market.
In Japan, where a large number of employers rely on firm-specific
skills, employers advocate wage subsidies during economic downturns
in order to avoid layoffs. Japanese unions, composed by protected core
workers from large corporations, support these policies. In Germany,
where large manufacturers rely on a combination of firm-specific and
industry-specific skills, we also observe that large German employers
and unions are interested in minimizing layoffs. Like Japan, Germany
provides wage subsidies from the unemployment insurance in order to
reduce layoffs. The Mittelstand sector of small firms complain about the
Social Protection and Skill Formation 161
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
7 For detailed empirical support for this hypothesis see Iversen and Soskice (2000), who
use comparative public opinion data to show the effects of skill type on social policy preferences among mass publics.
costs of these protections, and advocate deregulation of hiring and firing
rules, but they have not been able to impose their preferences on the
German political economy (Thelen and Kume 1999a).
In contrast to firm-specific and industry-specific skills, general skills
do not require any institutional guarantee. Indeed, employment and
unemployment protection undermines the incentives of workers to
invest in general skills. Employers pursuing product market strategies
based on general skills thus have no incentive to support either the
employment or unemployment protection we discussed above. Because
the median voter tends to have a good general education and understands that expanding benefits and job security can have adverse effects
on competitiveness, such opposition is likely also to carry the day in electoral politics. We might therefore expect that countries where the dominant product market strategy is based on general skills will have meager
employment protection and unemployment benefits. The USA and UK
provide good examples here.
4.4 Comparative Patterns
Our argument implies a tight coupling between employment protection,
unemployment protection, and skill formation. The dominant mode of
firm structure, as well as circumstances in the historical development of
different welfare production regimes, have led some countries to emphasize employment protection over unemployment protection, or vice versa. As
we noted in the theoretical discussion, political opposition to strong
employment protection legislation will be greater in countries with a high
proportion of small firms.
The predictions of our model are summarized in Fig. 4.1 above. When
neither employment nor unemployment protection is high, workers have
a strong incentive to protect themselves against labor market insecurities by investing heavily in highly portable skills. Since workers are reluctant to take on specific skills in this scenario—or at least unlikely to share
much of the cost of training such skills—firms have an incentive to use
technologies that rely least on specific skills. This, in turn, increases
demand for general skills, and availability of general skill jobs makes
general education more attractive for workers, thus creating a selfreinforcing dynamic. In this case we expect skill profiles to be heavily
tilted toward general and broad occupational skills, with a weak or
absent vocational training system.
When employment and unemployment protection is both high, on the
other hand, workers will find it more attractive to invest in firm- and
industry-specific skills. In turn, this makes it more cost-efficient for firms
162 Estevez-Abe, Iversen and Soskice
to engage in production that require large inputs of labor with specific
skills. As firms specialize in this type of production, the job market for
general skills shrinks. Note here that a standard trade argument supports
the idea of self-reinforcing dynamics in both types of systems: institutional comparative advantage makes an intensive use of relatively more
abundant skills an efficient production strategy.
Yet, not all countries necessarily conform to these two ideal types.
Where companies can offer very high levels of job protection and a large
and attractive internal labor market, firm-specific skill formation can
flourish in the absence of strong unemployment protection (represented
by the south-east corner of Fig. 4.1). If career opportunities are extensive
within the firm, and if the firm makes credible commitments to job security, the external labor market will be small and workers will have an
incentive to take advantage of internal career opportunities by investing
in company-specific skills. This, essentially, is the Japanese situation (see
Aoki 1988; Koike 1981). In most other cases, firms neither have the size
nor the resources and institutional capacity to commit credibly to lifetime employment. It is for this reason that we would ordinarily expect
the development of firm-specific skills to be coupled with generous
protection against unemployment.
On the flip side of the Japanese system, we find welfare production
regimes with extensive unemployment protection, but low or only
modest employment protection. Especially in economies dominated by
small firms, with small internal labor markets and little organizational
capacity to adapt to business cycles, employment protection is a costly
and unattractive option for employers. Denmark is an archetypal
example of an economy with a small-firm industrial structure. Yet, generous unemployment protection for skilled workers is still a requisite for
workers to invest in industry-specific skills in these cases, much the same
way as employment protection is a requisite for investment in firmspecific skills. In effect, unemployment protection increases employment
security within the industry, as opposed to security within a particular
firm. At a high level of abstraction, therefore, the industry in a country
with high unemployment and low employment protection becomes functionally equivalent to the firm in a country with low unemployment and
high employment protection.
4.4.1 Measuring Protection
There are no direct measures of job security, such as the risk of nonvoluntary dismissals, that can be used consistently across national cases.8
Social Protection and Skill Formation 163
1111
23456789
10111
123456789
20111
123456789
30111
123456789
4011
1
8 See OECD (1997c: 147) for a discussion of these measurement problems.
However, a series of indirect measures pertaining to legal and quasi-legal
rules governing individual hiring and firing have been developed by
the OECD to gauge the strictness of employment protection legislation
(EPL). The composite EPL index is based on provisions in the legal code
as well as in collective bargaining agreements, hereunder what constitutes just cause for dismissals, required length of advance notice, mandated severance pay, compensation for unfair dismissals, the rights of
employee representatives to be informed about dismissals and other
employment matters, and the rights of workers to challenge dismissals
in the courts. The OECD constructed this index to reflect as accurately
as possible the costs to employers of dismissing workers, and this is
directly relevant to our argument since such costs can be seen as a
measure of employer commitment to retain the workers they hire.
The composite EPL index is constructed for both regular and temporary employment, but our argument is only relevant for the former since
neither employers nor employees have much of an incentive to invest
in firm-specific skills when employment is time-limited. The regular
employment EPL index is calculated for two periods, the late 1980s
and the late 1990s, but since it is nearly perfectly correlated between
the two periods (r = 0.99) we have simply used an average (shown
first column of Table 4.1). The index is based on the regulation of individual contracts and does not incorporate measures for protection
against collective dismissals. In OECD’s latest update of the index (OECD
Employment Outlook 1999) a separate index was created to reflect the
regulation of collective dismissals, which is shown in the second column
of Table 4.1.
Neither of the OECD measures fully takes into account the employment protection that is built into the firm governance structure or into
the workings of the industrial relations system. As the OECD acknowledges, ‘non-legislated employment protection tends to be more difficult
to measure and may therefore be under-weighted’ (OECD Employment
Outlook 1999: 51). Japan illustrates the problem because companies in
Japan offer greater protection against dismissals for their skilled workers
than the EPL index would suggest (see OECD Employment Outlook 1994:
79–80). Indeed, dismissals and layoffs are extremely rare in Japan
compared to other countries (OECD 1997c: table 5.12).9 Instead, large
Japanese firms engage in special workforce loan practices with their
suppliers, called ‘Shukko’, which enable them to retain workers during
164 Estevez-Abe, Iversen and Soskice
9 These data are not fully comparable across countries, but the figure for Japan is of an
order of magnitude smaller than in any other country for which data are available.
Social Protection and Skill Formation 165
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
TABLE 4.1 Employment protection in eighteen OECD countries

Sweden 2.8 4.5 3 0.94
Germany 2.8 3.1 3 0.86
Austria 2.6 3.3 3 0.84
Italy 2.8 4.1 2 0.81
The Netherlands 3.1 2.8 2 0.80
Japan 2.7 1.5 3 0.76
Norway 2.4 2.8 2 0.66
Finland 2.4 2.4 2 0.64
France 2.3 2.1 2 0.61
Belgium 1.5 4.1 2 0.56
Denmark 1.6 3.1 2 0.53
Switzerland 1.2 3.9 2 0.49
Ireland 1.6 2.1 1 0.36
Canada 0.9 3.4 1 0.30
New Zealand 1.7 0.4 1 0.29
Australia 1.0 2.6 1 0.27
United Kingdom 0.8 2.9 1 0.25
United States 0.2 2.9 1 0.14
a Index of the ‘restrictiveness’ of individual hiring and firing rules contained in legislation and
collective agreements (high numbers mean more restrictive regimes). Weight: 5/9.
b Index of the ‘restrictiveness’ of collective dismissal rules contained in legislation and collective
agreements (high numbers mean more restrictive regimes). Weight: 2/9.
c Measure of company-level employment protection based on three criteria: (i) the presence of
employee-elected bodies with a significant role in company manpower decisions; (ii) the existence of strong external unions with some monitoring and sanctioning capacity (especially
through arbitration); and (iii) the systematic use of employee-sharing practices between parent
companies and subsidiaries or across companies. Where at least two of these conditions are met
to a considerable degree, we assigned a score of 3; where all three are largely absent, we assigned
a score of 1. Intermediary cases were assigned a score of 2. The French case has been assigned
a score of 2 even though company-level protection is weak. The reason is that the Inspectorat
du Travail can and does intervene to prevent redundancies, and this is not captured by OECD’s
legal measure of employment protection. Weight: 2/9.
d Weighted average of columns (1)(2) after each indicator has been standardized to vary
between 0 and 1.
Sources: OECD Employment Outlook (1998: 14252, 1999); Income Data Services (1996); Soskice
(1999).
recessions. In other countries, and to some extent also in Japan, firms
have to consult with works councils or other employee representative
bodies before making decisions about layoffs, and often industry unions
are in a strong position to oppose collective layoffs. This is only partly
reflected in the EPL index since it considers only the need for firms to
notify works councils or unions about impending dismissals—not the
power of unions or works councils to prevent or modify the implementation of decisions to dismiss.
We have captured these ‘private’ employment protection arrangements
in column 3 of Table 4.1 by a simple index that measures the strength of
institutions and practices at the firm level that increase the job security
of especially skilled workers in a company. The measure is based on three
criteria: (i) the presence of employee-elected bodies with a significant role
in company manpower decisions; (ii) the existence of strong external
unions with some monitoring and sanctioning capacity (especially
through arbitration); and (iii) the systematic use of employee-sharing
practices between parent companies and subsidiaries or across companies. Where at least two of these conditions are met to a considerable
degree, we assigned a score of 3; where all three are largely absent, we
assigned a score of 1. Intermediary cases were assigned a score of 2. With
the exception of Japan the index of company-based protection is consistent with the rank-ordering implied by the composite index.
The final column combines the OECD and company-based measures
in a composite index that captures both the legal and more informal
aspects of employment protection. The index is a weighted average with
the following weights: 5/9, 2/9, and 2/9. The first two weights are
adopted unchanged from OECD’s own weighing scheme (OECD Employment Outlook 1999: 118), and reflect the fact that collective dismissal rules
tend to build on individual dismissal rules, which are already part of the
EPL index. Since the influence of employee representative bodies over
firm-level manpower decisions is also partly captured by the EPL index,
we assigned the same (low) weight to the company protection indicator.
We feel that the resulting index of employment protection is about as
accurate as possible, and although some would quibble with the assignment of weights the relative numbers are not very sensitive to changes
in these weights.
Looking at the ranking of countries, it is not surprising to find the
Anglo-Saxon countries at the low end, and Japan and many of the continental European countries at the high end of protection. Belgium,
Denmark, and Switzerland are in the lower half of the table, most likely
because these countries have relatively large small-firm sectors, but in
166 Estevez-Abe, Iversen and Soskice
terms of actual numbers the break in the employment protection index
is between this group of countries and the Anglo-Saxon countries.
The measurement of unemployment protection is more straightforward, although there are some non-trivial issues concerning the administration of unemployment-benefit systems. The most obvious indicator,
and the most commonly used, is unemployment replacement rates; i.e.
the portion of a worker’s previous wage that is replaced by unemployment benefits (see column 1 of Table 4.2). We are here considering a
‘typical’ worker, defined as a 40-year-old industrial production worker,
‘averaged’ across several different family types (single, married to working spouse, and married to non-working spouse), and we are looking at
net replacement rates where cross-national differences in tax systems and
non-income subsidies for unemployed have been adjusted for (such as
rent support). Given that taxation of unemployment benefits varies
considerably across countries, gross replacement rates (for which much
more detailed data exist) can be quite misleading.
As in the case of employment protection, the Anglo-Saxon countries
again score at the bottom. But note that the three continental European
countries falling in the lower half of the employment protection index
M Belgium, Denmark, and Switzerland M now figure at or near the top
of the table. On the other hand, two countries M Italy and Japan M have
very low replacement rates compared to their position on the employment protection indicators. The pattern is broadly similar, though not
identical, when we look instead at the actual amount of money the
government spends on unemployment benefits (as a share of GDP),
compared to the number of unemployed people (as a share of the population). As before, the three countries in northern Europe with relatively
low employment protection are among the five countries with the most
generous unemployment-benefit systems.
Table 4.2 also includes a more qualitative measure of the administration of unemployment benefits: the restrictiveness of the definition of a
‘suitable job.’ All national unemployment systems stipulate that in order
to receive benefits a person cannot refuse a ‘suitable’ job, but what constitutes a ‘suitable job’ varies significantly from one system to another. In
principle, such variation is important for our purposes. For example, if
a skilled worker is required to take any available job, regardless of
whether it is commensurable with the worker’s skills, high unemployment benefits are of limited value from the perspective of reducing the
riskiness of specific skills investments. In practice it is difficult to get any
precise comparable figures for this variable. We therefore pieced together
only a very simple three-tiered classification based on a variety of
Social Protection and Skill Formation 167
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
168 Estevez-Abe, Iversen and Soskice
TABLE 4.2 Unemployment protection in eighteen OECD countries

Denmark 60 76 3 0.91
The Netherlands 58 74 3 0.89
Switzerland (40) 94 2 0.86
Belgium 57 99 2 0.82
Austria 43 78 3 0.81
Germany 43 66 3 0.77
Norway 40 40 3 0.64
Sweden 30 52 3 0.63
France 48 44 2 0.54
Finland 45 20 2 0.43
Ireland (38) 59 1 0.37
Japan 10 48 2 0.33
Canada 32 49 2 0.30
New Zealand 31 44 1 0.27
Australia 32 30 1 0.22
Italy 5 18 2 0.18
United Kingdom 23 15 1 0.11
United States 14 26 1 0.10
a Net unemployment replacement rates for a 40-year-old representative worker. Net figures for
Ireland and Switzerland are missing from the source and have instead been estimated by taking
gross replacement rates for these countries as proportions of average gross replacement rates and
then multiplying these proportions by average net replacement rates.
b The share of GDP paid in unemployment benefits as a percentage of the share of unemployed
in the total population. Average for the period 197389.
c Index that measures the restrictiveness of the definition of a ‘suitable job’ in the administration
of benefits to unemployed. (1) Any job qualifies as a suitable job; (2) Skilled unemployed are
given some discretion in rejecting jobs they deem unsuitable to their skills, but choice is restricted
in time and/or to certain job categories; (3) Skilled unemployed exercise wide discretion in
accepting or rejecting jobs on the grounds of the suitability of the job to their skills.
d Average of columns (1)(3) after each indicator has been standardized to vary between 0 and
1.
Sources: Col. 1: Restricted OECD data reported in Esping-Andersen (1999: table 2.2, p. 22); OECD,
Database on Unemployment Benefit Entitlements and Replacement Rates (undated). Col. 2: Huber et
al. (1997); OECD Economic Outlook (various years); OECD, Labour Force Statistics (various years).
Col. 3: OECD (1991: 199231); European Commission, Unemployment in Europe (various years);
and national sources.
national and international sources. Though basically reinforcing the
pattern revealed by the other two indicators, it does affect the rank-order
position of some countries slightly.
As in the case of employment protection, we combined the various
indicators into an index of unemployment protection (see column 4). With
the possible exception of Italy, this index gives a good sense of crossnational differences in the extent of unemployment protection. The number for Italy probably underestimates the extent of protection because
of quasi-public insurance schemes that do not show up in the official
statistics. Thus, about a third of (mainly large companies) covered by
Casa Integrazione have replacement rates between 70 and 80 per cent,
and there are normally good unemployment-benefit schemes for artisans
(i.e. craftsmen) administered at the regional level by associations representing small firms, in cooperation with regional governments.10
4.4.2 Measuring Skill Profiles
Workers’ skills are difficult to measure because they are not directly
observable. However, we can rely on a number of indirect measures. The
first is median enterprise tenure rates—the median number of years
workers have been with their current employer (based on national labor
force surveys). These numbers contain relevant information about the
firm specificity of skills because firms and individuals investing heavily
in such skills become increasingly dependent upon one another for their
future welfare. The greater the investment, the higher the opportunity
costs of severing the relationship, and the lower the incentive for either
party to do so. Indeed, short tenure rates may not only be an indicator
of the absence of firm-specific skills, but a positive measure of presence of
general skills. The reason is that general skills are developed in part by
accumulating job experience from many different firms.
The drawback of using tenure rates to measure firm-specific skills is
that they may also in part reflect the costs of dismissing workers as a
result of employment protection. However, if higher tenure rates were
unrelated to the extent of firm-specific skills, then the association
between employment protection and tenure rates would be weak at best
since most job switching is known to be voluntary. But, in fact, the crossnational association between the two variables is rather high (r = 0.75),
and where data are available tenure rates are strongly negatively related
to quit rates. From this it seems clear that at least part of the effect of
Social Protection and Skill Formation 169
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1 10 We thank Michele Salvati for providing this information.
170 Estevez-Abe, Iversen and Soskice
TABLE 4.3 Skill profiles in eighteen OECD countries
(1) (2) (3) (4) (5)
Median Vocational Vocational Upper- Skill profilee
length of training training secondary/
tenurea shareb systemc university
eductiond
Austria 6.9 22 Dual 71 Firm/industry/
apprenticeship 6 occupational
Germany 10.7 34 Dual 81 Firm/industry/
apprenticeship 13 occupational
Sweden 7.8 36 Vocational 74 Firm/industry/
colleges 13 occupational
Norway (6.5) 37 Vocational 82 Industry/
colleges 16 occupational
Belgium 8.4 53 Mixed 53 Industry/
11 occupational
Japan 8.3 16 Company- n.a. Firm/
based n.a. occupational
Finland 7.8 32 Vocational 67 Industry/
colleges 12 occupational
Italy 8.9 35 Company- 38 Firm/
based 8 occupational
France 7.7 28 Company- 60 Firm/
based 10 occupational
Ireland 5.3 6 Weak 50 Occupational/
11 general
The Netherlands 5.5 43 Mixed 62 Industry/
22 occupational
Switzerland 6.0 23 Dual 80 Industry/
apprenticeship 10 occupational
Denmark 4.4 31 Mixed 66 Industry/
15 occupational
Canada 5.9 5 Weak 76 Occupational/
17 general
Australia 3.4 9 Weak 57 Occupational
15 /general
New Zealand n.a. 7 Weak 60 Occupational/
11 general
United 5.0 11 Weak 76 Occupational/
Kingdom 13 general
United States 4.2 3 Weak 86 Occupational/
26 general
employment protection on tenure rates must go through the effect of the
former on the stock of firm-specific skills. This interpretation is supported by considerable evidence showing tenure rates across industries
within countries to be closely associated with the skill intensity of these
industries (OECD Employment Outlook 1993: 141–5).
Used as a measure of firm-specific skills (column 1 in Table 4.3), tenure
rates suggest that the stock of such skills is low in the Anglo-Saxon countries compared to Japan and most of the continental European countries
where workers stay with their firms for significantly longer periods of
time. The exceptions are Denmark, the Netherlands, and Switzerland,
where firm tenure rates also tend to be quite short.
The pattern of training suggested by tenure rates is reflected in the
character of vocational training systems (columns 2 and 3). Whereas such
systems are weak or absent in the Anglo-Saxon countries and Ireland, in
all the others vocational training is widespread and highly institutionalized. The share of an age cohort that goes through a vocational training
in the former varies between 3 and 11 per cent (counting short-term postsecondary degrees, such as the American junior college system), and
there is little involvement of companies in the training system. In the
remaining countries, the percentage of an age cohort going through a
vocational training is generally between a quarter and one half of an age
cohort. The figure for Japan is only 16 per cent, but much training in this
country goes on in large companies and is not recorded in the data.
The main difference among the countries with strong vocational
training systems is in the emphasis on company- as opposed to industrylevel training. Whereas in Japan, and to a lesser extent in France and
Italy, the emphasis is on company training, the remainder have some
Social Protection and Skill Formation 171
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
Notes for Table 4.3
a The median length of enterprise tenure in years, 1995 (Norwegian figure refers to 1991).
b The share of an age cohort in either secondary or post-secondary (ISCED5) vocational training.
c The character of the vocational training system according to whether most of the training occurs
at the company level (as in Japan), through a dual apprenticeship system (as in Germany),
through vocational colleges (as in Sweden), or through some mixture of the latter two (as in the
Netherlands). Where vocational training is weak, we have not distinguished between the type
of system.
d First entry is the percentage of 2534-year-olds with an upper-secondary education; the second
entry is the percentage of 2534-year-olds with a university degree (1996 figures). Data are not
available for Japan.
e Average of columns (1)(4) after each indicator has been standardized to vary between 0 and 1.
Sources: Col. 1: OECD Employment Outlook (1997: table 5.5). For Norway: OECD Employment
Outlook (1993: table 4.1). Col. 2: UNESCO (1999). Col. 3: Streeck (1992b); Finegold and Soskice
(1988); Soskice (1999). Col. 4: OECD (1999).
combination of on-the-job training and school-based training, with heavy
involvement of employer organizations and unions. Formally, the
systems can be divided into apprenticeship systems of the German type,
vocational school systems of the Swedish type, or mixtures between the
two, but all combine theoretical, industry-specific, and direct workplace
training (column 3). The weight between the three is difficult to gauge,
but Belgium, the Netherlands, and the Scandinavian countries (less so
Sweden) tend to place more emphasis on school-based training (i.e.
provision of non-firm-specific skills) than do Austria or Germany.
An additional indicator of skill profiles is the share of young people
with post-compulsory education (column 4). Unlike tenure rates, there
is no indication on this measure that Anglo-Saxon countries are less skill
intensive than continental European countries (no comparable data are
available for Japan). As in any advanced economy, a high average standard of living in the Anglo-Saxon countries depends on heavy investment in human capital. Indeed, there is some indication that countries
with only a modest stock of firm-specific skills compensate by investing
more heavily in higher education. For example, there is a negative relationship between tenure rates and university degrees. The USA can here
be highlighted as an archetypal case of a country with a weak company
and vocational training system, but a very advanced higher education
system. Indeed, a college education in this country is widely considered
the only effective insurance against an otherwise highly volatile and
uncertain labor market.
The figures for upper-secondary education hide more subtle differences
in the content of this education. In the Anglo-Saxon countries university
education tends to be very general, and even engineering and business
schools provide very broad training that is not linked to particular industries or trades. By contrast, in Japan and most continental European countries, many university degrees are more specialized and there tend to be
close linkages between engineering and trade schools to private industry.
Combined with the other two indicators, this paints a fairly clear picture
of the skill profile in different countries, summarized in column (5) of
Table 4.3. Needless to say, all training systems produce a whole range
of skills, but each system can be roughly characterized according to its
emphasis on firm, industry, occupational, or general skills.
4.4.3 Putting the Pieces Together
Fig. 4.2 plots the eighteen OECD countries on the employment and unemployment protection indexes. Countries are distributed along a primary
172 Estevez-Abe, Iversen and Soskice
axis, corresponding to the south-west–north-east diagonal in Fig. 4.2, with
some countries further divided along a secondary axis, corresponding to
the north-west–south-east diagonal in Fig. 4.2. The main axis separates
countries into two distinct welfare production regimes: one combining
weak employment and unemployment protection with a general skills
profile, represented by the Anglo-Saxon countries and Ireland; and one
combining high protection on at least one of the two social protection
dimensions with firm- and/or industry-specific skills, represented by the
continental European countries and Japan. The secondary axis divides the
latter group into one with greater emphasis on employment protection
and the creation of firm-specific skills, exemplified primarily by Japan
and Italy,11 and one with greater emphasis on unemployment protection
and the production of industry-specific skills, exemplified by Denmark,
the Netherlands, and Switzerland.
Social Protection and Skill Formation 173
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
11 Although the position of Italy is probably exaggerated by the failure to account for
semi-public unemployment insurance arrangements, as noted above.
Employment protection
0 0.2 0.4 0.6
[33]
[7]
5.3
4.8
8.3
7.3
8.6
0.8 1
Unemployment protection
1 Industry/occupational Firm/industry/occupational
Occupational/general Firm/occupational
0.8
0.6
0.4
0.2
0
Primary axis eS
condary axis
New Zealand
Canada
United Kingdom
Australia
Ireland
Finland
United States
France
Norway
Netherlands
Denmark
Switzerland
Japan
Italy
Sweden
Germany
Austria
Belgium
Fig. 4.2 Social protection and skill profiles
Notes: Bolded numbers are mean tenure rates for the cluster of countries circled; bracketed
numbers are the percentage of an age cohort going through a vocational training.
Source: See Tables 4.1, 4.2, and 4.3.
The data on skills presented in Table 4.3 have been summarized in the
form of averages for each cluster of countries (only tenure rates are relevant for the division along the secondary axis). The high protection countries are also those with the most developed vocational training systems,
and tenure rates decline with employment protection. Clearly, the empirical patterns we observe correspond rather closely to our main theoretical thesis, namely that skill formation is closely linked to social
protection.
The coupling of social protection and skill systems helps us understand the product market strategies of companies and the creation of
comparative advantages in the global economy. Thus, where there is a
large pool of workers with advanced and highly portable skills, and
where social protection is low, companies enjoy considerable flexibility
in attracting new workers, laying off old ones, or starting new production lines. This flexibility allows for high responsiveness to new business
opportunities, and facilitates the use of rapid product innovation strategies. In economies with a combination of firm- and industry-specific
skills, such strategies are hampered by the difficulty of quickly adapting
skills to new types of production, and by restrictions in the ability of
firms to hire and fire workers. On the other hand, these welfare-production regimes advantage companies that seek to develop deep competencies within established technologies, and to continuously upgrade and
diversify existing product lines (‘diversified quality production’ in the
terminology of Streeck 1991).
There is considerable case-oriented research to support these propositions (see especially Porter 1990; Soskice 1999; and Hollingsworth and
Boyer 1997), and they can be bolstered by quantitative evidence constructed by Thomas Cusack from US Patent Office data. Broken into
thirty technology classes, Cusack counted the number of references to
scientific articles for patents in each technology class and country, and
then divided this number by the world number of scientific citations per
technology class.12 The idea is that the number of scientific citations,
as opposed to citations to previous patents and non-scientific sources, is
a good proxy for the extent to which national firms are engaged in radical innovation strategies. The results are shown in the first column of
174 Estevez-Abe, Iversen and Soskice
12 The data are coded into references to previous patents and others, where many of the
latter are references to scientific articles. To get a good estimate for the number of scientific articles in the “other” category, the proportion of scientific references to other references was calculated for a random sample (6,000) for each country and technology class.
These factors were then used to correct the overall dataset so as to get a better measure of
scientific citations.
Table 4.4, with countries ranked by the average ratio of scientific citations
for patents secured by national firms. As it turns out, the Anglo-Saxon
countries and Ireland all have ratios that are significantly higher than in
the specific skills countries of continental Europe and Japan. Precisely as
we would expect.
At the low-tech end of product markets, we have to rely on a different
type of data to detect cross-national differences. In column (2) of Table
4.4 we used the proportion of the working-age population employed in
private social and personal services as a proxy. As argued by EspingAndersen (1990: ch. 8) and Iversen and Wren (1998), firms that rely
heavily on low-skilled and low-paid labor for profitability tend to be
concentrated in these industries. Although we only have data for a subset
Social Protection and Skill Formation 175
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
TABLE 4.4 Scientific citation rates and low-wage service employment in
eighteen OECD countries
(1) (2)
Scientific citation Private service
ratioa employmentb
Ireland 1.514 —
United States 1.310 23
New Zealand 1.267 —
Canada 1.032 20
United Kingdom 0.837 16
Australia 0.804 26
Sweden 0.757 14
The Netherlands 0.754 14
Norway 0.690 17
Switzerland 0.639 —
France 0.601 11
Belgium 0.598 13
Germany 0.592 14
Japan 0.586 —
Austria 0.575 —
Finland 0.552 11
Denmark 0.536 11
Italy 0.491 9
a The average number of scientific citations per patent by national firms in each of 30 technology
classes as a proportion of the average number of citations in each class for the entire world.
b The number of people employed in wholesale, retail trade, restaurants and hotels, and in community, social and personal services, 198291 as a percentage of the working-age population.
Source: Col. 1: United States Patent Office Data. Col. 2: OECD (1996b).
of countries, the numbers display a rather clear cross-national pattern.
Producers of standardized and low-productivity services thrive in
general skills countries such as Australia and the United States because
they can hire from a large pool of unskilled workers who are afforded
much job protection and whose wages are held down by low unemployment protection. By contrast, firms trying to compete in this space
in specific skills countries such as Germany and Sweden are inhibited
by higher labor costs and lower flexibility in hiring and firing. These
differences have magnified during the 1980s and 1990s, and Britain is
now closer to the mean for the general skills countries.
In an open international trading system, differences in product market
strategies will tend to be perpetuated, which in turn feed back into organized support for existing social protection regimes. Contrary to the
popular notion of a ‘race to the bottom’ in social policies, differences
across countries persist and are even attenuated through open trade.
Correspondingly, from the 1970s to the 1980s and 1990s, unemployment
benefits remained stable or rose in most continental European countries,
but they were cut in Ireland and all the Anglo-Saxon countries with the
exception of Australia.13 Moreover, whereas labor markets have become
even more deregulated in the latter countries, employment protection
has remained high in the former. Although some countries have seen a
notable relaxation in the protection of temporary employment, there is
no reduction in the level of protection for regular employment (OECD
Employment Outlook 1999). This evidence, and the theoretical explanation
we provide for it, seriously challenge the notion, popular in much of the
economic literature, that social protection is simply inefficient forms of
labor market ‘rigidities.’ Social protection can provide important competitive advantages. By the same token we question the prevalent approach
in the sociological and political science literature, which understands
social protection solely in terms of its redistributive effects.
4.4.4 Implications for Labor Market Stratification
That said, we are not implying that welfare production regimes are irrelevant for distributive outcomes. To the contrary, our argument has
important implications for equality and labor market stratification, and
it helps account for the political divisions over the welfare state. Partly
these effects are direct consequences of particular product market strategies and their associated skill profiles; partly they reflect the effects of
176 Estevez-Abe, Iversen and Soskice
13 Based on gross unemployment replacement rates published in OECD’s Database on
Unemployment Benefit Entitlements and Replacement Rates (undated).
the collective wage-bargaining system that is itself an important component of the wage protection system.
With respect to wage protection, the most important issue is what we
have previously referred to as wage protection for the unemployed. Such
protection implies that workers with similar skills are paid the same
amount across firms and industries, and in practice this is accomplished
through collective wage-bargaining at the industry level or at higher
levels. It is striking, though not surprising, that all countries with a strong
emphasis on industry-specific skills have developed effective wage coordination at the industry level. Conversely, general skills countries, and
countries with a strong emphasis on firm-specific skills (Japan in particular), lack such coordination.
Very extensive evidence has now been accumulated that demonstrates
the importance of the structure of the wage-bargaining system for the
wage structure (see especially Rowthorn 1992; Wallerstein 1999; and
Rueda and Pontusson 2000), but we believe the skill system is equally
important. Fig. 4.3, which uses the incidence of vocational training as the
indicator for skill system, clearly shows the empirical association
between skills and earnings equality, and there is a good reason. Because
specific skills systems generate high demand for workers with good
vocational training, young people who are not academically inclined
have career opportunities that are largely missing in general skills
systems. Whereas a large proportion of early school leavers in the former
acquire valuable skills through the vocational training system, in the
latter most early school leavers end up as low-paid unskilled workers
for most or all of their working lives.
In combination, the wage-bargaining system M i.e. whether it is industry coordinated or not M and the skill system—i.e. whether it is specific
skills or general skills biased—provides a powerful explanation of earnings inequality as we have illustrated in Fig. 4.4. The figure shows figures
for earnings and income inequality for each combination of bargaining
and skill system. The big drop in earnings equality occurs as we move
from specific skills systems with industry-coordinated bargaining to general skills systems where industry-coordinated wage-bargaining is lacking. By themselves this pair of dichotomous variables account for nearly
70 per cent of the cross-national variance in income inequality.14 Yet,
despite their importance for explaining inequality, neither variable is
Social Protection and Skill Formation 177
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
14 The estimated regression equation is :
Income equality = 0.23  0.048 × Specific skills  0.055 × Industry coordination,
where R2 = 0.69
accorded much attention in the established welfare state literature,
notwithstanding the focus on distribution in this literature. In our theoretical framework, on the other hand, they are integral parts of the story,
even though we have focused on micro mechanisms that emphasize the
importance of efficiency.
The hypothesized relationship between product market strategies, skill
composition, and equality points to another, and quite different, source
of evidence: academic test scores. Because specific skills systems create
strong incentives among young schoolgoers to do as well as they can in
school in order to get the best vocational training spots, whereas those
at the bottom of the academic ability distribution in general skills
systems have few such incentives, we should expect the number of early
school leavers who fail internationally standardized tests to be higher in
general skills countries than in specific skills countries.
Although the data are limited in coverage, this is in fact what we
observe (see Fig. 4.5). Whereas the percentage failing the test varies
178 Estevez-Abe, Iversen and Soskice
Incidence of vocational training(b)
0 10 20 30 40 50 60
D1/D9 wage ratios(a)
4.5
4
3.5
3
2.5
0
Austria
France
Switzerland
Japan
UK
Ireland
USA
Canada
New Zealand
Australia
Germany
Netherlands
Belgium
Finland
Italy
Denmark
Norway
Sweden
Fig. 4.3 Vocational training and wage inequality
a The share of an age cohort in either secondary or post-secondary (ISCED5) vocational training.
b The earnings of a worker in the top decile of the earnings distribution relative to a worker in
the bottom decile of the earnings distribution.
Sources: D1/D9 wage ratios: UNESCO (1999). Incidence of vocational training: OECD, Electronic
Data Base on Wage Dispersion (undated).
between 15 and 22 per cent in the Anglo-Saxon countries, it is only
between 8 and 14 in the countries emphasizing more specific skills for
which we have data. Although these differences could be due to the
overall quality of the educational system, it is not that case that the
Anglo-Saxon countries spend less money on primary education, and
there is no systematic difference in average scores. This points to the
importance of incentives outside the school system, which vary systematically according to the dominant product market strategies of firms and
their associated demand for particular skills.
But general skills systems are not necessarily bad for all types of
inequality. They perform better in terms of gender equality at work
(Estevez-Abe 1999b). When we compare degrees of occupational segregation, specific skills systems fare worse than general skills systems.
Specific skills systems segregate women into ‘female occupations’ such as
low-rank clerical and service jobs. Table 4.5 shows the occupational breakdown of women employed in the manufacturing sector expressed in
terms of a percentage of women over total workforce within the same category. While the data are not conclusive, it nonetheless shows that countries (see Germany and Sweden in Table 4.5) that adopt high-quality
Social Protection and Skill Formation 179
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1
Wage-bargaining system
Skills system
Industry
coordinated
Biased towards
specific skills
Not industry
coordinated
.49 (.34) (N=9)b .34 (.28) (N=3)c
Biased towards
general skills .29 (.23) (N=6)d
Fig. 4.4 Skills, the bargaining system, and equality
a Numbers are D9/D1 earnings ratios based on gross earnings (including all employer contributions for pensions, social security, etc.) of a worker in the bottom decile of the earnings distribution relative to a worker in the top decile. Figures are averages for the period 1977–1993.
Numbers in parentheses are D9/D1 income ratios based on disposable income of a person in the
bottom decile of the earnings distribution relative to a person in the top decile. Most figures are
from the early 1990s, with a few from the 1980s.
b Austria, Belgium, Denmark, Finland, Germany, the Netherlands, Norway, Sweden, Switzerland.
c France, Italy, Japan.
d Australia, Canada, Ireland, New Zealand, UK, USA.
Sources: Skills: see Table 4.4. Bargaining system: see Iversen (1999a: ch. 3). Inequality measures:
see OECD Employment Outlook (1991, 1996); Gottschalk and Smeeding (2000: fig. 2).
product market strategies—thus dependent on high industry-specific
skills—employ women for production jobs to a lesser degree. The USA,
the archetypal general skills system, shows significantly higher ratios of
women in technical and managerial positions when compared to specific
skills systems. Our findings support Esping-Andersen’s argument about
the US employment system being more gender equal than that found in
Germany and Sweden (Esping-Andersen 1999). Our explanation, however, differs from his.
4.5 Conclusion
Protection of employment and income is widely seen in the welfare state
literature as reducing workers’ dependence on the market and employers
(‘decommodification’ in Esping-Andersen’s terminology). In turn, this is
argued to reflect a particular balance of power between labor and capital.
We reject both theses. Although strong unions and left governments
180 Estevez-Abe, Iversen and Soskice
Percentage failing test
0 5 10 15 20 25
Sweden
Germany
The Netherlands
Switzerland
Belgium
Australia
Canada
New Zealand
United Kingdom
United States
Ireland
Fig. 4.5 The failure of early school leavers to pass standardized tests in eleven
OECD countries
Notes: The numbers are the percentage of all early school leavers taking the test who get a failing
score. Average across four test categories. The Belgian figure refers to Flanders only.
Source: OECD (2000).
undoubtedly affect distributive outcomes, we have argued that employment and income protection can be seen as efforts to increase workers’
dependence on particular employers, as well their exposure to labor
market risks. Moreover, social protection often stems from the strength
rather than the weakness of employers.
The key to our argument is the link between social protection and the
level and composition of skills. In a modern economy, skills are essential for firms to compete in international markets, and depending on the
particular product market strategy of firms, they rely on a workforce
with a certain combination of firm-specific, industry-specific, and general
skills. To be cost-effective firms need workers who are willing to make
personal investments in these skills. And if firms want to be competitive
in product markets that require an abundance of specific skills, workers
must be willing to acquire these skills at the cost of increasing their
dependence on a particular employer or group of employers. Because
investment in specific skills increases workers’ exposure to risks, only by
insuring against such risks can firms satisfy their need for specific skills.
The particular combination of employment protection and unemployment protection determines the profile of skills that is likely to emerge
in an economy. Thus employment protection increases the propensity of
workers to invest in firm-specific skills, whereas unemployment protection facilitates investment in industry-specific skills. The absence of both
Social Protection and Skill Formation 181
TABLE 4.5 Share of women by occupation in manufacturing (%)a

related
workers
USA (1989) 22 26 70 33 30 30
Japan (1988) 10 7 58 11 40 39 b
Germany (1986) 15 11 59 52 67 21
The Netherlands (1993) 14 0 55 32 45 12
Sweden (1989) 15 57 25 72 24
Australia (1987) 8 18 20 43 76 31
a Percentages represent the ratio of women over the total of men and women employed within each occupational category in manufacturing.
b The female ratio for occupational category (6) in Japan is exceptionally high due to a demographically
shrinking pool of young male workers (Estevez-Abe 1999b).
Source: ILO (198990).
gives people strong incentives to invest in general skills. These predictions are borne out by the comparative data, which show that most countries combine either low protection with general skills, or high protection
with specific skills.
Two factors contribute to the distinctiveness and resilience of particular welfare production regimes. The first is that such regimes tend to
be reinforced by institutions M collective wage-bargaining systems, business organizations, employee representation, and financial systems—that
facilitate the credible commitment of actors to particular strategies, such
as wage restraint and long-term employment, that are necessary to
sustain cooperation in the provision of specific skills. The second is that
those workers and employers who are being most advantaged by these
institutional complementaries also tend to be in strong political positions,
in terms of both economic clout and cheer numbers. For example, the
more a welfare production system emphasizes the creation of specific
skills, the more likely it is that the median voter will be someone with
considerable investments in specific skills, and the more likely it is that
employers’ interest organizations will be dominated by firms pursuing
specific skills strategies. Both will contribute to perpetuating institutions
and policies that advantage firms and workers with heavy investments
in specific skills.
Our argument has broader implications for our understanding of the
welfare state that reach well beyond the immediate effects of employment and income protection. In particular, earnings dispersion, by far
the most important determinant of the overall distribution of income, is
closely related to particular skill systems as well as the wage-bargaining
institutions that tend to go with these systems. Similarly, the combination of particular product market strategies and skills has distinct effects
on the career opportunities of particular groups, especially women. Thus,
our theory implies that gender-based segmentation of the labor market
varies systematically across welfare production systems.
Clearly, what we have done in this chapter is to outline a broad
research agenda for the study of the welfare state rather than testing a
specific set of hypotheses that follows from it. Much work needs to be
done, for example, in testing whether public opinion, voting behavior,
and the preferences of employers conform to the predictions of the
theory. Another big task is to rewrite social history to take into account
the preferences of employers, and the attempt by firms to engage in
particular product market strategies. This new emphasis on firms differs
from the earlier focus on the role of welfare capitalists—particularly in
case studies of the American social policy—in its explicit effort to high-
182 Estevez-Abe, Iversen and Soskice
light links between skills, product markets, and different welfare
systems. Some work has already been undertaken (cf. Estevez-Abe 1999a;
Mares 1998; Swenson 1997, forthcoming), but there are ample opportunities to expand on their pioneering research. Finally, much work
remains to be done to explore the implications of our argument for labor
market stratification. We have indicated the empirical relationship
between skill profiles, wage-bargaining systems, and labor market stratification (including women’s position in the labor market); most of the
empirical work is still to be done.
Social Protection and Skill Formation 183
1111
23456789
10111
123456789
20111
123456789
30111
123456789
40
1