. 6
( 11)


countries over a period that runs from the late 1960s (the ¬rst observation is
1967) to the late 1990s (the last observation is 1997). All fourteen countries
have been democracies since the Second World War. There are a total of

Credible Commitment, Political Institutions, and Social Protection

sixty-one observations, with the number of observations for each country
ranging from two to seven. About one ¬fth of the observations are from
the 1970s and late 1960s, about 40 percent are from the 1980s, and the
remainder are from the 1990s. The data are collected from separate national
surveys, but considerable effort has gone into harmonizing the data (or
“Lissifying” them) to ensure they are comparable across countries and time.
The LIS data are widely considered to be of high quality and the best
available for the purposes of studying distribution and redistribution (see
OECD 1995; Brady 2003).
I use the data speci¬cally to explore the determinants of redistribution as
measured by the percentage reduction in the Gini coef¬cient from before
to after taxes and transfers. The Gini coef¬cient is a summary measure
of inequality, which falls as income is shifted from those with higher to
those with lower incomes. It varies from 0 (when there is a perfectly even
distribution of income) to 1 (when all income goes to the top decile). Using
an adjusted version of the LIS data “ constructed by Huber, Stephens,
and their associates (Bradley et al. in press)12 “ I include only working
age families, primarily because generous public pension systems (especially
in Scandinavia) discourage private savings and, therefore, exaggerate the
degree of redistribution among older people. Furthermore, because data
are only available at the household level, income is adjusted for household
size using a standard square root divisor (see OECD 1995).
On the independent side, the key variable for explaining redistribution is
government partisanship, which is an index of the partisan left“right “center
of gravity” of the cabinet based on (i) the average of three expert classi¬-
cations of government parties™ placement on a left“right scale, weighted
by (ii) their decimal share of cabinet portfolios. The index was conceived
by Thomas Cusack who generously shared all the data from a new com-
prehensive source on parties and partisanship (see Cusack and Fuchs 2002
and Cusack and Engelhardt 2002 for details). The expert codings are from
Castles and Mair (1984), Huber and Inglehart (1995), and Laver and Hunt
(1992). For the purpose of explaining partisanship, the key variable is elec-
toral system. I use several different measures that are explained in detail in
the partisanship section later in this chapter.
I also controlled for variables that are commonly assumed to affect redis-
tribution, most notably income inequality. These variables, with de¬nitions,

12 I am grateful to the authors for letting us use their data.

Political Foundations of Social Policy

sources, and a short discussion of causal logic, are listed here. Country
means and a variable correlation matrix are provided in Tables A4.3 and
A4.4 in Appendix 4.E.

Pretax and transfer inequality. This variable is included to capture the
Meltzer-Richard logic that more inequality will lead to more pressure
for redistribution. It is measured as the earnings of a worker in the
ninetieth percentile of the earnings distribution as a share of the earn-
ings of the worker with a median income. The data are from OECD™s
wage dispersion data set (unpublished electronic data).
Constitutional veto points. This variable is Huber et al.™s (1993) compos-
ite measure of federalism, presidentialism, bicameralism, and the fre-
quency of referenda. The more independent decision nodes available,
the more veto points there are. One can raise de¬nitional objections to
the inclusion of referenda as a veto point, but it is clearly the case that
referenda are typically used to block legislation that would otherwise
have passed by a majority (see Lijphart 1999, pp. 230“1).
Unionization. According to power resource arguments, high union den-
sity should lead to more political pressure for redistribution while
simultaneously affecting the primary income distribution (see Huber
and Stephens 2001 and Bradley et al. 2003). The data are from Visser
(1989, 1996).
Voter turnout. Meltzer and Richard (1981) argue that the extension of
the franchise reduced the income of the median voter and raised the
demand for redistribution. A similar logic may apply to voter turnout if
non-turnout is concentrated among the poor as some research suggests
(Lijphart 1997). The data are from annual records in Mackie and
Rose (1991) and International Institute for Democracy and Electoral
Assistance (1997).
Vocational training. Iversen and Soskice (2001) argue that people with
speci¬c skills are more likely to support social insurance with a redis-
tributive component. As an indicator of the extent to which workers
are schooled in speci¬c vocational skills, as opposed to general aca-
demic skills, I use the share of an age cohort that goes through a
secondary or short-term postsecondary vocational training. The data
are from the UNESCO Statistical Yearbook (UNESCO, 1999).
Unemployment. Because the unemployed receive no wage income, they
are typically poor without transfers. Because all countries have public
unemployment insurance, higher unemployment will “automatically”
Credible Commitment, Political Institutions, and Social Protection

be linked to more redistribution. The unemployment ¬gures are stan-
dardized rates from OECD, Labour Force Statistics (various years).
Real per capita income. This variable is a standard control to capture
“Wagner™s law,” which says that demand for social insurance is in-
come elastic and, therefore, will tend to raise spending and re-
distribution. The data are expressed in constant 1985 dollars and
are from the World Bank™s Global Development Network Growth
Database (http://www.worldbank.org/research/growth/GDNdata.htm) “
itself based on Penn World Table, Global Development Finance, and World
Development Indicators.
Female labor force participation. Women™s participation in the job mar-
ket varies considerably across countries and time, and it is likely that
such participation matters for redistribution because it entitles some
women to bene¬ts (unemployment insurance, health insurance, etc.)
that they would otherwise not get. Whether this leads to more re-
distribution depends on the position of working women in the in-
come distribution as well as their family status, but there is a com-
mon presumption that women are more likely to be in low-paid jobs
and from low-income (single-parent) households. The measure is fe-
male labor force participation as a percentage of the working age
population and is taken from OECD, Labour Force Statistics (various

4.5.3. Statistical Model
The starting point is a simple error correction model. In this model, current
redistribution is equal to past redistribution plus a contribution from re-
distributive partisan policies that deviate from policies that would preserve
the status quo level of redistribution:

Ri,t = » · [± + β · Pi,t ’ Rt,t’1 ] + Ri,t’1 + ui,t

where u is identically and independently distributed with mean 0 and
variance s u .
With the available data on redistribution, however, one cannot estimate
this model directly because the observations on the dependent variable
for each country are unequally spaced, varying between 2 and as many as
10 years. To deal with this problem, I use a modi¬ed version of the model
where I substitute the preceding expression for Ri,t’1 , Ri,t’2 , and so on,
Political Foundations of Social Policy

until one gets to another observation of the lagged dependent variable.
This procedure yields the following expression:
Ri,t = » · ± · (1 ’ ») + » · β · (1 ’ »)s ·Pi,t’s

s =0 s =0
+ (1 ’ ») · Ri,t’N+1 + (1 ’ »)s · ui,t’s

s =0


Ri,t ’ (1 ’ ») N+1 · Ri,t’N+1
=»·±· (1 ’ ») + » · β · (1 ’ ») ·Pi,t’s + (1 ’ »)s · ui,t’s
s s

s =0 s =0 s =0

The second term in the last expression is a measure of the cumulative
effect of partisanship over a period of N years, where N is the gap between
the current and previous observation. Of course, insofar as other variables
affect redistribution, we need to calculate the cumulative effects of these
in precisely the same manner as for partisanship. Because there are annual
observations for partisanship and all control variables, the estimated model
is based on complete time series except for the dependent variable. The
model is estimated by choosing a value for » that maximizes the explained
Given the assumption that the composite errors are serially uncorre-
lated,13 but because the error term depends on N, there is heteroscedastic-
ity. To adjust for this, as well as contemporaneous correlation of errors, I
use panel corrected standard errors as is common when analyzing pooled
cross-sectional time-series data (see Beck and Katz 1995).
The model used to explain partisanship in the second part of the analysis
is a simple ordinary least squares regression that is explained later.

4.5.4. Findings Redistribution I begin the presentation with the results from
estimating a simple baseline model with economic variables only (col-
umn 1 in Table 4.1). As expected, female labor force participation and

E( s =1 [(1 ’ »)s ui,t’s ] · s =1 [(1 ’ »)s ui,t’(N1 +1)’s ]) = 0 since the errors in the ¬rst
13 1 2

square bracket run from ui,t to ui,t’N1 and in the second from ui,t’(N1 +1) to ui,t’(N1 +1)’N2 .

Credible Commitment, Political Institutions, and Social Protection

Table 4.1. Regression Results for Reduction in Inequality (standard errors
in parentheses)

1 2
Inequality 8.94
(5.68) (7.87)
Political-institutional variables:
Partisanship (right) “
Voter turnout “ 0.00
Unionization “
Number of veto points “
Vocational training “
’0.0014——— ’0.0005
Per capita income
(0.0005) (0.0006)
0.73——— 0.39——
Female labor force participation
(0.11) (0.18)
0.81——— 0.90———
(0.27) (0.26)
» .4 .7
0.61 0.73
Adjusted R-squared
47 47
Signi¬cance levels: ——— <.01; —— <.05; — <.10 (two-tailed tests).
Note: All independent variables are measures of the cumulative effect of these
variables between observations on the dependent variable. See regression equa-
tion and text for details.

unemployment are associated with more redistribution. Contrary to
Wagner™s law, higher per capita income slightly reduces redistribution.
With the exception of unemployment, none of these effects are robust
across model speci¬cations.
As in other studies, I also ¬nd that inequality of pretax and transfer
income has a negative effect on redistribution, contrary to the theoretical
expectation of Meltzer and Richard. This negative effect is statistically sig-
ni¬cant at a .01 level, and the substantive impact is also strong; a 1-standard-
deviation increase in inequality is associated with a 0.3-standard-deviation
reduction in redistribution.
Political Foundations of Social Policy

Model 2 introduces ¬ve political-institutional variables: government par-
tisanship, voter turnout, unionization, veto points, and vocational training.
All variables carry the expected sign, and all but voter turnout have statisti-
cally signi¬cant effects. The effects of partisanship and vocational training
are the strongest both substantively and statistically. Thus, a 1-standard-
deviation change in either partisanship or vocational training is associated
with a 0.25-standard-deviation reduction in redistribution.
Another notable change in moving from the baseline model to the full
model is that the effect of inequality reverses (though the positive effect is not
signi¬cant). One likely reason for this change is that left governments (and
strong unions) not only increase redistribution but also reduce inequality.
For example, partisan differences in educational policies are likely to have
an effect on before tax and transfer inequality. If so, excluding partisanship
produces an omitted variable bias on the coef¬cient for inequality.
Such a bias may also be caused by other variables. Experimentation with
including one variable at a time shows that vocational training and the num-
ber of veto points also contribute to the shift in the sign for the inequality
variable. In the case of vocational training, the likely reason is that em-
phasis on speci¬c skills simultaneously produces a more compressed skill
and wage structure and increases electoral support for spending. Similarly,
multiple veto points are likely to impede policies to both redistribute and re-
duce inequality, thereby contributing to a negative sign on inequality when
the veto points variable is excluded. Obviously, these conjectures need to
be substantiated by further empirical analysis. For my purposes, the key re-
sult is the one for partisanship, which is strong and consistent across model
speci¬cations. This ¬nding is largely con¬rmation of previous research,
especially Bradley et al. (in press).
To check the robustness of the results, I also estimated the model using
reduction in the poverty rate instead of reduction in the Gini coef¬cient as
the dependent variable. Redistribution in the poverty rate is the percentage
change in the share of families below 50 percent of the median income, from
before to after taxes and transfers. The results by and large con¬rm those in
Table 4.1. Partisanship and vocational training continue to be the strongest
predictors (and signi¬cant at a .01 level). However, the effect of turnout
is now signi¬cant, while the sign on unionization turns negative and is
borderline signi¬cant. Some of the negative effect of inequality also remains
after inclusion of all controls. Clearly, one must be cautious interpreting
the effect of inequality given how unstable it is across measures and model
Credible Commitment, Political Institutions, and Social Protection Partisanship While government partisanship is important in ex-
plaining redistribution, partisanship itself is a function of coalitional poli-
tics, which is shaped by the electoral system. A key implication of the argu-
ment is that center-left governments tend to dominate over long periods
of time under PR, whereas center-right governments tend to dominate un-
der majoritarianism. Put differently, partisanship is the mechanism through
which electoral systems exert an effect on redistribution.
To test this implication, I use the partisan center of gravity (CoG) index
as a dependent variable and indicators for party and electoral systems as in-
dependent variables. There are data for eighteen countries that have been
democracies since the Second World War, beginning with the ¬rst demo-
cratic election after the war and ending in 1998. One country “ Switzerland “
has a collective executive that prevents coalition politics from having any
in¬‚uence on the composition of the government. I, therefore, exclude this
case from the analysis, although every result reported in this section holds
even when Switzerland is included.
In the theoretical analysis, a distinction was made between majoritarian
two-party systems and proportional multiparty systems. In the former, only
one party can win the election, which determines who forms government,
whereas in the latter no party can form the government without the sup-
port of one or more other parties. The distinction underscores the impor-
tance of whether governments are formed through postelection coalitions
or as direct outcomes of elections. Yet, in practice, the dichotomy is com-
plicated by the fact that voters™ expectations about government formation
affect the partisan distribution of support. Where a single party can rea-
sonably be expected to form a government alone, the model implies that
strategic voting will favor the right and, thus, government composition,
even if the government is ultimately formed as a coalition. One, therefore,
cannot simply look at the number of parties in government at any given
moment in time but must take into account the institutionally mediated
expectations of voters.
There are no direct measures of voter expectations, but we do know the
nature of national electoral systems, which are distinguished in the ¬rst
column of Table 4.2. The strategy is simply to link electoral rules to the
expectation voters can reasonably be assumed to have concerning the na-
ture of the government formation process. With the possible exception of
Austria (because of the strong position of the two main parties), all PR sys-
tems clearly give rise to expectations of governments based on support from
more than one party. This is not the case in any of the non-PR systems,
Political Foundations of Social Policy

Table 4.2. Key Indicators of Party and Electoral Systems

Expectation That
Single-Party Effective
Government Forms Number of Proportionality
Electoral without Need for Legislative of Electoral
System Third-Party Support Parties System
Australia Yes 2.5 0.19
Canada SMP Yes 2.2 0.13
France Yes 3.8 0.16

Ireland Ambiguous 2.8 0.70
Japan Yes 2.7 0.61
New Zealand SMP Yes 2.0 0.00
United SMP Yes 2.1 0.16
United States SMP Yes 1.9 0.39
Average 2.5 0.30

Austria PR Ambiguous 2.4 0.89
Belgium PR No 5.2 0.86
Denmark PR No 4.4 0.96

Finland PR No 5.1 0.87
Germany PR No 2.6 0.91
Italy PR No 4.0 0.91
Netherlands PR No 4.6 1.00
Norway PR No 3.3 0.76
Sweden PR No 3.3 0.90
Average 3.9 0.90
a The use of the single transferable vote in single-member constituencies makes the Australian electoral
system a majority rather than plurality system.
b The two-round run-off system has been in place for most of the postwar period with short interruptions
of PR (1945 until early 1950s and 1986“8).
c The Irish single transferable vote system (STV) is unique. Although sometimes classi¬ed as a PR
system, the low constituency size (¬ve or less) and the strong centripetal incentives for parties in the
system makes it similar to a median voter-dominated SMP system.
d The single nontransferable voting (SNTV) in Japan (until 1994) deviates from SMP in that more than
one candidate is elected from each district, but small district size and nontransferability make it clearly
distinct from PR list systems.

although Australia and Ireland have experienced several instances of coali-
tion governments. Ireland is perhaps the most ambiguous case, but the
inclusion or exclusion of these cases makes little difference to the results.
The division into PR and majoritarian systems is buttressed by quan-
titative measures of party and electoral systems. First, countries with
Credible Commitment, Political Institutions, and Social Protection

Table 4.3. Electoral System and the Number of Years with Left and Right Governments,

Government Partisanship
Proportion of
Left Right Right Governments
Proportional 342 120 0.26
Electoral system (8) (1)
Majoritarian 86 256 0.75
(0) (8)
Note: Excludes centrist governments (see text for details).

majoritarian systems tend to have fewer parties than countries with PR
systems. This is indicated in the third column of Table 4.2 using Laasko
and Taagepera™s (1979) measure of the effective number of parties in par-
liament.14 France is somewhat of an outlier, but at least in presidential
elections the second round of voting in the French run-off system typically
involves only candidates from two parties.
The second quantitative indicator, the proportionality of the electoral
system, is a composite index of two widely used indices of electoral sys-
tem. One is Lijphart™s measure of the effective threshold of representation
based on national election laws. It indicates the actual threshold of electoral
support that a party must get in order to secure representation. The other
is Gallagher™s measure of the disproportionality between votes and seats,
which is an indication of the extent to which smaller parties are being rep-
resented at their full strength. Both indicators were standardized to have a
mean of 0 and a standard deviation of 1 before averaged into an index that
varies from low proportionality (0) to high proportionality (1). The data
are from Lijphart (1994).
The proportionality index is consistent with the division into a majoritar-
ian and a proportional group. There are no cases that should be “switched”
based on their value on the index, although Ireland and Japan have relatively
high scores among the majoritarian countries. Coupled with the other in-
formation in Table 4.2, the dichotomus division of countries into two types,
thus, seems reasonable.
Table 4.3 is a simple cross-tabulation of electoral system and government
partisanship using annual observations as the unit of analysis (the table is

14 The effective number of parties is de¬ned as one divided by the sum of the square root of
the shares of seats held by different parties (or one divided by the Hilferding index).

Political Foundations of Social Policy

identical to Table 1.2, but reproduced here for convenience). Governments
are coded as being left-of-center if their position on the composite left“right
index is to the left of the overall mean. This is somewhat arbitrary because
the mean may not correspond to a centrist position. An alternative would
be to de¬ne the center as the middle of the scale. But in two of the three
expert surveys, the middle of the scale is not explicitly de¬ned as centrist in
terms of a common standard, and experts may well equate it instead with
the observed center of a party system, whether or not this center is shifted
to the left or right. In practice, the choice has little effect on the results.
Identifying a centrist position, however, is important for a different rea-
son. If a CL leadership party in a majoritarian system is centrist, then the
model implies that it stands a good chance of winning. Observing such a
party in government is therefore consistent with the model. At the same
time, it cannot be counted as con¬rmatory evidence because we do not
have any measure to determine whether the party platform is credible. The
relative frequency of center and center-right governments therefore can-
not be hypothesized a priori. Moreover, because the theory implies that
the political space in majoritarian systems is tilted to the right (owing to
strategic voting in a setting of incomplete platform commitment), if we
include governments that are centrist in an absolute sense, these would be
counted as center-left in terms of their relative position. Using a scale such as
the composite CoG index, the results would therefore be biased against the
theory because the center on this scale is almost certainly affected by rel-
ative assessments. The solution is to use one of the component measures
in the CoG index by Castles and Mair to exclude governments that are
centrist in the absolute sense. The Castles-Mair measure is the only one
that explicitly de¬nes the middle value (3) as a party having a centrist left“
right ideology. An alternative strategy of measuring the left“right leanings
of governments against the position of the median legislator is discussed
later in this chapter.
Only one country, Germany, does not conform to the predicted pat-
tern. In this case, there were 34 years with center-right governments and
only 16 years with center-left governments. There is an interesting po-
tential explanation for this. If a center-right party can appeal to voters on
grounds other than class, especially religion, and thereby capture some sup-
port among middle- and lower-income people, such a party can credibly
claim to be closer to the center than when it represents only high-income
voters. This makes the party a more attractive coalition partner for center
parties. This could be why the small pivotal liberal party (FDP) for most
Credible Commitment, Political Institutions, and Social Protection

of the postwar period chose to ally with the Christian Democratic Party
(CDU/CSU) instead of the Social Democratic Party (SPD). The result-
ing equilibrium is still heavily in¬‚uenced by PR because the right must be
moderate and prepared to accept some redistribution. The space is thus
shifted to the left compared to majoritarian systems. But it does produce
less redistribution than in a typical PR system based entirely on class.
Germany aside, it can be objected to the evidence in Table 4.3 that it
does not take into account that the left“right balance of governments is also
affected by the left“right balance of power in the legislature. It may be that
whole electorates are shifted to the left or right for reasons that are outside
the model and that this, not party politics, explains why the composition
of the government varies across countries. Note again, however, that the
theoretical model implies strategic voting in majoritarian systems that does
shift the legislative center to the right. Also, the distribution of seats in PR
systems does not matter in principle as long as coalitions that are either to
the left or to the right of the center can be formed. Still, we cannot exclude
the possibility that the distribution of prestrategic voter preferences differs
across countries and affects government partisanship.
To test the possibility, I computed the position of the median legislator,
de¬ned as the left“right location of the median legislative party. If poli-
cies were decided by majority voting, not bargaining, the preference of the
median party would be the Condorcet winner. I therefore calculated the
difference between the government position, as measured by the CoG in-
dex, and the position of the median party in the legislature. Governments
to the left of the median are called center-left; governments to the right,
center-right. This procedure, however, does not work for single-party gov-
ernments with a majority in the legislature. If the Conservatives in Britain
control government power, for example, they will also command a majority
in the lower house. In such instances, the median in the legislature will
be identical to the government™s position. Yet, clearly, the government can
still sensibly be counted as either left or right. In these cases, I therefore
compared the government position to the average of party position, in the
legislature, weighted by the parties™ share of seats. The results are reported
in Table 4.4
As one would expect, the results are somewhat weaker than in Table 4.1,
yet they are entirely consistent with these. About two thirds of governments
under PR are to the left of the legislative median, whereas two thirds of
governments under majoritarian institutions are to the right. All but one
country conforms to this pattern. The “outlier” is no longer Germany
Political Foundations of Social Policy

Table 4.4. Electoral System and the Number of Years with Governments Farther to the
Left or to the Right Than the Median Legislator, 1945“1998

Government Partisanship
Proportion of
Left Right Right Governments
Proportional 291 171 0.37
Electoral system (9) (0)
Majoritarian 116 226 0.66
(1) (7)
Note: Excludes governments coded as centrist on the Castles-Mair scale.

because most governments were in fact to the left of the median, even as
they tended to be right compared to other PR systems. Instead, the deviant
case is France where twenty-nine of ¬fty-two observations are to the left
of the legislative median. But this is hardly discon¬rming evidence since
the government CoG is based on the distribution of cabinet portfolios,
which in the French system is decided through coalition bargaining between
parties in the National Assembly. Clearly, such a system does not preclude
governments forming to the left of the median “ indeed, we should expect
it. The French president, on the other hand, is chosen through a two-
candidate contest (barring an outright majority in the ¬rst round), and here
the argument implies that the center-right will win more of the time. This
has indeed been the case (the sole exception is the Socialist Presidency of
Mitterran). Hence, if the position of the presidential party were used instead
of the CoG cabinet measure, the results would clearly support the argument.
Also note that because French political parties take policy positions that are
at least partly adopted with an eye to winning the presidency, the rightist
bias in the presidential election will also tilt the legislature to the right. This
is why France conforms to the predicted when we use the absolute CoG
measure (Table 4.1).
What alternative explanations might there be for the pattern observed in
Table 4.4? Because I use the difference between the position of the govern-
ment and the median legislator, I have limited such alternatives to variables
that affect the postelection partisan composition of governments. I thus im-
plicitly “control” for all variables that may affect the left“right balance in the
legislature. Although there are obviously a plethora of situationally speci¬c
factors that shape each instance of government formation, variables that
would systematically bias the composition of governments in one ideological
direction or the other are in fact not easy to think of.
Credible Commitment, Political Institutions, and Social Protection

One important exception is the extent of party fractionalization on ei-
ther side of the center. Where the left (right) is relatively more divided than
the right (left), we would expect government formation between left (right)
parties to be more complicated under PR rules. Similarly, as argued by
Powell (2002), we would expect such fragmentation to produce more elec-
toral defeats under majoritarian rules. If so, this could confound the rela-
tionship between electoral system and government partisanship. Specif-
ically, Rokkan (1970) and Boix (1999) have argued that at the time of
the extension of the franchise, when a united right faced a rising but di-
vided left, the governing right chose to retain majoritarian institutions.
Conversely, when a divided right faced a rising and united left, the re-
sponse was to opt for PR. If this pattern of fractionalization persisted in
the postwar period, the right would tend to have an advantage in majori-
tarian systems, whereas the left would tend to have an advantage under
PR. This is precisely the pattern that our model predicts, but for different
Testing this alternative requires us to use multiple regression. Because
there is little meaningful variance in electoral systems over time, I ran a sim-
ple cross-sectional regression on the averages from 1950 to 1996 (for which
complete data exist on several control variables). The results are shown in
Table 4.5. The fractionalization variable is the difference between party
fractionalization on the left and right, where fractionalization is de¬ned as
one minus the sum of the squared seat shares held by parties to the left or
to the right of center (Rae 1967).
As expected, both electoral system (PR) and fractionalization on the
left signi¬cantly reduce the likelihood of getting a government that is to
the left of the legislative median, and both effects remain when the vari-
ables are entered simultaneously (column 3). In substantive terms, going
from a majoritarian to a PR system reduces the predicted center of grav-
ity of the government by 0.15 (relative to the legislative median). This
difference is roughly equivalent to the difference between a typical social
democratic and a typical Christian democratic government, or between
the latter and a typical conservative government. Another way to convey
the ¬nding is that the effect is roughly equal to a one standard devia-
tion on the government CoG variable “ a large impact by any standard.
The effect of a one-standard-deviation increase in left fractionalization is
about one third of this effect. Note also that including both variables si-
multaneously does not notably affect the estimated parameter for electoral
Political Foundations of Social Policy

Table 4.5. Regression Results for Government Partisanship, 1950“1996
(standard errors in parentheses)

(1) (2) (3) (4) (5)
Government Government Government
CoG Minus CoG Minus CoG Minus
Legislative Legislative Legislative Government Government
Median Median Median CoG CoG
0.653——— 0.593——— 0.664——— 0.163——— ’0.085
(0.039) (0.031) (0.033) (0.018) (0.529)
’0.173——— ’0.147——— ’0.202——— ’0.242———

Electoral system
(PR) (0.054) (0.047) (0.050) (0.086)
0.303—— ’0.241——

Fragmentation 0.159 0.011
(left minus right) (0.115) (0.094) (0.099) (0.171)
’ ’ ’ ’
Electoral 0.007
participation (0.006)
’ ’ ’ ’ ’0.005
’ ’ ’ ’
Female labor force 0.007
participation (0.005)
0.37 0.27 0.54 0.56 0.54
Adjusted R-squared
17 17 17 17 17
——— < .01; —— < .05; — < .10 (two-tailed tests).
Signi¬cance levels:

In column (4), I use the absolute government CoG measure as the depen-
dent variable. The effect of left fractionalization on this measure is weaker
and no longer statistically signi¬cant. The effect of electoral system, how-
ever, increases, mirroring the observed differences in strength between the
results presented in columns (1) and (4). This continues to be the case if we
control for other variables that may reasonably be expected to affect the po-
litical center (column 5). Predictably, high unionization rates are associated
with most left-leaning governments, but the effect is weak and statistically
insigni¬cant. Electoral participation and female labor force participation do
not have the effects one might have predicted, but, again, the coef¬cients
are not signi¬cantly different from zero. The only strong predictor contin-
ues to be the electoral system, and the fact that the effect is stronger than
when using the position of the government relative to the median suggests
that the electoral system affects not only government formation but also
the center of the political space.

Credible Commitment, Political Institutions, and Social Protection

4.6. Conclusion
This chapter has explored an overlooked problem in the politics of social
policy making: The inability of current voters to commit future voters to
particular policies, even when these policies are in their common inter-
est. This time-inconsistency problem is particularly notable in the area of
social insurance because those who are pivotal in electoral politics (“the
median voters”) tend to be employed and unlikely to bene¬t from social
spending, except at some time in the future when they will no longer be
pivotal. The implication of this problem is a serious underprovision of so-
cial protection compared to (long-term) voter preferences. And that in turn
undermines the institutional foundations for speci¬c skills systems, includ-
ing coordinated wage bargaining and well-functioning vocational training
There are two related solutions to the problem. One is disciplined and
responsible parties with close ties to private groups representing workers
with speci¬c skills. If parties offer a set of public goods that cannot be pro-
vided ef¬ciently without support from private actors “ a vocational train-
ing system, for example, requires information and sponsorship by unions
and employer associations to work ef¬ciently, and incomes policies can-
not succeed without the cooperation of workers and their unions “ the
groups whose cooperation is required can gain in¬‚uence over policy. In
the institutionalized party model, resources come in the form of “fees”
to political parties, broadly conceived, that are exchanged for in¬‚uence
over party platforms and leadership selection. When the incentives of party
leaders to defect are not too great, as may be the case in winner-take-all
electoral systems, party organization can alleviate the time-inconsistency
The second institutional mechanism is electoral systems and the dis-
tributive politics to which they give rise. Redistributive transfers not only
shift income between groups but also provide insurance against income
loss in the event of unemployment, sickness, and so on (Moene and Waller-
stein 2001). Insofar as there is a strategic complementarity between such
insurance and individuals™ decisions to invest in particular types of skills,
the ability of the government to credibly commit to redistributive spend-
ing serves as insurance against the loss of income when speci¬c skills are
rendered obsolete by technological and other forces of change. Because
PR promotes left party dominance and redistribution, it serves as a key

Political Foundations of Social Policy

commitment mechanism in political economies that depend on workers
making heavy investments in highly speci¬c skills. It is probably no ac-
cident, therefore, that PR and vocational training are highly correlated
(r = .7).
The key insight of this analysis is that political institutions “ the electoral
system and political party organizations “ support, and are being supported
by, economic as well as social institutions. Although the historical inter-
play between these institutions is a major topic for future research, the
clustering into two institutional con¬gurations “ one characterized by PR,
disciplined and responsible parties, generous welfare states, and speci¬c
skills production and the other by majoritarianism, leadership-dominated
parties, small welfare states, and general skills “ should not come as a sur-
prise. They re¬‚ect, in Hall and Soskice™s (2001), conceptional framework,
extensive institutional complementarities.

Appendix 4.A: Mathematical Proofs for the Party
Institutionalization Model

De¬nition of Utility Functions
The Group Utility Function Given that platforms are chosen behind the
veil, if the employed group member receives a pretax income of 1, then
the expected utility of an individual is at „ = 0 before the individual knows
whether or not he is unemployed. Compared to the group utility function
there is no probability of unemployment.

δ „ (1 ’ p)·log(1 ’ t„ + ±(s ·a „ ’ c (a „ )))
V0 ≡
„ =0
+ p ·log(t„ ·(1 ’ p)/ p + ±(s ·a „ ’ c (a „ )) ’ γ (s )) (A4.1)
where δ ∈ [0, 1] is the discount factor and where u(x) is the utility derived
from net receipts (income or unemployment bene¬t), as before. The neg-
ative skill-speci¬city contingent cost γ (s ) is incurred in the event of unem-
ployment because it is harder for those with more speci¬c skills to ¬nd jobs
that are suitable to their skills. γ (s ), in other words, can be conceived as a
search cost.
Credible Commitment, Political Institutions, and Social Protection

The Median Voter Utility Function The preferences of s m , the median
employed voter, after the veil has been raised, are given by

Vs m = log(1 ’ t + ±(s m a ’ c )) (A4.2)

where c is whatever contributions the median voter has already paid to one
or the other of the two parties; since that has already been paid, it will not
in¬‚uence his or her vote.

The Condition for Party Leader Cooperation in Stage ii of Game
The condition for party leader cooperation is derived as follows. The base
can observe if the leader sets the nominated platform tL or not, but the
leader may not always have the power to set tL and may instead be forced
to set a lower tax rate of tL ’ µ (because of, say, an unexpected recession). If
the base observes the platform tL ’ µ , it does not know whether the leader
has been forced to choose this platform or whether the leader has done it
deliberately to undercut the other party™s platform and win the election.
Concretely assume there is a random variable µi = {0, µ}, i = L, H;
only the leader i knows the value of µi . A leader then has two possible
strategies: Cooperate with the party base by always setting t = ti ’ µi or
defect by setting t = ti ’ µ. Assuming the other leader opts for a coopera-
tive strategy, how high does 1 ’ q (the probability of non-reelection) have to
be to ensure that leader i cooperates? The answer is given by the following
r · (1 ’ ») B(s )/2
(1 ’ q ) ≥ · (A4.3)
» B(s )/2 + L(s )
where r is the probability that the random variables take the values (0, 0).
From this follows the substantive interpretation of the condition provided
in the body of the text.

Proof of (A4.3). Let V be the present value of the maximum utility for a
reselected politician at the start of the current period. Because the game is
stationary, we have

B ·[q 2 /2 + q ·(1 ’ q ) + (1 ’ q )2 /2] + L + »V
≥ B ·[q 2 + q ·(1 ’ q )/2 + q ·(1 ’ q ) + (1 ’ q )2 /2] + L + »γ V
Political Foundations of Social Policy

The left-hand side is the present value of cooperation: q 2 is the probability
that the random variables take the values (0, 0), and 1 is the probability
of the leader winning in that case, and so on. The right-hand side is the
present value of defection: Thus, when the random variables take the values
(0, 0) with probability q 2 , the leader sets the platform t = ti ’ µ and thus
wins with probability 1 and so on. In both cases, the other leader pursues the
cooperative strategy. Because the square-bracketed term on the left-hand
side is 1 and that on the right-hand side is (1 + q )/2, this implies

»·(1 ’ γ )V ≥ q B/2 (A4.5a)
We also have from the left-hand side
V = B/2 + L + »V = (B/2 + L) /(1 ’ ») (A4.5b)
which together imply Equation (A4.3). q.e.d.
The relation between s and the probability of dismissal can likewise be
deduced from this equation. We have
‚ ln B ‚ ln L ‚(1 ’ γ )
> ” >0 (A4.6)
‚ ln s ‚ ln s ‚s
which is assumed to hold in (A4.3). Thus, as the level of speci¬c assets of
the median voter in the party base grows, so does the leader™s punishment
for not following the party platform. This is the key behavioral mechanism
in this stage of the game because it implies that leaders in H parties will go
along with higher rates of taxation than leaders in L parties. Given that the
leader incentives to defect are known to the groups who propose platforms,
it is assumed in the following that the platforms chosen by the groups satisfy
the constraint in (A4.3).15

Derivation of Equation (4.5) in Stage i of Game
Assuming that t„ and a „ have the same values ∀„ = 0, . . . , ∞, we can ¬nd
the utility-maximizing values, t — and a — , from Equation (4.4). For simplicity,
c (a) ≡ ’a 2 /2 (A4.7)

15 In a more complete model, institutionalization would be determined endogenously as the
outcome of a bargaining game over a and (1 ’ γ ). Yet, all that is needed for the logic of the
argument to go through is that (1 ’ γ ) is a rising function of membership fees, c(a), which
are in turn rising in s. As long as (1 ’ γ ) is valuable to the group, and a is valuable to the
party, this condition will hold in any reasonable bargaining game.

Credible Commitment, Political Institutions, and Social Protection

which yields the following optimal tax rate:

‚ V0 1’ p
‚t 1 ’ t + ± (s a ’ a 2 /2)
1’ p
+ · =0
(1 ’ p)t/ p + ± (s a ’ a 2 /2) ’ γ (s ) p
’ p · (1 ’ t) = (1 ’ p) · t ’ γ (s )
’ t — (s ) = p · (1 + γ (s )) (A4.8)

which is equivalent to Equation (4.5).
To ¬nd a — , it is assumed, without serious loss of generality, that γ (s ) =
s /2. It follows that

‚ V0
= s ’ a = 0 ’ a — (s ) = s (A4.9)
so that the preferred level of a rises in s .
The resulting equilibrium is robust to the possibility of third-party entry
as shown later. However, depending on the values of sML and sMH compared
to the median voter, the median voter may always prefer one party over
the other. For the losing party to avoid this, it would have to adjust the
nominated platform so that it had an equal chance of winning. Such a “con-
tingency” requirement would still produce a subgame perfect equilibrium.
More importantly, this is a problem that arises only in majoritarian systems,
and it is not endemic to the time-inconsistency problem.16

Proof That No Third Party Will Enter
The model equilibrium of the model is robust to the possibility of third-
party entry. Such entry will happen if the entrant can win outright. The
median voter will prefer a new entrant with a platform of t = a = 0 to the
platform of the L or H parties unless:

log(1 ’ c ) ¤ log(1 ’ p ·(1 + s ML /2) + ±s M s ML ’ c )

where the left-hand side of the inequality is the payoff from the new entrant™s
platform with a zero tax rate and level of a, and the right-hand side is the

16 Alternatively, one can allow uncertainty about the election outcome. If parties seek to
maximize expected policy outcomes, party platforms will diverge (Whitman 1973).

Political Foundations of Social Policy

payoff from the platform of the L party. This inequality implies that there
is a level of sM below which a new entrant will enter; this is given implicitly

±L + 2 L
± ’1
sM ’ sM > p’
±4 8

Taking a Taylor expansion of the left-hand side around implies

1 8p ’ L
sM > +
2 ±L + 2

This says that (i) the larger p is, the higher the tax rate will be and, therefore,
the greater sM will have to be for this to be compensated by the public good
(the value of which increases with sM ; (ii) the larger ± is, the smaller sM can
be consistently with the new entrant being kept out because the greater is
the value of the public good. This is a critical condition because its failure
to hold means that an institutional party system “ even with relatively low
degrees of institutionalization “ will cease to be viable. Still, it is clear that it
is perfectly possible to have two parties that set tax rates above zero, despite
the ability of new entrants to enter and do so. Under many circumstances,
if they did enter, they would not win.

Appendix 4.B: Proof That the Win-set of m* Is Empty
Refer to Figure A4.1. The transfer g is on the horizontal axis, and the
tax t is on the vertical axis. The indifference curves for L and H are drawn
through m— = {g, t} = {0, 0.5}, the median voter™s ideal point. The rel-
evant indifference curve of L, u L (m— ), is downward sloping with a gra-
dient of ’1. That of H, u H (m— ), is downward sloping with a gradient
of ’(1 + ±)(1 ’ µ)· u H (m— ) that is steeper than u L (m— ) if ± > µ/(1 ’ µ),
which is assumed to be the case. Utility for L improves in a northeasterly
direction with increasing g and t; for H the opposite is the case. Thus, it
can be seen that the LH win-set of m— is empty so that no alternative plat-
form will attract the votes for both L and H. Hence, m— is the Condorcet
Credible Commitment, Political Institutions, and Social Protection

Figure A4.1 The indifference curves for L and H and the empty LH win-set of m— .

Appendix 4.C: Rubinstein Bargaining Solution for LM
and MH Coalitions

LM Coalition
The Rubinstein solution is derived in the absence of outside options. The
bargaining over t ranges from 1 to 1 and the bargain over g ranges from 0
to g — . The normalized utility functions for L and M can be written as

u L = u L (t, g) ’ u L (.5, 0) = t + g ’ 0.5
u M = u M (t) ’ u M (1) = ’|t ’ 0.5| + |0.5|
Political Foundations of Social Policy

In the following proof it is assumed that µ = 0, but the result holds for
small enough µ.17 Two conditions need to be satis¬ed in a multidimensional
bargain (Kreps, 1990, p. 561, Proposition 2): First, L™s offer to M must be
worth at least as much to M now as M™s offer to L next period will be worth
to M now:

um(t L ) = δ· um(t M ) (A4.10)

This implies

’ |t L ’ 0.5| + 0.5 = ’δ|t M ’ 0.5| + 0.5δ
’ t L = (1 ’ δ) + δt M

And, second, M™s offer to L must be worth at least as much to L now as L™s
offer to M next period will be worth to L now:

u L (t M , g M ) = δ · u L (t L , g L ) (A4.11)

which implies

t M + g M ’ 0.5 = δ · (t L + g M ’ 0.5)

Solving for t M in terms of g M and g L gives

1 + 0.5δ (δg M ’ δ 2 g L )
= ’
1+δ (1 ’ δ)·(1 + δ)

As δ ’ 1, so the difference between ¬rst and second mover offers goes to
zero, so that

t = 0.75 ’ g/2 (A4.12)

(A4.12) is a necessary condition for the unique subgame perfect equilibrium
(SGPE) of this bargaining game. If (A4.12) is substituted into u M and u L
so that both are functions of g alone, the assumption of Pareto optimality
implies that g is maximized so that

tLM = 0.75 ’ g — /2

Follow the proof through using u M = ’|t ’ 0.5| + |1 ’ 0.5| + (1 + ±)µ(g — ’ g). This gen-
17 ˆ
erates the necessary condition t = 0.75 ’ g/2 + (1 + ±)·µ·(g — ’ g)/2. The condition
for ‚ u M /‚g > 0 is µ < 1/(1 + ±) .

Credible Commitment, Political Institutions, and Social Protection

MH Coalition
Bargaining is over t in the range [0, 0.5]. It is in the common interest
of both parties to agree on g = 0 . The normalized utility functions are
u M = ’ |t ’ 0.5| and u H = ’t. The conditions for a SGPE are

’ |t H ’ 0.5| = ’δ|t M ’ 0.5| (A4.13)


’ t M = ’δt H (A4.14)

and these imply

t H = 0.5/(1 + δ) (A4.15)

or as δ ’ 1, tH M ’ 0.25.
Can H break an LM coalition by offering M a deal that is closer to
M™s ideal policy? Figure A4.2 shows the argument that it cannot as an ex-
tensive game with complete and perfect information. Without serious loss

Figure A4.2 The structure of the coalition game.

Political Foundations of Social Policy

of generality, M is charged with coalition formation (the decision node at
the top of the game) and can either choose L or H to enter into coalition
bargaining.18 Suppose M chooses L. At that point, a Rubinstein-type al-
ternative offers in¬nite-move bargaining, and the subgame begins. During
the subgame, there is a discount factor, δ s , attached to the payoff after s
bargaining rounds. Without signi¬cant loss of generality, it is assumed that
the party to whom the proposal is made has the ¬rst move. Thus, we are
at the top-left L decision node. It is assumed that g = g — as part of the ML
bargaining19 so that the bargaining subgame entails alternating offers of
the tax rate. Starting with L™s ¬rst move, the closed interval of possible tax
rates, t µ [0, 1], is given by the base line of the triangle at the apex, which
is L™s decision node. L™s choice of a tax rate offer is indicated by a line from
L™s decision node to the base of the triangle. M now has the move and three
alternatives: (1) to accept L™s offer, in which case the game ends and an ML
coalition is established with g = g — and the tax rate being that offered by L;
(2) to reject L™s offer and to make a counteroffer to L (the line from M ™s
decision node down to the base of the triangle); or (3) to break off negoti-
ations with L and enter into negotiations with H. If (2), L can then choose
whether to accept M ™s offer, so that the game ends, or to reject it and make
a counteroffer. M again has the threefold options of acceptance, of rejec-
tion and making a new offer, or of breaking off negotiations with L. And
so on. It is assumed that whenever the game ends with the establishment
of a coalition, a discount factor δ S is applied to the utility of the parties
where S is the number of bargaining rounds that have elapsed.
It is further assumed, realistically, that if M enters into and then breaks
a coalition, M incurs a cost of C > 0. This cost can be interpreted as a
transaction cost of negotiations after they have started, or it can be in-
terpreted as the cost of breaking up a coalition government after it has
formed. The key is that negotiators can anticipate that M cannot defect
with impunity from a coalition that it has agreed to enter into. Coalition
breakups are accompanied by discord and put on public display the inabil-
ity of parties to bargain and govern effectively. Whatever the magnitude
of the cost, any positive cost of breaking off negotiations will prevent H
from underbidding a coalition of L and M that is based on a Rubinstein

18 Again for simplicity, it is assumed that parties do not reject an offer of coalition bargaining.
19 Which implies from the solution to the Rubinstein subgame that the equilibrium tax rate
will be lower (in M ™s favor).

Credible Commitment, Political Institutions, and Social Protection

For simplicity, also assume that M can only break off negotiations once.
If M, for instance, breaks off negotiations with L, then M must continue
bargaining with H until a coalition agreement has been reached. In fact,
the results go through in a model in which M can break off negotiations
an in¬nite number of times, so long as C is strictly positive and incurred
on each break-off situation. The proof is available from the authors on
The subgame perfect equilibrium can be worked out through backward
1. The SGPE solution to the bargaining subgame between M and L,
absent M ™s outside option of breaking off negotiations and switching
to negotiate with H, is for L at its ¬rst move to offer t = 0.75 ’ g — /2 to
M and for M to accept this offer, as δ goes to unity. This is the standard
Rubinstein result (see Appendix 4.A).
2. Similarly, the SGPE solution to the bargaining subgame between M
and H as a result of M breaking with L is t = 0.25, as δ goes to unity.
3. The SGPE of the bargaining subgame between M and L, including
M ™s outside option to switch to bargaining with H is the same as the
solution without the outside option (i.e., that in step 1). This follows
from a minor modi¬cation of Proposition 5.1 of Muthoo [1999]: If
the value to M of the outside option is less than the value of the
subgame without it, the outside option is irrelevant.
4. The SGPE of the bargaining subgame between M and H, includ-
ing M™s outside option to switch to bargaining with L requires us to
evaluate the payoff to M if M responds to an offer by H by breaking
negotiations and switching to L. From step 1, the outcome of the sub-
sequent bargaining subgame between M and L is tML = 0.75 ’ g — /2.
However, this result can now be incurred only at a cost of C. H will
therefore offer M a deal that is worse than 0.75 tML = 0.75 ’ g — /2 by
an amount equal to C, and M will accept this deal.
In combination, steps 1“4 imply that M chooses initially to negoti-
ate with L. This is because an initial negotiation with L results in
u M (tML , g — ) where tML = 0.75 ’ g — /2 (from step 1), but an initial negoti-
ation with H results in u M (tML , g — ) ’ C (from step 4). Hence, an ML coali-
tion will result with tML = .75 ’ g — /2 and g = g — . The intuition is simple;
because g falls disproportionately on H, M can use g as a bargaining chip
in negotiations with L to get a better deal. Although H would then have
an incentive to offer M an even better deal, after an HM coalition had
Political Foundations of Social Policy

formed, H would have an incentive to exploit its bargaining power to the
point where the deal offered M was exactly the same as the one M could
get from L minus the cost to M of breaking up the coalition. H, thus, faces a
time-inconsistency problem, and M will, therefore, prefer L as the coalition

Appendix 4.D: The Measurement of Party System

Centralized Parties
Carey and Shugart (1995) analyzed the conditions under which parties
can control their candidates. Con¬rming a long-standing intuition among
students of political parties, the incentives for politicians to campaign on
the party platform is critically dependent on the ability of parties™ to control
politicians™ reelection chances.
The best known means to accomplish such control is a closed party
list system where a candidate™s rank on the list determines that candidate™s
likelihood of reelection. In closed list systems like the Norwegian or the
Swedish, failure to adhere to the party platform during or after the election
can severely curtail a politician™s reelection chances. In open list systems
like the Finnish or the Dutch, the party controls who gets on the list, but
voters can choose among the candidates on the list, thereby reducing the
party™s control over who gets elected. This furnishes politicians with a rea-
son to take advantage of electoral opportunities even when these require
them to deviate from the party platform. This problem, however, is mag-
ni¬ed in systems with primaries, such as elections to the U.S. Congress,
because political parties do not control who gets on the ballot. Politicians,
thus, have incentives to run their campaigns with little regard for the party
platform, although after the primaries are over and the candidates face an
opponent from another party, the party label still carries some value. In
an extreme electoral system like that of the Japanese before the reforms
in 1994, even this incentive to use the party label is dissipated because an
open nomination process is coupled with elections where candidates from
the same party compete against each other for a single nontransferable
Credible Commitment, Political Institutions, and Social Protection

Table A4.1. Electoral Systems and Incentives of Politicians to Campaign on the Party Platform

Incentives to
Campaign on Party
Electoral System Countries Platform (rank score)
Closed list Norway, Spain, Sweden 6
Flexible lists Denmark, Germany, Greece, 5
Italy, Austria, Belgium,
Single member district plurality Britain, Canada, New Zealand 4
Single member district majority France 4
with run-off
Open lists Finland, Netherlands 4
Single transferable vote, party Australia, Ireland 3
Primary system United States 2
Single nontransferable vote, Japan (pre-1993) 1
open endorsement
Source: Adapted from Carey and Shugart (1995).

vote. Policy differentiation, not coordination, within the party is the
Table A4.1 shows the classi¬cation of electoral systems according to
the incentives of politicians to campaign on the platform of their parties.
Higher numbers indicate greater incentives to toe the party line. The classi-
¬cation follows Carey and Shugart™s, with a few quali¬cations. First, Carey
and Shugart only distinguish between open and closed list systems, but as
Cox (1997) points out, many countries have “¬‚exible” list systems where
the voter can cast a vote for both individual candidates and for the list
as a whole. In these systems, voters have some capacity to “break” the
list, but the party retains considerable control over who receives the list
votes and hence who gets elected (Cox 1997, p. 61). In practice, ¬‚exible
list systems function very similarly to closed list systems, and they have
been ranked just below closed list systems in terms of the incentives they
create for politicians to campaign on a party platform as opposed to other
Second, it is not clear that open lists can be unambiguously ranked
below single member plurality systems as Carey and Shugart (1995) do.
Their justi¬cation for doing so is that the party in single member plurality
(SMP) systems controls the list, whereas in open list systems the voter has
discretion over who on the list is picked. Yet, in SMP systems the total
Political Foundations of Social Policy

number of listed candidates is large “ as large as the number of seats in
the legislature if a party ¬elds candidates in each district “ and each candi-
date must appeal to local constituencies where strict adherence to the party
platform is often not conducive to electoral success. By contrast, larger dis-
tricts in open list systems create greater competition for a smaller number
of nominations, thereby increasing the leverage of the party leadership. On
the other hand, open list systems generate competition between candidates
from a single party that are absent under SMP voting. It is unclear which of
these effects dominates, and the two systems are, therefore, ranked similarly.
The same has been done for run-off majority systems (France) because, as
Carey and Shugart acknowledge, the incentives are only marginally dif-
ferent from SMP systems. The rest of the classi¬cation follows Carey and
Shugart™s scheme.

Turning to corporatism, by and large Schmitter™s conceptualization is used,
which includes (a) the capacity of interests groups to aggregate and artic-
ulate demands on behalf of their members and to implement policy com-
mitments (“intermediation”), and (b) the extent to which there is coordi-
nation of demands between groups and political parties (“concertation”).
In the theoretical model, corporatist intermediation refers to the capacity
of groups to commit resources to parties, while concertation concerns the
potential in¬‚uence over party platforms. Speci¬cally, a recent composite in-
dex developed by Siaroff (1998), which Lijphart (1999) considers to be the
most encompassing in terms of my concerns with policy in¬‚uence and ca-
pacity for implementation, is used. The index values are listed in Table A4.2
along with the electoral system numbers. To get a composite measure of
party institutionalization, we simply add the two measures together after
standardization (column 3).

Credible Commitment, Political Institutions, and Social Protection

Table A4.2. Political Institutions and Capacity for Commitment

(1) (2) (3)
Corporatism Incentives to Campaign Institutional Capacity
for Commitmenta
Index on Party Platform
Norway 4.6 6.0 1.5
Sweden 4.5 6.0 1.5
Austria 4.4 5.0 1.4
Germany 3.6 5.0 1.3
Switzerland 4.0 5.0 1.3
Belgium 3.8 5.0 1.3
Denmark 3.9 5.0 1.3
Netherlands 3.8 4.0 1.2
Finland 4.0 4.0 1.2
Spain 1.8 6.0 1.1
Italy 2.0 5.0 1.0
France 2.0 4.0 0.9
United Kingdom 1.5 4.0 0.9
Canada 1.5 4.0 0.8
New Zealand 1.9 4.0 0.8
Australia 2.4 3.0 0.8
Ireland 2.1 3.0 0.7
Japan 3.8 1.0 0.7
United States 1.9 2.0 0.5
a Calculated as the sum of (1) and (2) after standardization.

Appendix 4E. Summary Statistics for Section 4.5.2

Table A4.3. Country Means for Variables Used in Regression Analysis
Redistribution Effective Per Capita Labor Manufacturing
(% reduction Inequality Partisanship Voter Vocational Number Income Force Work
in Gini (p90/p50 (left“right Turnout Veto Training Electoral of (1985 Participation Unemployment Force
coef¬cient) ratio) CoG index) (%) Unionization Points (%) System Parties Fragmentation dollars) (%) (%) (%)

Australia 23.97 1.70 0.59 84 46 3 0.9 0.20 2.5 10,909 46 4.63 21
Austria 0.37 87 54 0.90 2.4 8,311 51 2.76 26
’ ’ ’ ’ ’0.18
Belgium 35.56 1.64 0.46 88 48 1 56.3 0.87 5.2 8,949 43 7.89 23
Canada 21.26 1.82 0.45 68 30 2 4.6 0.14 2.2 0.18 11,670 48 6.91 18
Denmark 37.89 1.58 0.44 84 67 0 31.8 0.96 4.4 9,982 63 6.83 24
Finland 35.17 1.68 0.37 79 53 1 32.9 0.87 5.1 8,661 66 4.48 23
France 25.36 1.94 0.50 66 18 1 27.9 0.18 3.8 0.10 9,485 51 4.57 23
Germany 18.70 1.70 0.49 81 34 4 34.9 0.91 2.6 9,729 51 4.86 29
Ireland 0.53 75 48 0.70 2.8 5,807 37 9.09 16
’ ’ ’ ’ ’0.33
Japan 0.98 71 31 0.61 2.6 0.22 7,918 56 1.77 23
’ ’ ’ ’
Italy 12.13 1.63 0.46 93 34 1 37.5 0.92 3.8 0.20 7,777 38 8.12 20
Netherlands 30.59 1.64 0.38 85 33 1 42.5 1.00 4.6 0.18 9,269 35 4.62 20
New Zealand 0.54 0 37.4 2.0
’ ’ ’ ’ ’0.00 ’0.40 ’ ’ ’ ’
Norway 27.52 1.50 0.19 80 54 0.77 3.3 9,863 52 2.28 22
’ ’ ’0.02
Sweden 37.89 1.58 0.44 84 67 0 33.2 0.96 4.4 9,982 63 6.83 24
Switzerland 8.84 1.68 0.32 35 32 6 35.5 0.86 5.3 0.16 12,377 53 0.78 32
United 22.67 1.78 0.65 76 42 0 10.4 0.17 2.1 0.08 9,282 54 5.01 28
United States 17.60 2.07 0.50 56 23 5 2.9 0.39 1.9 0.00 13,651 53 5.74 20

Note: Time coverage is 1950“96 except for redistribution and inequality, which are restricted to the LIS observations. Excludes Switzerland.
Table A4.4. Correlation Matrix

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
(1) Redistribution 1.00
(2) Inequality 1.00
(3) Partisanship 0.48 1.00
(4) Voter turnout 0.32 1.00
’0.76 ’0.05
(5) Unionization 0.67 0.49 1.00
’0.71 ’0.42
(6) Electoral system 0.38 0.31 0.47 1.00
’0.70 ’0.42
(7) Effective number of parties 0.56 0.18 0.57 1.00
’0.66 ’0.39 ’0.03
(8) Left fragmentation 0.27 1.00
’ ’ ’0.40 ’0.78 ’0.24 ’0.07
(9) Number of veto points 0.64 0.45 0.58 1.00
’0.55 ’0.56 ’0.53 ’0.27 ’
(10) Vocational training 0.32 0.52 0.23 0.79 1
’0.6 ’0.46 ’0.83 ’ ’0.29
(11) Per capita income 0.39 0.10 0.53 1.00
’0.20 ’0.27 ’0.56 ’0.21 ’0.30 ’0.08 ’0.43
(12) Female labor force part 0.51 0.41 0.03 0.05 0.31 1.00
’0.22 ’0.07 ’0.19 ’0.26 ’0.31 ’0.12
(13) Unemployment 0.09 0.11 0.38 0.15 1.00
’0.11 ’0.04 ’0.09 ’0.03 ’0.21 ’0.04 ’0.31 ’0.49
(14) Manufacturing work force 0.12 0.22 0.16 0.25 0.50
’ ’ ’0.11 ’0.30 ’0.00 ’ ’ ’0.63
Note: Correlations based on period averages.


Forces of Change

Coping with Risk

One of the most remarkable facts about welfare spending in advanced
democracies is its rapid and almost uninterrupted expansion since the 1950s.
Figure 5.1 illustrates this expansion for two broad categories of spending:
government consumption of services and government transfers, both ex-
pressed as shares of GDP. Although the very rapid expansion beginning in
the mid-1960s slowed down in the 1980s, and the ¬scal retrenchment asso-
ciated with the reining in of public de¬cits in the late 1980s seems to have
caused a temporary reduction in public consumption, there are no signs
of any broad-scale retrenchment. This continued growth of the welfare
state presents an intriguing puzzle for political economy since the tradi-
tional blue-collar working class, the supposed pillar of the welfare state, has
everywhere declined during the past four decades (cf. Piven 1991).
One of the solutions to this puzzle proposed in the literature is that grow-
ing exposure to the international economy has increased labor market inse-
curities and propelled demands for social protection. Thus, Cameron (1978)
and Katzenstein (1985), and more recently Garrett (1998) and Rodrik
(1998), argue that even though integration into the international econ-
omy promises large potential welfare gains, such integration comes at the
cost of exposure to the ups and downs of global markets and reduced ca-
pacity for governments to counteract these cycles. The way governments


. 6
( 11)