Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=phd.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Ph.D. Program in Political Science of the City University of New York is collaborating with JSTOR to digitize,
preserve and extend access to Comparative Politics.
http://www.jstor.org
Political Culture and Democracy
AnalyzingCross-LevelLinkages
61
ComparativePolitics October 2003
62
Ronald Inglehart and Christian Welzel
the individuallevel to the aggregate level, and the findings at both levels of analysis
were genuine and important.
Similarly, in contemporary France the vote for the xenophobic National Front
tends to be highest in districts with high percentages of Islamic immigrants. This
correlation does not mean that the immigrants are supporting the National Front.
They are not. Conversely,the fact that the immigrantsare not voting for the National
Front does not mean that the linkage between ethnicity and politics is spurious. The
relatively high percentage of immigrants has a major impact on the vote for the
National Front, even though the correlation between vote and immigrant status
reversesits polarity from one level of analysis to another.
Likewise, the fact thatjobless Germansin the early 1930s did not show a stronger
tendency to vote for the Nazis than those Germanswho still were employed does not
mean that there was no causal linkage between unemployment and the Nazi vote
share. Sharply rising unemploymentrates created a climate of anxiety that affected
all social groups, whetheremployed or not, increasing their readiness to vote for the
Nazis. Thus, the rise in unemploymentlevels from the late 1920s to the early 1930s
was followed by a strong increase in the Nazi vote. The fact that the unemployed
were as likely to vote Communistas Nazi at the individual level does not mean that
unemploymentwas unimportant.
As these examples demonstrate,it is perfectly possible-and frequently true-
that an aggregate level linkage is not reflected at the individual level. Nevertheless,
this linkage is not somehow unreal or spurious. Quite the contrary,aggregate level
linkages often have more impact on society than those found at the individual level.
Assuming that rising unemploymenthas no impact on support for extremist parties
because there is no linkage between unemploymentand extremism at the individual
level would be committing the "individualistic fallacy."5 Seligson, nevertheless,
claims that the linkage that Inglehartfound between interpersonaltrust and democ-
ratic institutions at the aggregate level is spurious because he finds no linkage
between trust and supportfor democracyat the individuallevel. This conclusion is a
classic case of the individualisticfallacy.
Seligson's article examines the individual level correlations among a set of indica-
tors that Inglehart used in analysis of the 1981 World Values Surveys. Readers of
Seligson's article would probably assume that it also refers to Inglehart's recent
work. However,Inglehart'sanalysis of the 1990-91 surveys and his subsequentwork
moves beyond the indicatorstested in Seligson's article (life satisfaction and inter-
personal trust), incorporatingthem into a broader set of indicators of political cul-
ture. This fact is of relatively minor importance.The critique in this article applies
63
ComparativePolitics October2003
equally to Inglehart'soriginal findings and to his more recent work: societies with
relatively high levels of interpersonal trust and life satisfaction are significantly
more likely to have democraticinstitutionsthan societies with lower levels, and this
linkage is by no means spurious.Nevertheless, it is worth noting that Inglehartand
othershave identified a broadersyndromeof self-expression values that includes not
only interpersonaltrust and life satisfactionbut also several other attitudesthat seem
to play even more importantroles in promoting democracy.6The respectivepublics'
locations on this self-expression values dimension, together with economic indica-
tors, explain roughly 80 percent of the variance in democratic institutions. The
dependentvariablein this analysis is an indicatorof democraticinstitutionsthat will
be referredto as effective democracy.
64
Ronald Inglehart and Christian Welzel
Llsav Assyeleit
levelwithin adiv
ndividual e leve acre
eidul as Ag gat cron
(mn nton (pooled dat) usional level
Vanaions
ladindg)
Strongseltcaprcssionvalue
strongemphasisn the
dkreflect
followingattuzdes/behavior
-T 82
Meler'a tAdlvlSy .4"1 -6868
- Pub*lsc 187
S-elspreie••45' .65
- Lbertysad Putarelp"Atn' 354 19 12
Tra~t 34 47 64
Sflarpers•al
- Life Sthtctle .13 .44
'76
Weak selfvwmia valves
rolect weak emphasix aon the
v-mance
Explained 23% 29% 54%
Numberof ease 137national
survey 15803 137
-------_vdPfaMo- a
wmrwavewaunits
Abts; Hnries ae (actor taig plrative prikcipal meszana nreascion
scmdmpyon of hctorewiha
'Eigenvalues'aboveI avienorotation. Source ea/W Vales Fly.S
Surveys-tV
mneranadftr"disiked niui bors"acded""
"Notr at d qu-iebue d AW AsOarsadded or
acorain
weighbors withAIDS(V59)sadbcan sswual dgm(VW)
Agmegaee l avragesondhi 0. sck
datsare eaioWs t
"Havedone"for"signaingptitions(VI 18)coded"I"anddicb aainst "0*.
d•gd
Agpegatfdataarenaidon palrcenapshavedone.
Respondent'irstandsecondpriorities for"vinS people in importantgoverntent
ra•P y
decsions" and"protectingfeedom of speech"(V 106-107) addedto a four-pointindex, asigning 3
points fr both items on fir and second rank,2 pos for one of these
aSems on firstrnk. I ppointfor
otwofthees
batno osiaondra and0 for nneof 'ie man fws orsecondrank.
cancel each other out.11 Following the law of large numbers,this reduction of error
becomes more pronounced as the number of individuals being aggregated rises.
Consequently, the random term becomes smaller, and the systematic correlation
larger, at higher levels of aggregation. Consequently, aggregation to the national
level does not produce spurious correlations. Quite the contrary,aggregation often
reveals systematic correlationsthat may be hidden by measurementerrorat the indi-
vidual level within nations. Hence the syndrome of self-expression values is much
65
ComparativePolitics October 2003
more pronounced at the aggregate national level than at the individual level within
nations (compare columns 1 and 3 in Table 1). As Erikson, MacKuen, and Stimson
argue, analysis at the aggregate level often provides a more accurate view of the
underlyingrelationshipsthan is availablefrom individuallevel analysis.12
The strengthof the correlationsat the pooled individuallevel falls between these
two extremes (compare column 2 with columns 1 and 3 in Table 1). At the pooled
individual level the variation in an attitude is composed of the deviations from the
mean within nations, which may be randomto a relativelylarge degree, and the devi-
ations from the mean between nations, which are largely systematic. Thus, at the
pooled individuallevel there is more systematic variationthan at the individuallevel
within nations, and, in turn, the pooled individual level entails more randomvaria-
tion than the aggregate level. The factor loadings of the self-expression values syn-
drome thereforeincrease from the individuallevel within nations to the pooled indi-
vidual level to the aggregate level.
In short, relatively weak correlations at the individual level do not indicate that
relatively strong aggregate level correlations are somehow false or spurious.Quite
the contrary,aggregate correlationsmay reveal linkages that are obscuredby random
measurementerrorsat the individuallevel. Moreover,the aggregate level is precisely
the level at which democracyexists: democracyis an attributeof nations,not of indi-
viduals. Hence, if one is interested in the impact of mass attitudes on democracy,
what matters is a society's mass tendency in these attitudes,not the individuallevel
attitudinalstructure,as Seligson assumes.
66
Ronald Inglehart and ChristianWelzel
is, civil rights and political rights).16This scale has been reversed so that higher fig-
ures indicate a broaderscope of freedomrights. The scores from FreedomHouse are
expert ratings of the extent to which certain individual rights are guaranteed.17The
most recent Freedom House scores from 1999-2000 have been used in order to
ensure that the measure of freedom rights is subsequent in time to the political cul-
ture indicatorsthat are used as predictorsof democracy.
The Freedom House scores are imperfect measures of freedom rights. They do
not take into account the extent to which given rights are respected in actual elite
behavior.To overcome this problem, the corruptionperception indices developed by
TransparencyInternational are used.18 These scores are also expert ratings; they
judge how corruptthe political, bureaucratic,and economic officeholders of a coun-
try are. One indication of the validity of these estimates is their strong correlation
with aggregatemeasures of the citizens' perception of elite corruptionin representa-
tive surveys.19
The TransparencyInternationalscores range from 1 to 100, with 100 indicating
the greatestamountof corruption.Reversingthese scores provides a measure of law-
abiding elite behavior or elite integrity.20Effective democracy is operationalized
throughweighting freedom rights by elite integrity. Since elite integrity will operate
as a weighting factor and not as a compensatingfactor, it is standardizedto 1.0 as its
maximum, obtaining fractions from 0 to 1. Hence, to obtain effective democracy,
freedom rights (standardizedto a maximum of 100) are multiplied by fractions from
0 to 1 for elite integrity.This produces an index of effective democracy that has 100
as its maximum. Since the most recent Transparency International scores from
1999-2000 are used, a measure of effective democracyin 1999-2000 is obtained.
Even if a country comes close to a maximum elite integrity of 1.0 (that is, almost no
elite corruption),the weighting procedurewould not compensate for a low level of
freedom rights. When a regime reaches only five percent of the possible maximum
in the freedom rights measure, a maximum elite integrity of 1.0 can not do more
than reproducethese five percent.21In contrast, a freedom rights level close to the
maximum of 100 percent can be severely devalued if elite integrity is so low that it
reaches only a small fraction of 1.0. Hence given freedom rights levels are devalued
to the degree that elite integrityis absent, reflecting that given constitutionalguaran-
tees are made ineffective in proportion to elite corruption. High levels of elite
integrity can not produce effective democracy, in the absence of freedom rights.
High levels of freedom rights, in contrast, produce formal democracy, but formal
democracy is effective only to the degree that elites base their activities on rights
instead of bribes.
67
ComparativePolitics October2003
1g0
e
040
10
20
00 10 ** * *40 *0 9.
00 10 20 30 40 50
60 70 80 90 100 110
FreedomRights 1999-2000
68
Ronald Inglehart and ChristianWelzel
69
ComparativePolitics October2003
105 C.'rrmarL
0
95 3894 29 40 x
90
5
oR R • *=0.75 e swifw,, C* • Sas,
.
....
. ka
.E .s.s.....
0) 5 d. 5
45 ?
? ?W CA
685 + VV )
i~~Gsranul
Spain Frew*
S55 o ''a 0 0
S.M h"a CM.
45
? ?dQf~
skmwa" 0
t 05
25 ElSay Ppwiposvowie o
Ro" ?nik
& w m
,o,, ? oS-K•t?&w
aO ced 0 h
? 15M s•O 0 a Peudh
~l~GCA*
0 •'eu
wln.
10 Tw! 0 0 C'.IRw~,
(
05
E4F""e
0I Cal
0luub~U~~I
oW
c0
-05
-15-1 3 -1 -09-07-050- 3-0.1 0.1 0,3 0.5 09 1t1 13 15 1.7 19 2.1
0,7
Strengthof Self-ExpressionValues (1990)
their effect from beta=.86 in Model 1 to beta=.51 in Model 5. Consideredconverse-
ly, however,the inclusion of self-expression values diminishes the effect of econom-
ic developmentfrom beta=.84 in Model 2 to beta=.43 in Model 5, adding 10 percent
of explained varianceto what economic developmentalone explains. Thus, although
self-expression values and economic development are strongly correlatedwith each
other, they are not completely exchangeable, since both add a significant amountof
explained varianceto the effect of the other.
By contrast,the length of time a society has experiencedunder democraticinsti-
tutions adds very little to the effect of self-expression values on effective democracy
(2 percent, see Models 1 and 6).25Moreover,a society's experience with democracy
only slightly diminishes the effect of self-expression values on effective democracy
(the beta-coefficient shrinks from .86 in Model 1 to .73 in Model 6). Conversely,
however, experience with democracy'simpact on effective democracy shrinksfrom
beta=.75 in Model 3 to beta=.18 in Model 6, controlling for self-expression values,
implying that self-expression values do not result from the presence of preexisting
democratic institutions. If they did, the length of the society's experience with
democracy would capture significant parts of the effect of self-expression values,
but it does not.
70
Table 2 The Effect of Self-ExpressionValueson Effective Democracy,Controllingfor Rival Pre
ofi 1
ModSel 2MlE O MI4 MOd Mo
auppwfar s .AD
-arac (4)
W 74 .70 .5 .34 80 .7
".,4
ComparativePolitics October2003
72
Ronald Inglehart and ChristianWelzel
3 5 .---------
15
MCI"#
a 5wene
*4
bod
0 *USA.
e.
oapny%; -al t
N-
01
50
so ramteo A
20 UA
0s**AN N0 *
Aa
Sm, oo
010
73
ComparativePolitics October2003
S S
awo Cromia
w0f SwedenA
o y
9ur V 0 c "Ww kn
u 60 ,e
EO * A, SpmwS *
E Fw*4AcA uguay S F Fapd~
* 5o
SouthAfs
40 tydbac? tu ?
•O
30 B
y10
Moo~~*9TU* gpj4o y49*4+ 1.09
- sq,
020 ?R
u- =-35
-15 -1,3 1,5 2,3
-1,0 -0,8 -0,5-0,3 0,0 0,3 0,5 0,8 1,0 1,3 1,8 2,0
waMass-Emphasis on Self-Expression 1990 35
on S~f-IExprosslon I 90O
Malr~ssEmnphaslsa
found that individual supportfor democracy is determinedfar more by instrumental
motives than by normativecommitmentsto the values that are inherentto democra-
cy. This finding is perfectly replicatedin the analysis of the WorldValues Surveys, as
Table 3 shows. Although there is a linkage between supportfor democracyand self-
expression values at the individuallevel (see Model 2), what people think about the
performanceof democracyin runningthe economy and maintaininglaw and orderis
a much better predictorof their overt supportfor democracy (comparethe explained
variances of Models 1 and 2). To be sure, people with strong emphasis on self-
expression almost always preferdemocracy to autocracy,but there is a large number
of people who supportdemocracyfor reasons of expected performance,even if their
emphasis on self-expression is weak. Hence overt support for democracyis a poor
indicator of intrinsic support,since overt supportis inflated by instrumentallymoti-
vated lip service.
Conclusion
In analyzing data from the 1981 WorldValues Surveys, Inglehartfound that societies
with relativelyhigh levels of interpersonaltrustand life satisfactionwere much more
74
Table 3 PredictingSupportfor Democracyat the IndividualLevel
a
1pde Sup Demo 19954
I Model
Predlctors4
.B SE)
-M (.02)
Support
~ PitaN- . .1A."
..
... . Ow
lsoSupd
Ol SE
. . -.... J .R.R
PartiMl
-.
Oatwsaewsar am e dedalew w. -?2 -6*1
(.02w)
Dem0oo amrebadin malntefrlng
order" -45OM(2) -.18' -.
AV":(0) .1t" A
Constant -.58 LOS) 4455(06) -t
R
Adjusted .24 .11
N AM46050
y'Me
bVmtow" mrs ded(4-atoglysagrnS. rore, 1
b) VlSI*Dam.craclesareIndeaIsieandhove
s tooouch qutbl*ng Forcoding,see a).
-i"..
0d1)Po dIod fetrano$0Morvwi i tdn Tat 1(center
000nnM).
e-tas
ugnlcant
atthe.001-level.
Eletsobtained
a In g o lforuural
zones,
usIngm ntI0
forea
zones sawby 8). Efectsofcultural
andSaker(see obln0te zoned notdocumented
Orremoof s
Source:EuropeafoMd Valuaesm
Surveys11(1995-48).
.......
.
-4
th
ComparativePolitics October2003
Appendix
76
Ronald Inglehart and Christian Welzel
NOTES
77
ComparativePolitics October2003
78
Ronald Inglehart and Christian Welzel
in case of countries that were not independentbefore 1850) until 1995. Countriesthat emerged from the
dissolution of the Soviet Union and Yugoslaviahave been coded like their formermother country as long
as they belonged to it. A year has been counted as one under a democratic constitution if a country
obtained at least +7 points on the Autocracy-Democracy index from Gurr and Jaggers, note 201. This
index is based on an analysis of constitutions and considers the extent of restrictionson executive power
and the voters' opportunities to influence politics. Gurr and Jaggers classify countries as "coherent
democracies"if they reach +7 or more points on their -10 to +10 index. Data and methodological descrip-
tion can be obtained from the homepage of the "Polity 98" project, http://www.bsos.umd.edu/cidcm/poli-
ty. These data are used here because they reach fartherback in time than the scores from Freedom House
and are thereforemore adequatein measuringthe enduranceof the democratictradition.
26. Democracy scale accordingto Hans-DieterKlingemann,"MappingPolitical Supportin the 1990s:
A Global Analysis," in Pippa Norris, ed., Critical Citizens: Global Supportfor Democratic Governance
(New York:OxfordUniversity Press, 1999), pp. 31-56. In the first step, we added up respondent'ssupport
of the statements"Havinga democraticpolitical system" (V157) and "Democracymay have problemsbut
it's better than any other form of government"(V163). Supportfor these statementscould be expressed in
four categories: very good (code 3), fairly good (code 2), fairly bad (code 1), and very bad (code 0) in
case ofV157 and agree strongly(code 3), agree (code 2), disagree (code 1) and disagree strongly (code 0)
in case of V163. People's supportfor these statementshas been added up to a 0 to 6 scale, with 6 repre-
senting the highest supportfor democracy.In the second step, we added up people's support of the state-
ments "Having a strong leader who does not have to bother with parliamentand elections" (V154) and
"Having the army rule" (V156). Analogous to the first step, a 0 to 6 scale of supportfor autocracy was
created. In the third step, we subtractedthe support for autocracy scale from the support for democracy
scale to create an overall index of autocraticversus democraticsupport,ranging from -6 (maximum auto-
cratic support)to +6 (maximum democratic support). In the fourth step, we calculated for each country
the percentage of people scoring at least +4 on this index (since from +4 onward they are closer to the
maximum democraticsupport,+6, than to the neutralpoint, 0). The percentageof solid democratsis thus
obtained for each country.
27. Michael Bratton and Robert Mattes, "Support for Democracy in Africa: Intrinsic or
Instrumental?,"British Journal ofPolitical Science, 31 (2001), 447-74.
28. Welzel;Welzel, Inglehart,and Klingemann.
79