. 3
( 11)


A Brief History of Modern Welfare Production Regimes

are developed in part by accumulating job experience from many different
The drawback of using tenure rates to measure ¬rm-speci¬c skills is that
they may also in part re¬‚ect the costs of dismissing workers as a result of
employment protection. However, if higher tenure rates were unrelated
to the extent of ¬rm-speci¬c skills, then the association between employ-
ment protection and tenure rates would be weak at best because most job
switching is known to be voluntary. But, in fact, the cross-national associ-
ation between the two variables is rather high (r = 0.75). Where data are
available, tenure rates are strongly negatively related to quit rates. From this
relationship, it seems clear that at least part of the effect of employment pro-
tection on tenure rates must go through the effect of the former on the stock
of ¬rm-speci¬c skills. This interpretation is supported by considerable ev-
idence showing tenure rates across industries within countries to be closely
associated with the skill intensity of these industries (OECD 1993, 141“5).
Used as a measure of ¬rm-speci¬c skills (column 2 in Table 2.6), tenure
rates suggest that the stock of such skills is low in the Anglo-Saxon coun-
tries compared to Japan and most of the continental European countries.
The exceptions are Denmark, The Netherlands, and Switzerland, where
¬rm tenure rates also tend to be quite short. In these countries, ¬rms tend
to be small with lower organizational capacity to adapt to the business cy-
cle and limited R&D capacity. Although these countries have developed
vocational training systems, ¬rms typically depend more on industry tech-
nologies and skills, as well as employment and income protection at the
industry level. Generous unemployment bene¬ts, for example, allow small
¬rms to “park” skilled workers in the unemployment bene¬t system dur-
ing downturns without undermining the incentives of workers to invest in
relevant skills.
The general pattern emerging in Table 2.6 is supported by a recent
detailed comparative study of training systems in France, Germany, Italy,
Japan, Sweden, the United Kingdom, and the United States (Crouch et al.
1999). In this comparison, the United Kingdom and especially the United
States stand out as general skills countries, coupled with a weak, but van-
ishing, apprenticeship system in the case of the United Kingdom. Large
American machine-tool ¬rms with good apprenticeships abandoned these
in the 1970s as weak employers associations and fragmented wage bargain-
ing could not prevent rampant poaching behavior (Finegold et al. 1994).
In this comparison, France and especially Sweden stand out as having
good school-based vocational training systems (but with more emphasis
Welfare Production Regimes

on in-¬rm training and internal labor markets in France), while the
German system is highlighted for its capacity to cultivate deep vocational
skills through a combination of school and on-the-job training. Japan has
little school-based vocational training but combines a good basic education
with extensive training inside large corporations. Italy, ¬nally, ranks low
in terms of formal training institutions, but it provides good ¬rm-based
training in regions with well-organized unions and employer associations.

2.2.3. Product Market Strategies and the International Division of Labor
Figure 2.1 plots the eighteen OECD countries on the employment and
unemployment protection indexes. Because income protection is so closely
related to unemployment protection, the picture does not change radically if
the former is used in place of the latter. Note that countries are distributed
along a primary axis, corresponding to the southwest-northeast diagonal

Figure 2.1 Social protection and skill pro¬les.
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.
Sources: See Tables 2.2, 2.3, and 2.4.

A Brief History of Modern Welfare Production Regimes

in Figure 2.1, with some countries further divided along a secondary axis,
corresponding to the northwest-southeast diagonal in Figure 2.1. The main
axis separates countries into two distinct welfare-production regimes: one
combining weak employment and unemployment protection with a general
skills pro¬le, represented by the Anglo-Saxon countries and Ireland, and
one combining high protection on at least one of the two social protection
dimensions with ¬rm- and/or industry-speci¬c 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 ¬rm-speci¬c skills, exempli¬ed primarily by Japan and
Italy,15 and one with greater emphasis on unemployment protection and
the production of industry-speci¬c skills, exempli¬ed by Denmark, The
Netherlands, and Switzerland.
The data on skills presented in Table 2.6 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 best developed vocational training systems, and
tenure rates decline with employment protection. Clearly, the empirical
patterns we observe are quite consistent with the notion 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 compara-
tive advantages in the global economy. Thus, where there is a large pool
of workers with advanced and highly portable skills and where social pro-
tection is low, companies enjoy considerable ¬‚exibility in attracting new
workers, laying off old ones, or starting new production lines. This ¬‚exibil-
ity allows for hightened responsiveness to new business opportunities and
facilitates the use of rapid product innovation strategies. In economies with
a combination of ¬rm- and industry-speci¬c skills, such strategies are ham-
pered by the dif¬culty of quickly adapting skills to new types of production
and by restrictions in the ability of ¬rms to hire and ¬re workers. On the
other hand, these welfare-production regimes advantage companies that
seek to develop deep competencies within established technologies and to
upgrade and diversify existing product lines continuously.
Although this book is not intended as a comprehensive test of these
propositions, they are broadly consistent with accounts in the comparative

15 Although the position of Italy is probably exaggerated by the failure to account for semipub-
lic unemployment insurance arrangements, as noted earlier.

Welfare Production Regimes

literature of the international division of labor that had evolved during the
1950s and 1960s as a result of institutional and policy differences. Porter
(1990), Streeck (1991), Soskice (1999), and Hollingsworth and Boyer (1997)
all describe a pattern where ¬rms in continental European countries, espe-
cially in northern Europe, were producing for established product markets
but continuously upgrading and diversifying their product lines to take ad-
vantage of the most lucrative market niches. Porter, using data from the
early to mids 1980s, shows that countries such as Germany, Switzerland,
and Sweden had competitive advantages in a broad range of machinery in-
dustries and that the number of industries with an advantage was an order
of magnitude greater than in the United States or the United Kingdom.
The reverse pattern was apparent in industries using rapidly changing new
technologies, as well as in internationally traded service industries where
there is a premium on being able to hire highly mobile and well-educated
employees from the external market (Porter 1990).
These ¬ndings are echoed by quantitative evidence developed by
Thomas Cusack using U.S. Patent Of¬ce data. Broken into thirty tech-
nology classes, Cusack counted the number of references to scienti¬c ar-
ticles for patents in each technology class and country and then divided
this number by the world number of scienti¬c citations per technology
class.16 The idea is that the number of scienti¬c citations, as opposed to
citations to previous patents and nonscienti¬c sources, is a good proxy
for the extent to which national ¬rms are engaged in radical innovation
strategies. The results are shown in the ¬rst column of Table 2.7, with
countries ranked by the average ratio of scienti¬c citations for patents se-
cured by national ¬rms. As it turns out, the Anglo-Saxon countries and
Ireland all have ratios that are signi¬cantly higher than in the speci¬c skills
countries of continental Europe and Japan. This is precisely as expected,
although much work is clearly needed to establish a conclusive linkage be-
tween institutions and innovation strategies “ a task outside the scope of this
At the low-tech end of product markets, it is necessary to rely on a differ-
ent type of data to detect cross-national differences. Column 2 of Table 2.7

16 The data are coded into references to previous patents and “others,” where many of the
latter are references to scienti¬c articles. To get a good estimate of the number of scienti¬c
articles in the other category, the proportion of scienti¬c 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 data set so as to get a better measure of scienti¬c

A Brief History of Modern Welfare Production Regimes

Table 2.7. Scienti¬c Citation Rates and Low-Wage Service
Employment in Eighteen OECD Countries

(1) (2)
Scienti¬c Private Service
Citation 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
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 scienti¬c citations per patent by national
¬rms in each of thirty technology classes as a proportion of the
average number of citations in each class for the entire world, 1985“
95. Source: U.S. Patent Of¬ce data.
b The number of people employed in wholesale, retail trade, restau-
rants and hotels, and community, social and personal services, 1982“
91 as a percentage of the working age population. Source: OECD,
International Sectoral Data Base (1996).

uses the proportion of the working age population employed in private so-
cial and personal services as a proxy. As I will show later, ¬rms that rely
heavily on low-skilled and low-paid labor for pro¬tability tend to be con-
centrated in these industries. Although there is only data for a subset of
countries, the numbers display a rather clear cross-national pattern. Pro-
ducers of standardized and low-productivity services thrive in general skills
countries such as Australia and United States because they can hire from a
large pool of unskilled workers who are afforded little job protection and
whose wages are held down by low unemployment protection. By contrast,
¬rms trying to compete in this space in speci¬c skills countries such as
Welfare Production Regimes

Germany and Sweden are inhibited by higher labor costs and lower ¬‚ex-
ibility in hiring and ¬ring. These differences were magni¬ed during the
1980s and 1990s, and Britain is now closer to the mean for the general
skills countries.
One of the catalysts for greater diversi¬cation in industrial structures
and the supporting social and labor market institutions was the economic
crisis of the 1970s. Starting with the “hot” summer of 1968, wage mod-
eration collapsed in some countries, and in¬‚ation exploded. The increases
won by strikers in 1968“9 were about twice those of the preceding three
years (Allsopp 1983, Table 3.4). One of the sources of in¬‚ationary wage
pressure was the rising demand for labor, which caused unemployment to
reach unprecedented low levels across the continent. With the share of
employment in agriculture having declined to less than 15 percent, elas-
tic supplies of underemployed labor from the agricultural sector no longer
capped industrial wage demands. Johansen™s (1987, pp. 148“9) description
of the situation in Denmark is representative: “In the mid-1960s,” he writes,
“the registered unemployed were either workers who were in the process
of changing from one job to another and had a few idle days in between, or
older people staying in isolated municipalities in Northern Jutland or the
smaller islands from where they did not want to move.” Under such condi-
tions, the threat of unemployment no longer disciplined wages. Memories
of high unemployment faded as the postwar generation of workers aged and
retired. The Soviet threat was perceived as less immediate, removing one
incentive for labor and capital to pull together. With the weakening of the
Bretton Woods System and its breakdown in the early 1970s, in¬‚ationary
expectations lost their anchor.
The result was a great downward pressure on industrial employment that
was exacerbated by the beginning shift in consumer demand away from
manufactures and toward services. Governments responded to the con-
sequent unemployment by increasing their spending to sustain demand.
They responded to the breakdown of wage moderation by encouraging
further centralization of negotiations and more formal incomes policy ar-
rangements (Scharpf 1991). Unions were promised increased health and
unemployment payments and larger social security stipends as the quid
pro quo for restraint. Public spending as a share of gross domestic prod-
uct rose from 24 percent in 1967“9 to 30 percent in 1974“6.17 While the
growth in spending as a percentage of GDP had been rapid in the 1960s, its

17 This is an unweighted average for thirteen European countries.

A Brief History of Modern Welfare Production Regimes

expansion was even faster in the 1970s. It was particularly dramatic in The
Netherlands, Denmark, and Sweden, where public spending was tied to the
expansion of transfer payments and social programs.
This strategy worked best where the institutions of corporatism and
centralized wage bargaining were most advanced.18 Fiscal expansion and
accommodating monetary policy stimulated employment rather than in¬‚a-
tion, given agreements by the unions to restrain their wage demands. Where
private-sector employment growth lagged, governments supplemented it
with increases in public employment or early retirement schemes. Con-
trary to popular notions, the pressures of the international crisis, especially
rising unemployment, did not force or induce generous welfare states to
cut back. This is most evident with respect to unemployment protection,
as illustrated in Figure 2.2. The ¬gure splits countries into high- and a
low-protection categories, based on the unemployment protection index
introduced earlier, and then shows the evolution of average unemployment
replacement rates in each category during the period from 1971 to 1995.
Note that whereas high-protection countries signi¬cantly raised replace-
ment rates, low-protection countries kept them low.19
Although the administration and long-run generosity of unemployment
bene¬ts have everywhere been tightened, which is not apparent in the ¬g-
ure, the cross-country pattern is consistent with the argument in this book
since support for generous unemployment protection among unions, em-
ployers, and individual workers should be a function of their exposure to
risk. Because risk is linked to skills, and because speci¬c skill countries tend
to have relatively generous protection, when exogenous shocks raise the
level of risk, it is not surprising that the effect is greater in the speci¬c
skills countries with already high generous protection. The one exception
is Japan where unemployment bene¬ts were relatively low and remained
low. In this case, as argued in Chapter 1, because skills are almost entirely
¬rm-speci¬c, high-employment, not unemployment, protection is the key
to insurance against risk.
Of course, in all the speci¬c skills countries, there was a greater pre-
mium on retaining employment, partly to reduce the exposure to risk but
also to lighten the ¬scal burden of generous bene¬ts. In many countries, the
corporatist solutions to the economic crisis, described earlier, worked quite

18 Typically, in the smaller European democracies. See Katzenstein (1984, 1985).
19 There are exceptions to this pattern. In Ireland, replacement rates were raised during the
1970s only to be cut back again in the 1990s. In Germany, there is little change over time.

Welfare Production Regimes

well. In Austria and Sweden, for example, unemployment was kept remark-
ably low at 1.9 percent of the labor force in 1973“83 (see Table 2.8). In
Germany, where the unions similarly restrained wages but macroeconomic
stimulus was subdued as a result of the strong antiin¬‚ationary predisposition
of the Bundesbank and de¬cit reductions by capital-market-constrained
state and local governments, unemployment still averaged less than 5 per-
cent during this turbulent period. By comparison, in countries like Britain,
Italy, Canada, and the United States where corporatist institutions were
missing or less well developed, demand stimulus aggravated in¬‚ation in-
stead of reducing unemployment. In the United States, this was exacerbated
by the Vietnam War and the accommodating stance of the central bank.


High-protection countries

Replacement rate


0.3 Low-protection countries

1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995

Figure 2.2 Unemployment replacement rates, 1973“1995.
Notes: This division of countries into the high- and low-protection groups is based on the un-
employment protection index introduced in Chapter 1. High-protection countries: Austria,
Belgium, Finland, France, Germany, Netherlands, Norway, Sweden, and Switzerland; low-
protection countries: Australia, Canada, Ireland, Italy, Japan, New Zealand, United Kingdom,
and United States. Because time-series data for this index do not exist, the ¬gures in this
graph are based on gross replacement rates for the ¬rst year of unemployment for a repre-
sentative 40-year-old worker (averaged over different family situations). These rates do not
necessarily correspond to the protection index across countries, but it is assumed that change
over time in underlying protection levels is correlated with changes in gross replacement
Source: OECD, Database on Unemployment Bene¬t Entitlements and Replacement Rates (undated).

A Brief History of Modern Welfare Production Regimes

Table 2.8. Unemployment in Selected Countries, 1950“1996

1950“59 1960“72 1973“83 1984“96
Switzerlanda 0.8 0.1 0.4 1.7
Sweden 2.2 1.5 1.9 4.2
Austria 3.9 1.8 1.9 3.6
Norway 1.5 1.2 2.0 4.3
Japan 1.3 1.1 2.0 2.6
Finland 1.3 2.0 4.5 9.4
Germany 6.1 1.0 4.7 8.0
France 0.6 0.9 5.4 10.5
Netherlands 2.0 1.3 6.5 8.4
United Kingdom 1.6 2.3 6.5 9.1
Denmarka 9.5 3.5 7.1 7.9
United States 4.4 4.9 7.2 6.3
Italy 9.3 5.3 7.4 11.2
Canada 4.2 5.4 7.9 9.7
Belgium 9.1 3.4 8.2 10.2
Ireland “ 5.2 8.5 15.6
a Figures for Denmark and Switzerland are not directly comparable to the rest because of
different de¬nitions of unemployment.
Source: OECD Employment Outlook, various years.

In addition to differences in the capacity of governments to respond to
macroeconomic pressures, new technology and diverging product market
strategies started to create a more fundamental divide between the social
market economies of continental Europe, the corporate welfare model of
Japan, and the more liberal Anglo-Saxon countries. As the postwar wave of
Fordist mass production methods gave way to more skill-intensive, science-
based technologies and ¬‚exible specialization, ¬rms that were in long-term
cooperative relations with their employees and had access to workers with
good vocational training thrived. Firms in countries with adversarial labor-
management relations suffered not only because they had more dif¬culty
containing wage pressures, but also because they had no effective way to
provide their employees with the kind of deep ¬rm- and industry-speci¬c
skills that were required to compete effectively in the new high-quality, and
hence high-value-added, product markets.20

20 These differences are evident in a wave of industrial relations reforms in high-protection
countries during the 1970s. Although varying in depth and breadth, these reforms all
improved employee representation at the plant level and in some cases in corporate board
rooms. By and large, the reforms were consistent with the production model that emerged

Welfare Production Regimes

The loss of competitive edge for American ¬rms was a major concern
in several in-depth studies. Thus, the research team led by Michael Porter
found relative decline in such industries as automobiles, machinery, ma-
chine tools, consumer electronics, and of¬ce products, and they detected
a beginning deterioration in many others (Porter 1990, p. 519). A team of
MIT researchers came to similar conclusions, emphasizing that U.S. prod-
ucts were often lacking in quality, timeliness of service, ¬‚exibility, and other
business aspects where Japanese and European ¬rms excelled (Dertouzos
et al. 1989, p. 33). In both studies, the institutional sources of weakness
were precisely those that gave coordinated market economies an edge dur-
ing this period: a weak vocational training system, hierarchical corporate
governance structures, arms-length relationship between buyers and sup-
pliers, and short-termism in the equity-based ¬nancial system.
Based on the competitive advantage in diversi¬ed quality markets, and
despite the slowdown resulting from the global recession, ¬rms in both
Europe and Japan were well positioned to take advantage of new export
opportunities, and they kept unemployment below the level in most other
countries. In Japan, growth was actually accelerating, and equity markets
took off in the early 1980s. By contrast, growth fell sharply in Britain and
the United States, and unemployment rose more rapidly than elsewhere.
International trade, the one area of the world economy experiencing rapid
growth, became increasingly dominated by Japan and continental Europe,
and both Britain and the United States experienced a slowdown in exports in
the latter half of the 1970s (in Britain, export growth actually fell below the
domestic rate of growth). Commensurate with poor economic performance,
British and American equity markets went through protracted bear markets,
and commentators started to sound alarmist about the future of Anglo-
Saxon capitalism (e.g., Zysman and Cohen 1987; Barlett and Steele 1992;
Madrick 1995).21
It is against this backdrop that one must understand the radical deregula-
tion of the British and U.S. economies during the 1980s and 1990s. Instead

in these countries and were broadly consistent with the interests of high-quality producers
and skilled workers (Windolf 1989).
21 James Cramer, a highly successful hedge fund manager, recalls the defensive mood among
U.S. investors in the early 1980s: “my clients were scared to death that the Japanese were
going to destroy us in the ™80s as they took over every one of our industries. My pitch
was always the same: ˜One thing that is never going to happen is that you won™t ever see a
bottle of Mitsubishi ketchup on the table. It will always be good old Heinz.™” Commentary,
December 5, 1999, TheStreet.com (www.thestreet.com).

A Brief History of Modern Welfare Production Regimes

of trying to emulate the welfare production regimes of northern Europe and
Japan, and despite political rhetoric to that effect during the late 1970s and
early 1980s,22 these countries embarked on reforms that would strengthen
their comparative institutional advantage: low-cost mass-produced services
and high-tech radically new products. For the continental European coun-
tries and Japan, despite the manufacturing successes of the 1970s, new ten-
sions associated with the rise of services threatened the long-term viability
of the high-protection strategy during the 1980s and 1990s. As I discuss
in the next section, a strategy of wage compression, high job protection,
and high taxes was not conducive to the growth of labor and low-pay in-
tensive services, even though it was entirely compatible with manufacturing
To summarize, by the early 1970s, advanced economies had divided into
two broad categories: one with generous social protection, speci¬c skills,
and competitive advantages in established product markets, and one with
low social protection, general skills, and comparative advantages in new
product markets. There is no simple causal story behind this bifurcation. It
resulted from a complex interaction between economic conditions, inher-
ited organizational capacities of employers and unions, initial differences in
skill systems, and differences in political structures. Where institutional ca-
pacities were high, where there existed a craft tradition, where there were
consensual rather than majoritarian institutions, and where the potential
for economic growth was high, cooperative solutions of institutionalized
wage restraint and an expansion of the welfare state emerged. In turn, these
solutions reinforced comparative advantages in product markets requiring
speci¬c skills investments and long-term cooperative relationships between
¬rms and their employees. Where institutional capacities were low, on the
other hand, high social protection tended to exacerbate collective action
problems and pushed governments toward deregulation and welfare state
cutbacks, while encouraging individuals to seek self-insurance, primarily
through heavy investments in general skills.

2.3. A Change of Fortunes: The Rise of the Service Economy
The high-protection countries with relatively egalitarian distributions of
income performed remarkably well during the 1950s and 1960s on most
economic measures. The insecurity and inequality of capitalist markets

22 See especially Reich (1983).

Welfare Production Regimes

had been tempered by high levels of social protection and wage equal-
ity while simultaneously giving European ¬rms a comparative advantage
in relatively established but high-quality segments of international prod-
uct markets. It was not simply that European ¬rms learned to live with a
generous social protection system, but that they depended on this system
to overcome worker disincentives to invest in speci¬c skills and to secure
union cooperation in wages and work organization. Centralized wage bar-
gaining, in particular, appears to have played a key role in adjusting to the
international economic crisis of the 1970s, as argued by many students of
Yet, somewhere in the mid- to late 1970s, the fortunes of different welfare
production regimes started to change. In some countries, this was probably
tied to the extension of egalitarian norms to the point where it became coun-
terproductive for skill formation and interfered with the introduction of new
technology (Hibbs and Locking 2000). Relatively stable wages between skill
categories provide a protection of skill investments, but in some countries
solidaristic wage policies signi¬cantly reduced wage differentials across skill
groups. I have discussed these problems at length elsewhere (Iversen 1996).
On the whole, however, these problems probably only made a small dent
into the international market shares of European companies, and they were
dealt with by reorganizing collective bargaining institutions and macro-
economic policy regimes without signi¬cantly reducing the overall level of
social protection for full-time skilled employees.
But while industrial relations reforms may have helped restore any lost
competitiveness by European ¬rms, high-protection countries experienced
a growing employment problem that did not go away with the success
of European exporters. The problem is illustrated in Figure 2.3, which
shows the gap in unemployment and private sector employment perfor-
mance in manufacturing and services between the United States and eleven
high-protection countries in Europe starting in 1950.23 A positive number
for unemployment means that European rates exceed those in the United
States, while a positive number for employment means that private sec-
tor employment as a percentage of the working age population is higher
in Europe than in the United States. Note that Europe outperformed the
United States on unemployment until the mid-1980s, and that private sec-
tor employment in manufacturing and private services almost caught up to

23 The eleven countries are Austria, Belgium, Denmark, Finland, France, Germany, Italy,
Netherlands, Norway, Sweden, and Switzerland.

A Brief History of Modern Welfare Production Regimes

U.S. levels in the early 1960s where it remained until the 1970s. Thereafter
a gap in the United States™ favor quickly opens up.
Figure 2.4 provides a breakdown of the employment numbers by sector.
One trend is unsurprising: because Europe began the period with about two
to three times more employment in agriculture than in the United States
(19 percent of the working age population in Europe versus 8 percent in
the United States), agricultural employment was bound to decline faster in
Europe relative to that in the United States. By 1995, agricultural employ-
ment converged to the low U.S. ¬gure of about 2 percent. Expansion of
employment in European manufacturing by and large made up for losses
until 1963, when manufacturing employment likewise began a process of
convergence to U.S. levels. By 1995, average manufacturing employment
in Europe was 18.1 percent of the working age population compared to
17.4 percent in the Unted States. This is essentially complete convergence.
In a very similar fashion, although with “opposite sign,” the period up
until the late 1960s was characterized by a gradual convergence of European
Gap between Europe and the United States (percentage)

-9 Private employment

1950 1955 1960 1965 1970 1975 1980 1985 1990 1995

Figure 2.3 The (un)employment gap between continental Europe and the United
States. The gap is de¬ned as average European (un)employment minus U.S.
(un)employment. Above the zero line, absolute (un)employment is higher in Eu-
rope. Employment is measured as a percentage of the working age population in
private sector employment. Unemployment is measured as standardized unemploy-
ment rates. Continental Europe refers the following countries: Austria, Belgium,
Denmark, Finland, France, Germany, Italy, Netherlands, Norway, and Sweden.

Welfare Production Regimes

service employment to U.S. ¬gures. Had this trend toward convergence
continued, Europe and the United States would essentially have had an
identical sectoral employment structure by the mid-1990s (as indicated by
the dotted line). But starting around 1970, the process of convergence in
services ended, and Europe and the United States grew increasingly apart.
In 1968, employment in private services was about 26 percent in the United
States and about 21 percent in Europe. Twenty-seven years later, the ¬gures
were 42 percent for the United States and 28 percent for Europe. Measured
as a share of the adult population, almost 50 percent more people were
employed in private services in the United States than in Europe. If we are
to understand the shift in private employment performance captured by
Figure 2.3, it is therefore critical to understand the causes of divergence in
private service employment performance.
The explanation advanced in this book is simple and follows directly from
the welfare production regime argument. Even though the egalitarian so-
cial welfare regimes of northern Europe created comparative advantages
in high-value added manufactures, they created disadvantages in precisely

Gap between Europe and the United States (percentage)




Linear projection based
Private services on data from 1950-69


1950 1955 1960 1965 1970 1975 1980 1985 1990 1995

Figure 2.4 The employment gap between continental Europe and the United
States, by sector. The gap is de¬ned as European employment minus U.S. employ-
ment. Above the zero line, absolute employment in a sector is higher in Europe
than in the United States. Employment is measured as a percentage of the working
age population.

A Brief History of Modern Welfare Production Regimes

Table 2.9. Average Annual Rates of Growth in Total Factor Productivity for Fourteen OECD
Countries, 1970“1994

1970“ 1975“ 1979“ 1983“ 1987“ 1991“ 1970“
1974 1978 1982 1986 1990 1994 1994
4.6 1.2 3.6 2.5 2.7 3.7
INDUSTRY 2.0 1.5 1.5 1.4 1.8 1.6
Manufacturing 2.8 1.8 0.1 1.6 1.3 2.7 2.1
Textiles 3.2 0.8 0.3 1.3 0.2 2.2 1.7
Chemicals 3.8 3.7 0.6 0.7 1.4 4.1 2.6
Machinery and equipment 3.6 0.7 0.5 1.7 1.6 3.4 2.6
Electricity, gas, and water 3.0 1.5 1.0 0.6 0.5 1.8
Transport, storage, and 2.2 1.7 0.3 2.1 3.2 2.0 2.7
’1.2 ’0.4 ’0.4 ’0.4
Construction 0.0 0.7 0.3
’0.5 ’0.4 ’0.5 ’0.2
SERVICES 0.2 0.2 0.4
’0.6 ’0.4 ’0.5
Private services 0.8 0.2 0.7 0.4
’0.9 ’0.5
Wholesale, retail, restaurants, 1.8 0.5 1.1 0.1 0.9
and hotels
’0.4 ’0.4 ’0.2
Finance, insurance, and 0.4 0.2 0.9 0.0
real estate
’0.4 ’0.7 ’0.6 ’0.2
Community, social, 0.4 0.1 0.3
and personal services
Productivity gapb 1.8 1.3 0.4 1.1 1.8 2.3 1.8
a Based on estimated labor productivity.
b Difference between industry and private services.
c n.a.: data not available.
Source: OECD, International Sectoral Data Base (1996).

the type of service production that was generating the most jobs during
the 1980s and 1990s. As we shall see in Chapter 6, most of the differential
in service sector employment performance was the result of relatively low-
skilled industries that were hurt by wage compression at the bottom end of
the wage scale as well as by in¬‚exible labor contracts and high costs of so-
cial protection for low-skilled workers.24 In addition, because productivity
growth in most services seriously lagged behind that in manufacturing (see
Table 2.9), a tightly coupled wage-setting system raised the relative prices

24 Although the problem could potentially have been addressed by specializing in high-value-
added services for exports, this option was unavailable at the time because of the very limited
trade in most services. I return to the issue of liberalization of services trade in Chapter 6.

Welfare Production Regimes

on these services and reduced demand.25 How important this “Baumol cost
disease” (Baumol 1967) is for employment depends on the price elasticity
of demand, but in a range of services where the product is not essential to
the consumer, the effect is substantial.
This argument is related to problems of unemployment in the follow-
ing way. Although the government can cope with sluggish private sector
employment growth by limiting the entry of new workers into the labor
market, by encouraging the exit of older workers, or by increasing pub-
lic sector employment, the incentives and opportunities for unemployed
workers with speci¬c skills to ¬nd work outside the industry or sector where
their skills are appropriate are very small. Therefore, when unemployment
is rising not as a result of temporary swings in the business cycle or shifts
in the relative performance of companies within a sector but because of
a secular decline in an entire industry, unemployment is likely to change
from a transitional to a structural problem as workers with speci¬c skills
become longterm unemployed. Combined with high unemployment pro-
tection, unemployed workers have neither good opportunities nor strong
incentives to look for employment outside their sector.
Finally, the con¬‚uence of workers with high industry- and sector-speci¬c
skills, rapidly declining employment in manufacturing, and sluggish growth
in new employment opportunities in services help explain the rapid growth
of the welfare state. Essentially, the new labor market insecurities could not
be dealt with inside the “private” protection system of wage coordination
and employment security. Deindustrialization threatened to force workers
out of the industries for which their skills had been developed. This had two
effects. First, if the incentives underpinning the speci¬c skills equilibrium
were to be retained, income protection through the public insurance system
had to rise to compensate for a lower effective rate of “private” wage protec-
tion. Second, the higher the probability of transitioning between jobs with
different skill requirements, the higher the preferred level of protection
at any given level of skill speci¬city and income. This implication of the

25 Productivity in services is notoriously dif¬cult to measure. However, a large empirical litera-
ture supports the productivity gap evident in the OECD data. See especially Gordon (1987)
and Van Ark (1996). Gordon, widely considered the leading authority on productivity, ¬nds
that services productivity lagged manufacturing productivity in Europe and the United
States by a factor of about 2 to 4 from the early 1970s to the mid-1980s. There is also evi-
dence for a productivity gap between services and manufacturing in the sizable economic
literature on real exchange rates. I discuss this literature in Chapter 6.

A Brief History of Modern Welfare Production Regimes

theory of social policy preferences is developed in Chapter 3. The broader
political, institutional, and economic consequences of deindustrialization
are considered in Part III.

2.4. Conclusion
This chapter has outlined the historical origins of the different welfare
production regimes that came to de¬ne the postwar political economies of
Europe, North America, and Japan. The emphasis has been on tracing the
processes leading to divergent institutional developments that are roughly
approximated by two ideal types or institutional equilibria. Each is charac-
terized by distinct couplings of social protection, skill systems, and political
institutions, and a key objective of this study is to provide an understanding
of the political underpinnings of these couplings.
Following Hall and Soskice (2001), a central theme is why social, po-
litical, and economic institutions tend to coevolve in a manner that rein-
forces rather than undermines one another. The welfare state is not “pol-
itics against markets,” as commonly assumed, but politics with markets.
Although it is popular to think that markets, especially global ones, in-
terfere with the welfare state, and vice versa, this notion is simply incon-
sistent with the postwar record of actual welfare state development. The
United States, which has a comparatively small welfare state and ¬‚exible
labor markets, has performed well in terms of jobs and growth during the
past two decades; however, before then the countries with the largest wel-
fare states and the most heavily regulated labor markets exceeded those in
the United States on almost any gauge of economic competitiveness and
Despite the change in economic fortunes, the relationship between so-
cial protection and product market strategies continues to hold. Northern
Europe and Japan still dominate high-quality markets for machine tools
and consumer durables, whereas the United States dominates software,
biotech, and other high-tech industries. There is every reason that ¬rms
and governments will try to preserve the institutions that give rise to these
comparative advantages, and here the social protection system (broadly con-
strued to include job security and protection through the industrial relations
system) plays a key role. The reason is that social insurance shapes the in-
centives workers and ¬rms have for investing in particular types of skills,
and skills are critical for competitive advantage in human-capital-intensive

Welfare Production Regimes

economies. Firms do not develop competitive advantages in spite of systems
of social protection but because of it.
Continuing this line of argument, the changing economic fortunes of
different welfare production regimes probably has very little to do with
growing competitive pressure from the international economy. To the
contrary, it will be argued in Chapter 6 that the main problem for Eu-
rope is the growing reliance on services that have traditionally been closed
to trade. In particular, labor-intensive, low-productivity jobs do not thrive
in the context of high social protection and intensive labor-market regu-
lation, and without international trade, countries cannot specialize in high
value-added services. Lack of international trade and competition, there-
fore, not the growth of these, is the cause of current employment problems
in high-protection countries.
The rest of this study seeks to spell out both the political foundations of
the persistence of institutions and the political responses to the economic
forces that challenge them. Part II considers the comparative politics of
social protection and focuses on electoral politics. Speci¬cally, Chapter 3
develops a human capital model of popular preferences for social protec-
tion, while Chapter 4 explores the institutional conditions that facilitate or
retard the translation of these preferences into policy outcomes. Finally,
Part III focuses on the challenges to existing welfare production regimes
caused by deindustrialization and other forces of structural change. First,
Chapter 5 relates deindustrialization to the expansion of the welfare state,
and how it has been conditioned by skill systems and political institutions.
Second, Chapter 6 discusses the economic tradeoffs engendered by dein-
dustrialization and how governments have responded to these.


Political Foundations of Social Policy

Explaining Individual Social
Policy Preferences—

As noted in Chapter 1, human capital rivals physical capital as a source of
personal and national wealth, and it is the single most important deter-
minant of personal income in advanced industrialized countries. Viewing
human capital as an investment or asset, this chapter asks how the character
of that investment affects workers™ preferences for social protection. The
fundamental idea is that investment in skills that are speci¬c to a particular
¬rm, industry, or occupation exposes their owners to risks for which they
will seek nonmarket protection, whereas skills that are portable, by contrast,
do not require extensive nonmarket protection.
The asset theory of preferences does not necessarily contradict a long
tradition in the study of the welfare state that emphasizes redistribution as
a key political motive behind the welfare state (e.g., Korpi 1983; Esping-
Andersen 1990). Indeed, Meltzer and Richard™s (1981) in¬‚uential median
voter result for government spending, which focuses on the redistributive
aspect of social protection, emerges as a special case in the model. Given a
particular composition of skills, workers with higher incomes are likely to
demand less social protection than workers with lower incomes. The asset
model parts ways with the Meltzer-Richard model, however, because it
explicitly recognizes that social protection also has an insurance aspect (Sinn
1995; Moene and Wallerstein 2001) and that demand for insurance varies
between workers according to their degree of exposure to labor market risks
(Baldwin 1990).

— This chapter is an expanded version of a co-authored paper with David Soskice published
in the American Political Science Review under the title of “An Asset Theory of Social Policy
Preferences” (Iversen and Soskice 2001).

Political Foundations of Social Policy

The model is tested on public opinion data from eleven advanced democ-
racies. Income and skill speci¬city, which is measured in a variety of ways,
turn out to be the overriding determinants of preferences for a range of
different social polices that broadly correspond to the various types of pro-
tection identi¬ed in Chapter 1. Gender is also important for explaining
policy preferences, and the effect is precisely what we would predict from
the discussion of gender and social protection in Chapter 1. Unlike the
impression one often gets from a largely nontheoretical comparative public
opinion literature, there is, thus, a great deal of structure to peoples™ social
policy preferences.
The chapter is organized into three sections. In the ¬rst, the model
and its main empirical implications are presented. The second tests these
implications on public opinion data from the International Social Survey
Program. The concluding section discusses the broader implications of the
model for explaining differences in social protection across countries.

3.1. The Model

3.1.1. Assumptions
Workers derive their incomes from skills that can be either general or
speci¬c. Speci¬c skills are skills that are valuable only to a single ¬rm or to
a group of ¬rms (whether an industry or a sector), whereas general skills
are portable across all ¬rms. Three different employment situations, or
states of the world, are distinguished, each associated with distinct levels
of income. In State I, a worker is employed in a ¬rm that utilizes both his
or her speci¬c and general skills; in State II the worker is employed in a
¬rm that only utilizes his or her general skills; and in State III the worker
is unemployed (i.e., none of his skills are being utilized).
In State II, g de¬nes the market value of a worker™s general skills when
his or her speci¬c skills are not being used. In State I (when the speci¬c skills
are being used as well), the worker is paid s g, the value of his or her com-
bined speci¬c and general skills. A worker that has no speci¬c skills (s = 1)
is always employed at the market value of his or her general skills. The key
assumption is that general skills are marketable in all sectors of the econ-
omy, whereas speci¬c skills are marketable only in one sector (the size of
which is de¬ned by the speci¬city of skills).
In addition to market income, which includes both wage and non-
wage compensation, workers receive transfer income from the government,
Explaining Individual Social Policy Preferences

hereunder unemployment bene¬ts, healthcare bene¬ts, pensions, and other
forms of nonwage compensation. Although some of these bene¬ts are re-
ceived by people who are outside the labor market, what matters to the
argument is that they are viewed by workers as part of their compensation
(what in the neocorporatist literature is sometimes referred to as the “social
wage”). Those who are most fearful of losing the labor market power of
their skills, and, hence, their ability to secure good health and pension plans
through their employer, will also be most concerned about guaranteeing a
high level of bene¬ts, even if the bene¬ts are “deferred” to the future.
Transfers are assumed to come in the form of a ¬‚at-rate payment, R,
which incorporates the idea in the Meltzer-Richard model that there is a
redistributive aspect to social protection.1 Following the terminology in
Chapter 1, one can distinguish between transfers that go to support the
income of employed workers, wage protection, and transfers that go to the
unemployed, unemployment protection. As the model is developed, we will
discuss what happens if R only goes to unemployment protection. But in
the main model, we will assume that all workers receive the same ¬‚at-rate
subsidy, which may simply be referred to as income protection. Employment
protection could be modeled as a probability of keeping a job where a worker™s
skills are fully utilized, and it will be considered separately in the empirical
analysis along with the other types of protection.
Transfers are paid out of a ¬‚at-rate tax (t) on all wages. Total per capita
receipts are T, and all receipts are spent on transfers (i.e., budgets are as-
sumed to be balanced). As in the Meltzer-Richard model, taxation is as-
sumed to create work disincentives, captured here by the following simple
labor supply function:

l(t) = 1/(1 + t) (3.1)

where l(t) is the number of hours worked or the intensity of effort (the
particular form of this function is chosen for mathematical convenience).
De¬ne w as average hourly pretax earnings. Then tax income per capita is

T = t · w · l(t) = =R (3.2)

1 This is also a realistic assumption with after-tax income distributed signi¬cantly more equally
than pre-tax income (see Gottschalk and Smeeding 2000; Huber and Stephens 2001; Bradley
et al. 2003).

Political Foundations of Social Policy

State I State II
(employed using specific (employed using only
State III
and general skills) general skills)
Share of LF: ± = r . q/( p + q) Share of LF: β = (1 ’ r) . q/(p + q)
r.q (1 ’ r) . q
Share of LF: γ = p/( p + q)
Disposable income: Disposable income:
(1 ’ t)
(1’t) Disposable income:
…sg + R …g + R
sg =
(1 + t)
(1 +t) R
p = (1 ’ 2R/w)…g + R
= (1’2R/w)… sg + R

Figure 3.1 The three states of an individual in the labor market.

Figure 3.1 illustrates the three states of labor market and shows the
disposable (after tax) income associated with each state: I, s g; III, R;
and II, g.
For a given period of time, there is a probability, p, of losing one™s job
and another probability, q , of reemployment. In equilibrium p · e = q · γ ,
where e is the share of employed workers and γ is the share of the work
force that is unemployed (e = 1 ’ γ ). This implies that in equilibrium e =
q /( p + q ). Furthermore, if r is the probability that an employed worker
is reemployed into State I (i.e., into a job where both general and speci¬c
skills are utilized), then the equilibrium share of the labor force employed
in State I is

± = r · e = r · q /( p + q ) (3.3)

Likewise, the share of the labor force employed in State II is

β = (1 ’ r) · e = (1 ’ r) · q /( p + q ) (3.4)

while the share of the labor force in State III (unemployment) is

γ = p/( p + q ) (3.5)

For any individual worker with both speci¬c and general skills, the propor-
tions ±, β, and γ can be interpreted as probabilities in a lottery with three
possible outcomes. An employed sg-worker will therefore seek to maximize
the expected utility of income across all three states. Ignoring the discount-
ing of future income (which makes no substantive difference to the results),
this is captured by the following utility function:

V = ± · u(s g) + β · u(g) + γ · u(R) (3.6)
Explaining Individual Social Policy Preferences

where u(·) is the worker™s utility from net income, which for simplicity
is assumed to be spent on consumption. Using standard assumptions, the
following constraints are imposed on u:

u c > 0,
u c c < 0, (3.7)


lim u (c ) = ∞
c ’0

A number of the following results hold for this general form of utility
function (notably the Meltzer-Richard results). However, because the in-
surance function of the social wage will play an important role and because
we then need speci¬c conditions on risk aversion, the standard assumption
of a constant Arrow-Pratt relative risk aversion (RRA) utility function is
used. Speci¬cally,
c 1’a
u(c ) = ∀ a > 0, =1 (3.7a)
= log c a =1

With these assumptions in mind, it is not possible to determine workers™
utility-maximizing preferences for social protection.

3.1.2. Optimizing Social Preferences
The logic of the presentation in this section is as follows. First, we consider a
simple base-line model with no insurance effects, no tax disincentives, only
general skills, and no unemployment (Model I). Then tax disincentives are
introduced to get the Meltzer-Richard result (Model II), followed by in-
surance effects (and unemployment) to explore the effects of risk-aversion
(Model III). Finally, we see what happens to the demand for social protec-
tion when the composition of skills is allowed to vary (Model IV). To keep
the presentation simple, all proofs are put in the appendix at the end of this

Model I. No insurance effects, no disincentive effects: the t = 1 model.
In solving the workers™ maximization problem, begin by assuming a labor
force with only general skills (s = 1) and no unemployment (e = 1). The
simplest case is where there are no tax disincentive effects on the number of
Political Foundations of Social Policy

hours supplied [so that l(t) = 1 rather than l(t) = 1/(1 + t) as subsequently
will be assumed].
When s = 1, e = 1, and l(t) = 1, Equation (3.6) reduces to

V = u((1 ’ t)g + tw) = u ( g(1 ’ R/w) + R) (3.8)
· ( g(1 ’ R/w) + R)1’a
We want to choose R to maximize V, where R is bounded between R = 0,
corresponding to t = 0, and R = w, corresponding to t = 1. Because

VR = ( g(1 ’ R/w) + R)’a · (1 ’ g/w)

and, because 0 ¤ R ¤ w, g > w implies that VR is uniformly negative for
all values of R and hence t: Maximization of V, therefore, requires t = 0.
Thus, voters with above average income will choose a zero tax rate. An
analogous argument for g < w implies that voters with below average in-
come will choose the maximum tax rate of 100 percent. This is the standard
result: In the absence of insurance functions and tax disincentives, voters will
want the maximum R (i.e., t = 1) if g < w and a zero R (t = 0) if g > w.
Thus, if the median voter, M, has an income less than the average income
of w, the median voter will always vote for a maximum tax rate. The result
is illustrated in Figure 3.2a.

Model II. Disincentive effects, no insurance effects: the Meltzer-Richard
model. If we now include the tax disincentive effect that l(t) = 1/(1 + t),
we have
1’t 2R
V=u g+ =u g 1’ +R (3.9)
1+t 1+t w
1 2R
= · g 1’ +R
1’a w
This implies

VR = ( g(1 ’ 2R/w) + R)’a · (1 ’ 2g/w) (3.10)

so that in the Meltzer-Richard model, only voters with a g level below that
of half the average hourly wage (g = w/2) will vote for a maximum tax rate.
As illustrated in Figure 3.2b, if the median voter has a g level above w/2 he
or she will, therefore, not vote for the maximum tax rate. Because of the
simplicity of the tax disincentive function, voters with g levels below w/2
Explaining Individual Social Policy Preferences

(b) The Meltzer-Richard Model
(a) The t=1 Model
1 1

0 0

M w w/2 M w
Income Income

(c) The Insurance Model (d) The Asset Model
R R sg
0 <RRA<
RRA> 1 s g ’ w /2
1 1

High s/g

Low s/g


M w
M w Income

Figure 3.2 Four models of social policy preferences. Arrows indicate preferred
level of social protection by the median voter.

will vote for t = 1, and voters with g levels above w/2 will vote for t = 0.2
With the more complex tax disincentive function used by Meltzer-Richard,
workers with income in the range [w/2, w] will prefer taxation up to the
point where the bene¬ts to them from redistribution are exactly outweighed
by the ef¬ciency costs of tax disincentives. If the median voter is in this
range, as the Meltzer-Richard model assumes, then he or she may vote for
a positive tax rate less than 1 (as illustrated in Figure 3.2).
One of the implications of the Meltzer-Richard model is that voter
turnout will be positively related to the level of government transfers
because nonvoting tends to be concentrated among low-income people

2 Meltzer-Richard have a more general tax disincentive function than that used here. Conse-
quently, the tax rate that maximizes tax revenue can be less than one.

Political Foundations of Social Policy

(Lijphart 1997). There is some cross-national evidence for this proposition
(see Franzese 1998). On the other hand, there is little empirical support for
another key implication of the model, developed by Alesina and Rodrik
(1994), namely that relatively inegalitarian societies will exhibit greater
pressures for redistributive spending than relatively egalitarian ones (see
Perotti 1996 for a review of the evidence). Among advanced countries, the
relationship is actually the reverse (B´ nabou 1996).

Model III. Disincentive effects, insurance effects: the insurance model.
Moene and Wallerstein (2001) offer one possible explanation for this puzzle
by introducing insurance effects (see also Sinn 1995). For insurance effects
to matter, we need at least two states of the world. By contrast to the Moene-
Wallerstein model (where there are two categories of employees, high and
low wage, in addition to the unemployed), it is assumed that workers can
either be employed at a gross wage equal to their “tax-incentivised” skill
level g/(1 + t) or be unemployed. There are no speci¬c skills (s = 1), so
Equation (3.6) becomes

V = β · u(g) + γ · u(R) (3.11)

In this model, if relative risk aversion is constant and greater than unity,
workers will choose a higher tax rate as they become wealthier; in other
words, their aversion to risk outweighs the increased cost to them of insur-
ance as their income increases. Thus, the relationship between income and
preferred level of spending is positive (see Figure 3.2c).3
To get this result, it is assumed that g = g · (1 ’ 2R/w) so that R is only
paid to those who are unemployed. It can then be shown (as is done in
Appendix 3.A) that

d R/d g < 0 RRA < 1
iff (3.12)


d R/d g > 0 RRA > 1
iff (3.13)

3 This differs from the Moene-Wallerstein model because workers with high incomes in
that model are assumed not to be exposed to labor market risks and, hence, prefer zero
spending and taxation. Because no distinctions are made in the risk exposure of different
income groups, model III is referred to as the insurance model rather than as the Moene-
Wallerstein model. In the empirical section the insurance model, not the Moene-Wallerstein
model, is tested.

Explaining Individual Social Policy Preferences

A key implication of the result for RRA > 1 is that, contrary to the Meltzer-
Richard model, a means-preserving increase in inequality will reduce the
median voter™s preferred level of social protection (provided that the in-
come distribution is skewed to the right). The reason is that such a rise in
inequality lowers the income of the median voter, and because the insur-
ance motive dominates the redistribution motive (RRA > 1), demand for
social protection will decline. In the Meltzer-Richard model, there is no
insurance motive, so a fall in the income of the median voter always leads
to a rise in the demand for social protection. Risk aversion, thus, poses one
potential solution to the empirical puzzle of why income equality is linked
to redistributive spending.
Yet, despite the neatness of this result, the econometric estimations
clearly reject its implication that people prefer less redistribution at lower
levels of income. This leaves the negative relationship between redistri-
bution and inequality as an important unsolved puzzle for comparative
political economy. In the conclusion to this chapter, how the distinction
between speci¬c and general skills permits an alternative and more plausi-
ble interpretation is discussed.

Model IV. Disincentive effects, insurance effects, speci¬c and general skills:
the asset model. This is the most general model and requires us to consider
all three states in Figure 3.1. We therefore return to the present value of
utility given by Equation (3.6):
V = ± · u(s g) + β · u(g) + γ · u(R) (3.6)
s g = s · g · (1 ’ 2R/w) + R
g = g · (1 ’ 2R/w) + R
In addition, an important new variable is introduced, expected hourly in-
come before taxes and transfers, y. It is simply de¬ned as
y ≡ ± · sg + β · g
Now we can ask, ¬rst, up to what value of y is the chosen R maximal
(i.e., t = 1); second, under what RRA conditions does R fall or rise as y
rises above this value; and, third, what happens to the choice of R as the
balance of general and speci¬c skills changes, holding y constant?
In the ¬rst result, a worker will only choose the maximum tax rate if y ¤
w/2. It is stated formally (the proof is in Appendix 3.A) in Result I.
Political Foundations of Social Policy

Result I. Given the assumptions of Model IV, t = 1 iff y ¤ w/2.
With this property established, we now consider what happens to R when y
increases. As shown in Appendix 3.A, this yields Result II.
Result II. Given the assumptions of Model IV, holding s constant and with
y > w/2,

‚R sg
= sgn RRA(s g) ’
‚y s g ’ w/2

What this equality says is that the direction of the relationship between R
and income (the sign, sgn, of ‚ R/‚ y) depends on the level of risk aversion,
just as in the simple insurance model. However, for income to be positively
related to support for spending, the RRA requirement is more stringent
(RRA > s g/(s g ’ w/2)) than before (RRA > 1). The reason is that R now
goes to the employed as well as to the unemployed, and because employed
workers in the insurance model only have an insurance incentive in rela-
tionship to unemployment, RRA must be higher for the insurance motive to
dominate the redistribution motive. This implication is also demonstrated
by Moene and Wallerstein.
Now we come to the critical result, which differentiates the proposed
approach from previous ones. Central to the argument in this chapter
is the proposition that an increase in speci¬c skills relative to general
skills, holding constant the level of expected income, implies an increase
in preferred R; put broadly, workers with speci¬c skills will prefer higher
taxes and social protection than workers with general skills. The following
result is also proved in Appendix 3.A.
Result III. Assuming a constant relative risk aversion utility function and
R R A > 0, ‚ R/‚s > 0 holding y constant. In other words, as s rises, the pre-
ferred level of R also rises. The intuition behind this key result is that
workers with speci¬c skills have more to fear if they lose their job than
workers with general skills. This is because speci¬c skill workers who are
laid off face the risk of being reemployed in a sector where their skills are
not needed. If this happens, they will lose some of their previous income, in-
cluding employer-provided insurance against illness and old age. General
skill workers do not face this problem because they are always compen-
sated at the value of their general skills. Hence, the more income derived
from speci¬c as opposed to general skills (i.e., the higher the ratio s /g),
the greater the demand for income protection, R. The logic is illustrated in
Explaining Individual Social Policy Preferences

Figure 3.2d and implies that the median voter™s support for social protection
depends on the composition of his or her skills.
The results in this section can be summarized as follows: With the sim-
plest set of assumptions “ only one state of the world (employment), only
general skills, and no tax disincentives “ the politics of social spending is all
about redistribution (class politics if you will); those with a wage below the
mean will want a maximum rate of taxation (t = 1) whereas those above the
mean will want zero taxation. If we add tax disincentives, however, the cost
of redistribution may deter those low-income workers closest to the mean
from demanding con¬scatory taxation, and the median voter is likely to be
among those workers. This is the Meltzer-Richard model.
When an unemployment state is added to the model, an entirely new
motive enters into workers™ calculations of their interests: insurance against
loss of income. If workers are suf¬ciently risk-averse and if all transfers go
to the unemployed, rising income may in fact be associated with higher
demand for protection because high-income workers have more to lose
than low-income workers. This is the simple insurance model. If some
transfers go to the employed, however, the threshold of risk-aversion for
which this relationship holds goes up because transfers to the employed
only serve redistributive purposes.
Finally, when differences in the speci¬city of skills, which require at least
two employment states (States I and II in the model), are introduced, the
insurance motive plays a crucial role even when workers are only moder-
ately risk-averse (0 < RRA < 1) and even when transfers are distributed
to both employed and unemployed workers. The reason is that employed
workers risk losing the income from their speci¬c skills, regardless of their
exposure to unemployment. This coupling between skills and demand for
insurance thus transforms the relationship between income and social policy

3.1.3. The Role of Gender
The asset model has implications for understanding gender differences in
social policy preferences. The possibility of a gender gap emerges when
marriage contracting is incomplete and termination of the contract is an
ever-present possibility. In this case, spouses will have con¬‚icting prefer-
ences over who receives family bene¬ts, and they will differ over any policies
that affect their outside options should the marriage break up. This is not
simply because spouses could one day be forced to take the outside option
Political Foundations of Social Policy

but also because outside options affect the current bargaining power inside
the family (Iversen and Rosenbluth 2003).
As soon as outside options matter for family bargaining power, men
and women will differ over the policies that affect these options. The most
obvious matter of disagreement between husbands and wives is over publicly
subsidized daycare. Because women are much more likely to end up as
the primary care givers, their welfare is disproportionately affected by the
availability of high-quality, low-cost daycare. Men may prefer to spare the
public purse, and hence their tax bill, if their wives are default child-care
givers. This logic also applies to public care for the elderly and the sick
because it helps women escape some of their traditional duties and thereby
permits more time to be spent in paid employment. In addition, as has been
stressed throughout, the welfare state is an important source of employment
for women precisely because so many of the jobs replace caring functions
that are otherwise provided “for free” in the family.
Men and women are also likely to disagree about a range of social transfer
programs. First, as pointed out in Chapter 1, research based on the Luxem-
bourg Income Study has shown that transfers always result in a reduction of
inequality (Bradley et al. 2003), and because women are on average paid less
than men, they tend to have a stronger interest in redistributive spending
(assuming again that they care about their outside options).4 Social transfers
are also important insofar as they allow women temporary career interrup-
tions without signi¬cant loss of income. As argued by Estevez-Abe (1999),
social protection for women involves two factors that are not equally rel-
evant for men: (i) protection against dismissal during and after pregnancy,
such as maternity, parental, and family leave policies; and (ii) income main-
tenance during leaves and guarantees of reinstatement to the same job at
the same wage level upon return to work.
The asset argument enters into the story because women will ¬nd it
harder to get jobs in occupations that require extensive speci¬c skills. For
a woman to invest in speci¬c skills, she needs to be assured that potential
career interruptions, if temporary, will not lead to dismissal or reduce her
4 The gap in income between men and women may be the result of “statistical discrimination”
where employers pay women less than men because women are on average more likely than
men to leave the labor market for purposes of child birth and rearing, and because women
are more likely than men to trade off working time for time on domestic duties, including
care for sick family members (see Daly 1994; Rubery, Fagan, and Maier 1996). There is a
vicious circle here because as women are paid less their bargaining position in the family
becomes weaker and they will spend more time on household duties to the detriment of
their attractiveness to employers.

Explaining Individual Social Policy Preferences

wage level in the long run. A high probability of dismissal reduces the
incentives to acquire ¬rm-speci¬c skills. Likewise, a high probability of
reduction in wages after becoming a mother reduces the incentives to invest
in either ¬rm-speci¬c or industry-speci¬c skills (Estevez-Abe 2002).
One key implication of the asset model is that women, provided that out-
side options do matter, at any given level of income and skill speci¬city will
prefer higher social protection than men. This gender gap is magni¬ed by
an indirect effect through income because women earn less than men. On
the other hand, the effect is reduced to the extent that women invest more
in general than in speci¬c skills. An interesting corollary of this argument
is that women should be more supportive than men of public investment in
general education. The reason is that inexpensive access to good formal ed-
ucation presumably bene¬ts women disproportionately because they have
a comparative advantage in general skills.
Using the logic developed in this section, we can revisit some broader
claims that are sometimes made about the gender and political prefer-
ences. Orloff (1999) and O™Connor, Orloff, and Shaver (1999), for example,


. 3
( 11)