Sei sulla pagina 1di 15

Society for Political Methodology

The Measurement and Meaning of Trust in Government


Author(s): Stanley Feldman
Source: Political Methodology, Vol. 9, No. 3 (1983), pp. 341-354
Published by: Oxford University Press on behalf of the Society for Political Methodology
Stable URL: http://www.jstor.org/stable/25791197 .
Accessed: 21/06/2014 17:13
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
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.

Oxford University Press and Society for Political Methodology are collaborating with JSTOR to digitize,
preserve and extend access to Political Methodology.

http://www.jstor.org

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

The Measurement
Trust

and Meaning of
inGovernment

Stanley Feldman

In political
Perhaps the most well documented trend
over the past 20 years has been the sharp
attitudes
increase
in political
in the 1950s and early
Research
cynicism.
1960s found Americans
of political
loyal and highly trusting
authorities
(Almond and Verba,
1965).
Since
1963; Lane,
the mid-1960s,
a large number of studies
using a number of
different
in
indicators
declines
have shown substantial
of both
and public evaluations
trust,
political
confidence,
and nonpolitical
and" leaders
Institutions
(Miller,
political
House and
1974a; Hill and Luttbeg,
1980; Ladd,
1976-77;
the
1976).
1975; Wright,
Mason,
Despite
large body of evi
dence on the decline
of political
trust,
major questions
still
of
and consequences
remain about the meaning, causes,
the observed
of these
trends.
Although the significance
a major
in political
trends
attitudes
remains unclear,
issue

that

needs

to

be

addressed

is

the

meaning

of

responses

to the items used to document these trends.


A good example of the problems surrounding
the meaning
of such social
is the popular trust
indicators
and political
in a long series
that has been included
in-government scale
of SRC/CPS National
These
items have
Election
surveys.
and
become popular measures of political
trust/cynicism
but consid
have been Included
In other
surveys,
large-scale
over their
erable
conflict
has developed
meaning.
precise
The interpretation
around the problem of
issue revolves
of alle
diffuse
1975):
(Easton,
general
feelings
support
Miller
the
for
and
(1974a,
political
regime.
giance
support
the
for interpreting
1974b) made the case most strongly
were collected
AUTHOR'S NOTE: The data used in this analysis
and made available
Studies
by
by the Center for Political
and Social
the
Consortium
for Political
Interuniversity
for the
neither of which bears any responsibility
Research,
or interpretations
here.
results
reported

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

342

Political Methodology

trust
items as measures of diffuse
political
support,
arguing
that mistrust
is in large part a consequence
of fundamental
and political
social
conflict
and has potentially
significant
for the political
consequences
Miller,
system.
Goldenberg,
and Erbring
have more recently
(1979:79)
argued that cynicism
"reflects
dissatisfaction
with government perfor
general
a 'leading*
mance" and may "be considered
Indicator
of dif
fuse

support."

An alternative

is that the trust-in


interpretation
reference
to the "govern
of their
items, because
government
ment in Washington"
and "the people
in Washington,"
tend to
reflect
attitudes
toward the incumbent authorities
rather
than toward the political
Citrin
system more generally.
latter
in part because
(1974) argued for the
interpretation,
he found that expressions
of cynicism were often accompanied
in the political
of pride
Muller and
by feelings
system.
Jukam's (1977) data from a German survey show a clear
dis
tinction
between attitudes
toward political
incumbents and
toward the political
They argue that political
system.
is more related
trust
to incumbent affect than to system
Most recently,
(1982)
Muller,
Jukam, and Seligson
support.
trust measure does not relate
have shown that the political
to anti-system
behavior
in the way that a measure of diffuse
The key issue
in this debate
is
political
support should.
the seemingly ambiguous meaning of "government"
in the trust
do people
items:
leaders
interpret this to mean the specific
in power at the time or the more established
governmental
institutions
from which long-term legitimacy
derives
(Hill,
1981)?
One data set may offer some answers to this question.
were
Election
In the 1978 National
study, four new questions
included with the five trust-in-government
Using
questions.
the same wording as the "do-what-is-right"
and "few big
interests"
items of the trust-in-government
two items
scale,
were changed to refer to "President
and "the Carter
Carter"
while two others
"the U.S. Congress."
administration,"
specify
(See the appendix
for the wording of these questions.)
The
more specific
of these questions
referents
make it possible
to examine the relationship
between the general
trust
items
and trust
in both the President
and Congress,
and thus to
of the meaning of trust
in gov
gain a better understanding
ernment.

Two problems stand


in the way of a simple
Interpretation
of the relationships
among the three sets of trust
items,
as Abramson and Finifter
however.
(1981) recognized
First,
in their analysis
of these data, the use of identical
ques
the three referents
the
tion wording across
introduces
of contamination
because of common methods
possibility

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

343

Stanley Feldman

variance.

Thus,

correlation

between,

say,

trust

in

govern

ment and trust


in Congress could reflect
the covariance
gen
erated
to the common question
by people
reacting
wording.
Abramson and Finifter's
(1981) analysis
of these
items showed
that common method variance
is a potential
least
problem, at
for the few-big-interests
Even with this source
questions.
of bias recognized,
found
however, Abramson and Finifter
substantial
between the original
trust measure
relationships
and the new measures of trust
in Carter and Congress.
awareness
the question
their
of
Despite
wording problem
and their careful
attempts to deal with it, their methods
of dropping
items from scales
while
(examining the effects
for changes
in scale
in
fall short
correcting
reliability)
two respects:
they are able to detect only quite
large
and they cannot accurately
methods effects
estimate
either
or the true correlations
the magnitude of those effects
removed.
with methods effects
In
among the trust measures
that attempts to estimate
unbiased
addition,
any analysis
constructs
must take
into
among attitudinal
parameters
in the
account
both systematic
and random error components
1968).
measured variables
(Blalock,
The second problem that must be dealt
with before these
new trust
the meaning of the
items can be used to assess
scale

trust-in-government

is

the

causal

nature

of

the

Plausible
theoretical
observed
relationships.
arguments
can be used to defend three different
of a
interpretations
and trust
in
trust scale
correlation
between the general
as Miller
there
Carter or Congress.
(1979) argues,
First,
may

be

"spillover"

effect

in

which

general

distrust

of

in increasing
of the
distrust
authorities
results
political
of
of the government or the occupants
institutions
specific
the direction
of causal
those
institutions.
Alternatively,
of
with generalized
distrust
influence could be reversed,
of
of distrust
authorities
being a consequence
political
the President
and Congress:
distrust
of the major political
and incumbents builds
institutions
up and leads to a more
sense of distrust
and the political
of politics
diffuse
of this
less
substantive,
Another,
interpretation
system.
in "government"
causal
flow is that, when asked about trust
or the "people
in Washington,"
respond on the basis
people
of the national
features
of the most salient
government:
correlations
observed
the President
and Congress.
Finally,
could simply be spurious.
trust measures
among the several
that the government has not solved
important prob
Feelings
to people's
lems or is generally
may
opinions
unresponsive
of the government simul
of various
lead to distrust
aspects
of
among the three measures
Thus, correlations
taneously.
causal
direct
demonstrate
trust do not necessarily
political
among them.
relationships

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

344

Political Methodology

more precisely
the relation
I estimate
In this paper,
in government
trust
items and the
ships between the general
new items measuring trust
in President
Carter and the Con
with the issues of systematic
and
gress.
Only by dealing
of causal
random measurement error and the structure
effects
can the new trust
items provide
useful
information for
the meaning and significance
of political
understanding
trust.

DATA AND METHODS


Election
paper uses the same 1978 CPS National
As noted
(1981).
employed by Abramson and Finifter
to the standard
in addition
five trust
items, four
earlier,
new items were included.
the word
The new items duplicate
the
the
referents
of
of
two
standard
items,
changing
ing
and "the Carter administration"
Carter"
first to "President
the other three trust
Because
and then to "the Congress."
not be
in the new form, they will
items were not repeated
cannot
in this analysis:
included
question
wording effects
be estimated
with a single
item. The analysis
will thus
the relationships
items represent
consider
among six trust
constructs
(trust
in government,
ing three substantive
and two methods effects
and Congress)
Carter,
(question
This

Data

wording).

these six
The hypothesized
items is
model underlying
1.
shown in Figure
In this model, the measured variables
are enclosed
constructs
in
in rectangles
and the unobserved
Variation
in each
item may result
circles.
from three dis
a substantive
the
tinct
influences:
factor representing
dimension
trust
of
Carter,
specific
political
(government,
a question
or a random error
and Congress),
wording effect,
The estimates
term.
of the correlations
among the three
are thus free of the effects
trust
factors
of both system
atic and random measurement error.
In addition
to the mea
surement model shown in Figure
six
variables
1,
exogenous
were included
as predictors
in the analysis
of the three
trust
factors.
This was done both to aid
in the identifica
tion of the model and to provide evidence
of discriminant
The six vari
factors.
among the three estimated
validity
are:
ables
of the job performance of
respondents'
ratings
Carter and Congress,
of the government's
approval
handling
of the economy and the respondents'
most
self-identIfied
important problem, an index of government responsiveness,
and party identification.1
The entire model, with the six trust
indicators
and six
was estimated
exogenous
variables,
using LISREL (Joreskog
and Sorbom, 1978), which provides
fuI I-Information,
maximum

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

J
?lV^RESW
R.GHT

GOVERNMENT
GOVERNMENT
CARTER
CARTER
CONGRESS
CONGRESS
DO
WHAT
FEW
IS
BIG
DO
WHAT
IS
FEW
BIG
DO
WHAT
IS
FEW
BIG

y,
Az,
a/,
A/
\v
a/
Figure
Factor
Structure
the
Political
Items
I.of
Trust
RIGHT-INTERESTSRIGHTINTERESTSRIGHTINTERESTS e, /
\?2e4
e3
/
\ e5
/\e6

I)4
GOVERN(CARTER
)4
(CONGIITRUST
\TRUST
/
\TRUST

J
J
V
\V
MENT
7RESS

/V.
r\

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

*n
______

346

Political Methodology

likelihood

estimates

constructs.

Two

for

means

identified

are

models
for

provided

with

unobserved
goodness

assessing

errors of the parameter estimates


and the
standard
of-fit:
errors can be used to
The standard
likelihood
ratio value.
for statistical
and test each parameter
compute t-values
In
the
likelihood
ratio
value
significance.
large samples,
a
as a chi-square
and provides
is distributed
distribution
the
for
entire
In
to
model.
test
contrast
goodness-of-fit
a large chi-square
most inferential
value
indi
statistics,
that the model does not fit the data.
One drawback
cates
ratio value
is a direct
of this test
is that the
likelihood
that any non
function of sample size,
making it unlikely
trivial
model will
value with a
produce a small chi-square
In this case,
a more useful
data set from a large sample.
to degrees of freedom
is to examine the chi-square
strategy
of under five are consid
ratio.
As a rough guide, ratios
ered evidence
of good fit for a sample of the size
used
here.

RESULTS
1 provides
the LISREL estimate
Table
of the basic mea
of the six exogenous
varia
surement model and the effects
bles on the three trust constructs.
All coefficients
have
been

The

standardized.

of

chi-square/degrees

freedom

ratio

a very good fit for a sample size of over 2000.2


is 3.58,
the two
The estimates
in all three cases
that
show, first,
On
trust.
of political
items are equally
good indicators
the other hand,
in all six cases
effects
wording
question
for more than half of the variance
and random error account
the underlying
in the indicator.
On the average,
trust
in
construct
determines
only 42 percent of the variance
these
It is thus clear
that any estimates
of
indicators.
measures
the relationships
the
will
be
three
trust
among
badly

biased

unless

both

sources

of

error

variance

are

con

led for.
As Abramson and Finifter
the impact of
concluded,
across
the items,
question
wording varies
significantly
us to see more precisely
how each of the six trust
allowing
The
items is affected
to question
by response
wording.
show a substantial
and consistent
estimates
word
question
for the few-big-interests
These
questions.
ing effect
are consistent
with Abramson and Finifterfs
large effects
The three
set effects
for these
items.
finding of response
items are not completely
free of question
do-what-is-right
however.
For two forms of this
item, the
wording effects,
the question
is
government and Carter,
wording effect
in little
and thus results
between pairs
correlation
slight
trol

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

Stanley Feldman

347

TABLE

OF THE POLITICAL

PARAMETER ESTIMATES
A:

of the

Estimates
Trust
Govern
ment

Indicator
Government?
Do what is right

.66

Government?
Few big
interests

,63

Measurement

Factors

Carter

Carter?Few
big Interests

.65

Congress

Congress^?Few
interests
big

.68
are

Impact of

Trust
Factor

Rating
of
Carter

Rating
of
Congress

Government

.09**

.17**

Carter

.48**

Congress

.06

* =
significant
** =
significant

-.01
.3.5**
at the
at the

=64.5
chi-square
of freedom =
degrees
= 3.58
chisquare/df

.05
.01

significant

Exogenous
_Exogenous

at

Most

.26

level.

on Trust

Variables

.40

.30

,52
.01

.41

.52

.51

the

Error

.53

.42

.67

B:

Factors
Few big
Interests

.21

Congress?Do
what is right

coefficients

Wording
Do what
is right

.42

.66

All

Model

.20

Carter?Do
what is right

Note:

TRUST MODEL

Factor

Variables_
Government

Handling
Economy

Important
Prob Iem

.16**

.15**

.34**

.06*

.15**

.30**

.12**

.13**

.38**

Responsiveness

level.
level.

18

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

Party
Identi f icat ion
-.05
.06*
-.04

348

Political Methodology

case of the Congress


In the specific
of the three
Items.
of this question,
version
however, the influence of question
is substantial.
In fact, question
wording
wording effects
are quite
for
both
the
items, approach
pronounced
Congress
levels of the substantive
trust
factor.
ing the
on
of the exogenous
variables
Turning to the effects
the three trust constructs,
has a
government responsiveness
The
substantial
effect on all three trust constructs.
to the
is not being responsive
belief
that the government
the
in Carter,
is clearly
linked to a lack of trust
public
and the government more generally.
Not too sur
Congress,
of the overall
performance of Carter and
prisingly,
ratings
are strongly
to expressed
in
the Congress
trust
related
and
each.
Note that the relationship
between evaluation
for Carter than for the
trust
is significantly
stronger
is more
that trust
in the President
suggesting
Congress,
to
is that
of
than
tied
evaluations
job performance
closely
the impact of job
In the case of Carter,
In the Congress.
than the impact of government
is greater
rating on trust
reverse
the
is true for the Congress.
while
responsiveness,
no effect
has virtually
Surprisingly,
party identification
a statistically
on trust,
coefficient
significant
appearing
of
in Carter.
holding ratings
Finally,
only for trust
of the govern
Carter and Congress
evaluations
constant,
ment's

handling

of

the

economy

and

the

respondent's

most

but substantively
small,
important problem have significant,
on all three trust constructs.
effects
taken
With both sources
of error variance
into account,
the correlations
to estimate
it is possible
among the three
trust constructs
free of the impact of random measurement
error and question
The estimates
of the
wording effects.
are:
correlations
between Carter and Congress,
.57; between
and the
Carter and the government,
.63; and between Congress
.78.
these
results
show that,
Overall,
purged
government,
there are very
of random and systematic
measurement error,
Thus,
among the three trust constructs.
strong correlations
co
the observed
inflates
although common question
wording
ran
of
the
variation
substantial
amounts
constructs,
among
dom error attenuate
the relationships
In
considerably.
terms of the meaning of trust
in government, the most impor
tant result
is the significantly
relationship
stronger
in government and trust
between trust
in Congress than
between the former and trust
in Carter.
The true significance
of the correlations
among the
nature of
trust constructs
however, on the causal
depends,
the relationships.
To assess
it is necessary
to
this,
a
in
model
which
rela
nonrecursive
simultaneous
specify
are permitted
between trust
in government and
tionships

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

349

Stanley Feldman

both trust
in Carter and trust
In Congress.
The newly
model was estimated
with the LISREL program.
The
specified
fit of this model
is somewhat better than the first,
reflect
of nonsignificant
causal
ing the elimination
paths between
the exogenous
variables
and the three trust constructs
(chi
= 3.35).
The estimates
of the measurement compo
square/df
nent of the model remained virtually
The esti
unchanged.
are (all coefficients
mates for the three trust constructs
are significant
at the .05 level):
IN GOVERNMENT=

TRUST

+ .20TRUST

.53TRUST

IN CARTER +

HANDLING OF ECONOMY +
-.08PARTY

.11 GOVERNMENT
.09RESP0NSIVENESS

IDENTIFICATION

IN CARTER =

TRUST

IN CONGRESS

.17TRUST

IN GOVERNMENT

+.46RATING

OF CARTER +

PROBLEM +

.24RESP0NSIVENESS

.15M0ST
+

IMPORTANT
.09PARTY

IDENTIFICATION
IN CONGRESS =

TRUST
+

.18TRUST

IN GOVERNMENT

.31 RATING OF CONGRESS +

HANDLING OF ECONOMY +
PROBLEM +

.09G0VERNMENT

.13M0ST

IMPORTANT

.31 RESPONSIVENESS

in govern
of trust
Looking first at the determinants
in Congress.
is trust
influence
ment, we see that the dominant
The standardized
for this construct
is more than
coefficient
two and a half times
larger than the next largest coefficient,
held
in Congress
and Carter
for trust
in Carter.
With trust
of
the
the
economy, govern
constant,
handling
government's
have only
and party identification
ment responsiveness,
on trust
It seems clear,
in government.
effects
marginal
their
level of
then, that when people are asked to express
trust
in the federal
primarily
government they are responding
on the basis of their trust
in Congress.
on the other hand, are
Trust
in Carter and Congress,
in government more gener
by trust
only somewhat influenced
to
for distrust
is some tendency
Although there
ally.

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

350

Political Methodology

over" from the "government" to the institutions


of
"spill
is not nearly as pronounced as the
government, the effect
effect of trust
in Congress on trust
in government.
Beyond
are
the determinants
of trust
in Carter and Congress
this,
to the previous
trust
very similar
analysis.
Specifically,
in Carter
is more responsive
to job performance
ratings
than is trust
in Congress
(.46 vs.
.31), while the reverse
is true of the impact of government responsiveness
(.31 vs.
.24).

CONCLUSIONS
cannot answer all the questions
Although this analysis
some impor
to the meaning of trust
relevant
in government,
can be drawn.
tant conclusions
the analysis
has
First,
that at
least two of the items (and pre
shown very clearly
scale
sumably all)
making up the popular trust-in-government
are contaminated
by both random and systematic
(question
more than half of the
Overall,
wording) measurement error.
in the items reflects
sources
variance
of error rather than
the substantive
construct
of political
trust.
underlying
the effects
are
of question
Second,
wording
quite evident,
for the few-big-interests
Thus, for
especially
questions.
this question,
and to a somewhat lesser extent the do-what
reflect
not only their
items, people's
responses
is-right
of trust but also their tendency to respond
in
feelings
to
consistent
certain
of
As
Abramson
ways
questions.
types
a problem
and Finifter
this
(1981)
is clearly
recognized,
when

the

referents

of

the

trust-in-government

scale

are

Some of the observed


covariation
be a result
will
changed.
of methods variance
rather than substantive
relationships.
the correla
inflates
Third,
although
question
wording
between the trust constructs,
tions
random measurement sig
them.
Estimated
correlations
among
depresses
nificantly
the three,
sources
are very
for
both
of error,
controlling
between trust
in
in government and trust
large, especially
These correlations
Congress.
suggest that either members
of the public
do not distinguish
between the "gov
clearly
and the Congress
in their
ernment," the incumbent President,
or that the sources
of trust or cynicism,
of mis
feelings
trust
in all three are quite
similar.
from a nonrecursive
model show that
Fourth, estimates
the dominant causal
these correla
relationship
underlying
tions
is the effect of trust
in govern
in Congress oni trust
The other three causal
ment.
in Carter
(trust
relationships
on government, and trust
in government on the other two),
are much less
while statistically
important
significant,
than the first.
Thus, when people
report that they

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

351

Stanley Feldman

the "government,"
(mis)trust
they are in large measure saying
some
that they (mis)trust
there
is also
Finally,
Congress.
in the President
that trust
evidence
is based more on the
in Congress
job performance of the incumbent, while trust
is more institutionally
based
in perceptions
(as reflected

of

responsiveness).

of
What does all of this mean for the interpretation
the trust-in-government
is
scale?
the
key
Clearly,
finding
the strong causal
in Congress
and trust
link between trust
An initial
in government.
asks why this relation
question
in Carter on
is so much stronger than that of trust
ship
answer
in government.
is that both the
One plausible
trust
are understood
in
"government" and Congress
by the public
Carter
to President
institutional
terms, while the reference
to separate
is in very personal
terms.
it is easier
Thus,
of the
of the President
from the institutions
evaluations
This still
leaves the problem of how to inter
government.
in government.
in Congress on trust
pret the impact of trust
of a process
the operation
On the one hand, this may indicate
in the government
trust
(or cynicism)
by which people develop
as a whole on the basis of their evaluations
of the perfor
mance of specific
institutions
1979).
(Miller,
governmental
when asked to
is that,
However, an alternative
explanation
in "government,"
their trust
express
people respond on the
in this case on
basis of particularly
salient
information,
1981).
the basis of their evaluation
of Congress
(Hill,
in large part by people's
eval
This
linkage may be created
as an institution.
shows
Evidence
uations
of "Congress"
of Con
between evaluations
that people clearly
distinguish
If this
of their own representative.
gress and evaluations
the
trust-in
that
it
is
correct,
suggests
explanation
levels of
actual
may not be measuring
government scale
evalua
for the political
regime, but may reflect
support
in
The
data
institutions.
tions of specific
governmental
between these two explana
cannot distinguish
this analysis
this should be a primary goal of future research.
tions;
APPENDIX
are the six
The following
here.
in the analysis
reported

political

trust

Trust

items used

in government:
"How much of the time do you think you can trust the
about
in Washington to do what is right?just
government
some
the
time?"
or
of
the
of
most
time,
only
always,
is pretty much run by a
"Would you say the government
or that
it is
themselves
for
out
few big interests
looking
run for the benefit of a I I the people?"

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

352

Political Methodology

Trust

in Carter:
"How much of the time do you think you can trust
President
Carter to do what is right?just
about always,
most of the time, or only some of the time?"
"Would you say that the Carter administration
is pretty
out
for
much run by a few big interests
themselves
looking
or that
it is run for the benefit of all the people?"
Trust

in Congress:
"How much of the time do you think you can trust the
to do what is right?just
about always,
most
U.S. Congress
of the time, or only some of the time?"
is pretty much
"Would you say that the U.S. Congress
or
run by a few big interests
out
for
themselves
looking
that
it is run for the benefit of all the people?"
NOTES
1.
Evaluations
of the job performance of President
were obtained
from five-point
Carter and the Congress
scales,
from very good to very poor.
Approval of the government's
from separate
ques
handling of the economy was constructed
Those
items
with unemployment and inflation.
tions
dealing
of the government's
and one measuring
approval
handling of
the

respondents'

alternatives
The

government

"most

of

"good

important

job,"

responsiveness

problem"

"only

scale

fair,"
was

had

response

and "poor

constructed

job."

from

three
items asking whether the government pays attention
whether parties
make the government
what people think;
listen.
and whether elections
make the government
listen;
of government
indicate
approval
High scores
high trust,
performance,

party

government

identification.

responsiveness,

and

to

Democratic

2.
To minimize the problem of shrinking
sample size,
was employed
in the
deletion
of missing
values
pairwise
of the correlation
There were only
calculation
matrix.
between this matrix and one based on
minor differences
deletion
of missing values.
listwise
3.
that the two question
It is possible
wording fac
tors also tap a more general
level of political
trust
in
Two pieces
addition
to methods variance.
of evidence
sug
no
there
is unlikely.
is virtually
First,
gest that this
between the two factors.
there
correlation
is no
Second,
that the exogenous
indication
variables
have any effect on
the do-what-is-right
while they may have a slight
factor,
factor.
This
line of
impact on the few-big-interests

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

353

Stanley Feldman

cannot be pursued
analysis
identification
problems.

too

far,

however,

because

of

4.
To Identify the model, certain
variables
exogenous
were excluded
from each equation.
These deletions
reflect
both substantive
no direct
considerations
effect of
(i.e.,
of Carter on trust
in Congress,
con
and no direct
ratings
nection
between trust
in Congress
and trust
in Carter)
and
of the estimates
from the initial
model.
inspection
Finally,
all
terms were dropped from the equations
and
nonsignificant
the model was reestimated.
Several
other specifications
were tried and the estimates
of the relationships
among the
trust factors
proved to be very robust.
REFERENCES
Abramson, Paul R. and Ada W. Finifter.
1981
"On the Meaning of Political
New Evidence
Trust:
from Items Introduced
in 1978."
American Journal
of Political
Science
25:297-307.
A. and Sidney
Almond, Gabriel
The Civic Culture.
1963

Verba.
Boston:

Hubert M.
Blalock,
1968
"The Measurement Problem."
and Ann Blalock,
Methodology
New York:
McGraw Hill.
Citrin,

Easton,

Jack.
The Political
1974
"Comment:
American
Government."
68:973-988.

Little,
In Hubert
in Social

Brown.

H. Blalock
Research.

of Trust
in
Relevance
Political
Science
Review

David.

1975

of the Concept of Political


"A Re-assessment
British
Journal of Political
Science

Support."
5:435-457.
Hi I I, David
1981

Hill,

David
1980

B.
and the Measurement
"Attitude
Generalization
Political
Trust
in American Leadership."
3:257-270.
Behavior
B. and Norman R. Luttbeg.
Trends
in American Electoral
F. E. Peacock.
Illinois:

Behavior.

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

of

Itasca,

354

House,

James

1975

S.

and

William

"Political
American

M.

Mason.

Alienation
Sociological

in America,
1952-1968."
Review 40:123-147.

Karl G. and Dag Sorbom.


Joreskog,
1978
Guide.
LISREL
IV User's
EducationaI

Ladd,

Lane,

Everett
1976-77

Robert

1965

Research.

Carl I, Jr.
"The Polls:
Public
Opinion

Political Methodology

Chicago:

National

The Question
of Confidence."
40:544-550.
Quarterly

E.

"The Politics
of Consensus
ence."
American Political
874-895.

Mi Iler, Arthur H.
1974a
"Political
1964-1970."
68:951-972.

in an Age of Afflu
Review 49:
Science

in Government:
Issues and Trust
Review
Science
American Political

to 'Comment1 by Jack Citrin:


or Ritualism?"
American
Discontent
Review 68:989-1001.
Science

1974b

"Rejoinder
Political
Political

1979

Focus of Political
"The Institutional
Distrust."
at the 1979 Annual Meeting of
Paper presented
the American Political
Science
Association,
D.C.

Washington,

Miller,

and Lutz Erbring.


Arthur H., Edie Goldenberg
1979
Politics:
Impact of Newspapers on
"Type-Set
Science
American Political
Public
Confidence."
Review 73:67-84.

Muller,

Edward N. and Thomas 0. Jukam.


1977
"On the Meaning of Political
Support."
Science
Review 86:1561-1595.
Political

Muller,

A. Seligson.
Edward N. and Mitchell
1982
Political
"Diffuse
Support and Antisystem
A Comparative
Political
Behavior:
Analysis."
Science
American Journal of Political
26:240-264.

Wright, James D.
The Dissent
1976
Academic

of the Governed.

New York:

Press.

This content downloaded from 200.120.118.64 on Sat, 21 Jun 2014 17:13:37 PM


All use subject to JSTOR Terms and Conditions

American

Potrebbero piacerti anche