Sie sind auf Seite 1von 24

The CASMIN Project and the American Dream

Author(s): Robert Erikson and John H. Goldthorpe


Source: European Sociological Review, Vol. 8, No. 3, Special Edition on Social Stratification and
Social Mobility (Dec., 1992), pp. 283-305
Published by: Oxford University Press
Stable URL: http://www.jstor.org/stable/522720 .
Accessed: 16/05/2013 12:28
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 is collaborating with JSTOR to digitize, preserve and extend access to European
Sociological Review.
http://www.jstor.org
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
European Sociological Review, Vol. 8 No. 3, December 1992
283
?Oxford
University
Press 1992
T h e
CASMINproject and th e American
dream
ROBERT ERIKSON AND JOHN H.
GOLDT HORPE
ABST RACT In th is
paper
we
respond
to th e
critiques
of our work undertaken under th e
auspices
of
th e
CASMIN
project
th at are
presented by
Hout and Hauser and
by
Srensen in
preceding papers
in th is
number.
We treat in turn issues
concerning
data
comparability
and th e class sch ema th at we use as th e basis
for
our
analyses
of
mobility;
our model of 'core social
fluidity';
and
empirical
results relevant to th e
evaluation
of th e
FJH-h ypoth esis.
In conclusion we
point
to certain
conceptual presuppositions
and related
research
interests wh ich we would see as
deeply
rooted in th e American tradition of work in th e field of
social
stratification and
mobility
and
wh ich ,
we
suggest,
th row
ligh t
on th e nature of th e reaction th at our
work
h as
provoked.
INT RODUCT ION
We are
gratified
th at our work
arising
out of
th e CASMIN
(Comparative Analysis
of Social
Mobility
in Industrial
Nations) Project
h as so
quickly
attracted
attention,
and we much
appreciate
th e efforts th at h ave been made
by
our critics both in th is number of th e
European
Sociological
Review and elsewh ere. We
regret
only
th at th e
speed
of th e critical reaction h as
been such th at th e content of our
major
publication (Erikson
and
Goldth orpe, 1992)
would seem not
always
to h ave been
fully
assimilated.
In th is
response,
we do not
attempt
to take
up every point
raised in th e
foregoing papers
by
Hout and Hauser and
by
S0rensen. We
concentrate, rath er,
on wh at would
appear
to
be th e more crucial issues on wh ich
disagreement
occurs;
and we
seek, moreover,
to
place
th ese
issues-even ones of a
seemingly
tech nical
nature-in th e context of certain broader
considerations
regarding
th e
comparative study
of social stratification and
mobility.
T o th is
end,
it
may
be
h elpful
if at th e outset
we note two concerns th at
guided
th e
conception
and th e
subsequent development
of th e
CASMIN
Project.
First and foremost was a
concern to
improve
th e
quality
of data used in
cross-national research
and,
above
all,
in
regard
to
comparability. By
th e
early 1980s it h ad
become
apparent
in
mobility research -as in
oth er fields-th at
dispiritingly
inconsistent and
confusing
results would continue to be obtained
from
comparative analyses
unless
problems
of
data received far more serious attention
(cf.
Goldth orpe, 1985b). Consequently,
with in th e
CASMIN
project
a
large part
of our available
resources were devoted to data
preparation:
in
particular,
to th e
recoding
of th e unit-record
data of national
mobility enquiries
to new
standardized
categories
th at could be
applied
cross-nationally
in a
reasonably
reliable
way (see
furth er Erikson and
Goldth orpe,
1992:
47-53;
and,
for
greater detail, Konig
et
al., 1987;
Erikson et
al., 1988).
We derive
particular
satisfaction from th e fact th at th e
approach
and
tech niques
followed in th is connection
by
th e
CASMIN team h ave influenced similar efforts
at
improving
data
quality
in
subsequent
com-
parative projects,
such as th e
Luxembourg
Income
Study,
th e
Development
of Unionism
in
Europe Study,
and th e International
Survey
on Social Justice. It
is,
we
believe,
with research
th at is th us committed to
taking
data
seriously
th at th e future of
comparative
macro-sociology
lies.'
Secondly,
we indicated from th e start of th e
CASMIN
project
our intention of
treating
mobility
with in th e context of a class structure
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
rath er th an th at of a
prestige
or status
h ierarch y.
We need not reh earse our
arguments
for th is
conceptual
ch oice since
th ey
are set out
fully
elsewh ere
(Erikson
and
Goldth orpe,
1992:
29-35;
and cf. also
Goldth orpe, 1985a).
It will
be
enough
h ere to
say
th at we wish ed to be able
to treat
adequately
th e effects on
mobility
of th e structural
development
of national
economies;
and furth er to make our
analyses
of
mobility
as relevant as
possible
to th e work
of oth er
sociologists-and
of social h istorians-
primarily
interested in
questions
of class
formation and
decomposition,
th e
linkages
between class structures and educational
systems
and th e class bases of
political beliefs, values,
and action.
T h ese two
concerns,
with data
quality
and a
class-structural
perspective,
are of immediate
relevance to th ose
arguments
of our critics
th at we address in th e
following section,
regarding
th e
degree
of
aggregation
of th e class
categories
th at we use in our
comparative
mobility analyses
and th e
range
of nations th at
is covered.
T h ence,
we
proceed
to a series of
issues
pertaining
to our model of 'core social
fluidity' and,
in
turn,
to issues of a more
substantive nature
arising
from
attempts
at th e
empirical testing
of th e 'FJH
h ypoth esis'.
T o
end
with ,
we
point
to certain
conceptual pre-
suppositions
and related research interests wh ich
are,
we
believe, deeply
rooted in th e American
tradition of work in th e field of social stratifica-
tion and
mobility,
and wh ich
h elp
make more
intelligible-to
us at least-th e oth erwise often
puzzling
nature of th e reaction th at our work
h as
provoked.
DAT A AND T HE CASMIN CLASS SCHEMA
It is one of th e main
complaints
of Hout and
Hauser th at th e class
categories
used in our
comparative analyses
of
mobility (Erikson
and
Goldth orpe, 1987a, 1987b, 1992)
are of a too
aggregated
kind and
th at,
in
consequence,
a
good
deal of
mobility
both with in and between
th ese
categories
is concealed.
T h ey
th erefore
proceed
to
re-analyse
our
data-or, rath er,
a
subset
th ereof-using
th e full 12 classes of th e
sch ema to wh ich our data are coded in th e
CASMIN IMS
Superfile
instead of th e
collapsed
seven-category
version th at we ourselves
adopt.
S0rensen,
in
contrast,
is content to
operate
with
only
a
six-category
version of th e sch ema; but
h e objects th at our model of core
fluidity
is
applicable,
if at
all, only
to th e small number
of nations th at we consider and seeks to
demonstrate th is
th rough
an
analysis
of
mobility
in 23
nations, drawing
on th e data-set
constituted
by
Ganzeboom et al.
(1989).
We
would,
first of
all,
wish to assure our
critics
(th ough
we find it odd th at th is sh ould
be
necessary)
th at we would indeed h ave
preferred
to work with more refined class
categories
and with a
larger
number of national
cases. T h e reason
wh y
we did not sh ould be
evident
enough :
it was not
possible
to do so
with out
compromising
th e standards of data
quality,
and
specifically
of data
comparability,
th at we would
regard
as
minimally necessary
for
effective
comparative
research and th at we h ad
expressly
set out to establish and maintain.
Correspondingly,
we can
only regard
our critics
as th emselves
sh owing
a lack of concern for data
quality
th at is
retrograde
and th at vitiates th eir
efforts from th e
very
start.
In our
recoding exercise,
we did indeed use
th e full version of th e class sch ema.
But,
in th e
course of th is
exercise,
it became
apparent
th at
a
satisfactory
standard of cross-national
comparability
could not be ach ieved at th is level.
In some
instances,
essential information was
simply
not available in th e
original
studies: for
example,
to enable th e distinction to be made
between small
proprietors
with
employees (Class
IVa)2
or with out
(Class IVb)
or to make
any
consistent differentiation
among farmers,
wh eth er in terms of
employees (wh ich
Hout and
Hauser
erroneously suppose
to h ave been th e
case)
or of indicators of farm size. In oth er
instances,
wh ile we could in
principle implement
th e sch ema for all our
nations,
we could not in
practice
do th is in th e same
way
from one nation
to anoth er: most
obviously,
in
drawing
th e line
between manual
supervisory
and lower tech nical
grades (Class V)
and skilled manual workers
(Class VI)
and also between th e
h igh er
and lower
grades
of th e service class
(Classes
Iand
II).
T o obviate such
difficulties,
resort to th e
collapsed
seven-class version of th e sch ema
284
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
was-unfortunately-necessary (for
furth er
details,
see Erikson and
Goldth orpe,
1992:
50-2).3
In
placing
our data in th e
public domain,
it
seemed
only
sensible to retain th e more detailed
form,
since some
potential
users
migh t
h ave an
interest in
particular
nations rath er th an in
comparative analysis.
But in
working papers,
conference
presentations,
and
personal
com-
munications,
as well as in th e
major report
on
our work
(Erikson
and
Goldth orpe,
1992:
52),
we h ave
emph asized
th at
only
at th e seven-class
level of th e sch ema can an
acceptable degree
of
data
comparability
be claimed: th at
is,
in th e
sense th at no
major
defects are evident and th at
if,
on th is
basis, statistically significant
cross-
national differences in
mobility
are
revealed,
it
would for th e most
part
be a reasonable
presumption
th at
th ey
are real rath er th an
artefactual.
It is not
apparent wh y
Hout and Hauser
sh ould ch oose to
disregard
th is
repeated message
(th ough
we do
subsequently suggest
a
reason).
However,
th e
consequences
are
inevitably
damaging.
As
th ey accept, th ey simply
h ave to
discard th e
th eoretically interesting
cases of
Hungary
and
Poland-and,
h ad
th ey
worked
with th e full CASMIN
data-set, th ey
would h ave
been forced to discard Australia and th e USA
as well.4 But
furth er,
and more
seriously,
in all
th e
analyses
of
mobility
th at
th ey
undertake
th ey
will of
necessity
introduce some
non-negligible
amount of cross-national 'variation' th at is of
an
entirely spurious
kind.
T h e
problem
of data th at arises with
S0rensen's contribution is
similar,
but still more
severe. T h e
Ganzeboom, Luijkx,
and T reiman
(GLT )
data-set to wh ich h e resorts is one
constructed
following
an
approach
to th e
problem
of data
comparability quite
different
from th at of th e CASMIN
Project. Briefly,
th e
GLT
strategy
is to accumulate
multiple mobility
tables for a
large
number of nations
by relaxing
standards of data
comparability (and also
by
accepting
data from
very
small
samples),
but th en
to
regard
each table as a
'data-point', subject
to
error. In all
analyses
based on th e data-set it is
th erefore
necessary
to ch eck on
possible
effects
due to
error,
at th e same time as substantive
issues are addressed
(Ganzeboom
et
al., 1989).
We h ave serious doubts about th e
viability of
th is
approach ,
at least as so far
implemented
(see
Erikson and
Goldth orpe,
1992: 53, n. 27,
100-1).
But wh at we would wish to
emph asize
h ere is th at S0rensen sh ows no awareness of its
essential
logic.
In
effect,
h e treats th e
mobility
tables th at h e takes over from th e GLT data-
set as if
th ey
were on a
par
with th ose
produced
with in th e CASMIN
Project
and carries out no
'quality
controls' wh atever. His
only
concession
to th e
problem
of data
comparability
is to
reject
tables th at score 0 and 1 on th e
six-point
GLT
quality grading.
T h is
is, h owever, entirely
inadequate. Only
tables
scoring
5 are
directly
comparable
to CASMIN tables
(Ganzeboom
et
al.,
1989:
21);
and it can be sh own th at even
tables
produced
in th e manner th at will
give
a
quality
score of 4
(by mimicking
th e class
sch ema via ISCO
occupational codes)
are
likely
to create
quite
serious
problems
of com-
parability.5 Furth ermore,
examination of
S0rensen's T able IIin th e
ligh t
of th e docu-
mentation of th e GLT data reveals th at th e
subset th at h e h as extracted is
especially patch y:
th us,
for nine of h is 23 nations h e can muster
in
total, only
th ree tables
(out of
35)
th at attain
a
quality
score of
4,
th e remainder
scoring only
3 or 2.
We do not th erefore find it at all
surprising
th at in
analysing
th ese data S0rensen sh ould
sh ow a
degree
of cross-national variation in
relative rates of social
mobility
th at is
quite
incompatible
with th e
FJH-h ypoth esis.
T h ere
is no reason to
suppose
th at
any
oth er outcome
would h ave been
possible.
T h at is to
say,
th e
FJH-h ypoth esis,
in
any version,
will
necessarily
be
rejected
on th e basis of th ese
data, simply
on account of th e lack of cross-national
comparability
inh erent in th em. Even in
using
th e CASMIN
data-set,
we would
argue,
some
'ch arity'
is due to th e
h ypoth esis
in virtue of th e
residual
non-comparability
th at is still
present
and th at will th erefore
count-unfairly-against
it
(cf.
Erikson and
Goldth orpe,
1992:
174).
But
S0rensen
appears quite h appy
to
exploit
th e
deficiencies of h is data to th e
advantage
of h is
own case.
As
against
both Hout and Hauser and
S0rensen,
we would th en wish to maintain
th at th e first
requirement
of
any
worth wh ile
285
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
sociological analysis
is th at th e limitations of
th e available data sh ould be
respected.
Hout and
Hauser are kind
enough
to
say
th at th e
CASMIN
project
'h as
produced
a wealth of new
information on
comparative
class
mobility'
(p. 239).
We are
sorry
th at we could not
produce
more. But
noth ing-but
trouble-will be
gained
by proceeding
as if th e data we use are more
detailed or more extensive th an is in fact th e
case.
T o conclude th is
section,
we
append
one
furth er comment
relating
to Hout and Hauser's
argument
for
wish ing
to work with th e full
version of th e class sch ema:
namely,
th at th e
classes of th e
collapsed
version are too h etero-
geneous
in th e
prestige
or status scores of th eir
constituent elements
(pp. 240-41).
T o
repeat,
we too would h ave liked to utilize th e full
sch ema,
h ad
requirements
of data
quality
allowed-but
not,
we sh ould
say,
for th e reason
th at Hout and Hauser h ere advance. T o
judge
th e
validity
of
any
class sch ema
by
reference to
prestige
or status makes little th eoretical
sense,
unless one
supposes,
as we would
not,
th at
classes are
just
more or less
arbitrary
slices of
a
prestige
or status continuum. So far at least
as our sch ema is
concerned,
th e
appropriate
criteria are th ose considered
by Evans,
in h is
paper
in th is
number;
th at
is,
ones
pertaining
to th e different
aspects
of
employment
relations
in terms of wh ich th e th eoretical rationale of
th e sch ema is
expressed (cf.
Erikson and
Goldth orpe,
1992:
35-47).
Evans' results
indicate
th at,
as we h ave
always supposed,
we
do indeed lose
someth ing
in
collapsing
th e
sch ema
(and
especially
Classes V and
VI);
but
prestige
or status
gradations
are not th e relevant
consideration. It would not
surprise
us if th e
latter could
always explain someth ing
more of
mobility
rates and
patterns
over and above wh at
can be
explained
in terms of
class-alth ough ,
we would
argue,
th e reverse is more
obviously
true,
in
particular
with
regard
to structural
effects.
But,
in
any event,
as we noted
earlier,
our own concern h as from th e
beginning
been
with class
mobility;
and th ose wh ose interest
centres, rath er,
on
mobility
in terms of
prestige
or status
surely
h ave th e
option
of
dispensing
with class
categories altogeth er
in favour of
some
appropriate
scale.
T HE CASMIN MODEL
Our critics are dissatisfied with th is model on
a
variety
of
counts,
wh ich we sh all consider in
turn.
T ailoring
Both Hout and Hauser
(p. 255)
and S0rensen
(pp. 268, 278)
contend th at our model of core
social
fluidity
is tailored to fit th e
mobility
data
to wh ich it is
applied,
and in
particular, th ey
believe, th rough
th e
'affinity'
terms th at it
includes. As it
stands, h owever,
th is
ch arge
is
simply
incorrect. T h e model is
tailored,
as we
h ave described in some detail
(Erikson
and
Goldth orpe, 1987a,
1992: ch .
4),
in order to fit
th e
empirical representation
of 'core'
fluidity
th at we derive from our data for
England
and
France
(th e
two nations wh ich
emerged
from a
MDSCAL
analysis
of data for nine
European
nations as th ose most central to th e
'space'
with in wh ich cross-national variation in
fluidity
occurs). But, beyond th is,
no
tailoring
is
undertaken,
and such fit as th e core model
may
ach ieve to
mobility
data for oth er nations cannot
be devalued on th is account.6
It
ough t, moreover,
to be
recognized-and
more
clearly
th an our critics
accept-just
wh at
th e
purpose
of th e above-mentioned
procedure
was: th at
is,
to
provide
an
adequate
test of th e
FJH
h ypoth esis,
taken in its actual verbal
formulation as
claiming only
a 'basic'
(not total)
similarity
in cross-national
fluidity patterns.
T o
th is
end,
it would not h ave been
appropriate
simply
to
postulate
a model of such
basic,
or
core, fluidity
a
priori,
and th en to deem th e
h ypoth esis
to h ave been refuted if th is model
did not fit th e data well. Some
attempt
h ad to
be made to
identify empirically
th e best
'candidate' for such a core
pattern.
S0rensen
writes at one
point (p. 268):
'T h e
danger
is th at
a model such as th at
proposed by
Erikson and
Goldth orpe,
wh ile
coming
close to
fitting
adequately
in all
nations,
none th e less
may
not
provide
a true
representation
of th e
pattern
of
relative
mobility
in a
given
nation.' But a model
of th is kind is
just
wh at one h as to look for in
th e interests of
testing
th e
FJH-h ypoth esis
not
only
as
rigorously
as
possible
but also
fairly-
someth ing
with
wh ich ,
as we h ave
already
286
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
observed,
S0rensen h imself seems rath er little
concerned.
Symmetry
Hout and Hauser dislike our model of core
social
fluidity
because it is
asymmetrical. T h ey
are forced to
acknowledge
th at in a
majority
of
th e nations
th ey
consider
asymmetries
in relative
mobility rates, or,
th at
is,
in th e
pattern
of social
fluidity,
do in fact
exist;
but
th ey
still contend
th at a
symmetrical
model sh ould be favoured
on
grounds
of
parsimony
and
interpretability.
On th e issues of
parsimony,
as
against fit, wh at
would seem
ultimately
to be involved are
differing preferences
th at cannot
perh aps
be
usefully
debated. But on th e issue of inter-
pretability,
th e main
argument
th at Hout and
Hauser advance is one th at we must
regard
as
at best
misleading
and at worst
just wrong.
T h eir claim
(p. 249)
is th at 'Intrinsic
asymmetry [i.e. asymmetry
in
fluidity patterns]
is
incompatible
with th e
analysis
of structural
mobility'.
T h is statement
needs, h owever,
to be
glossed
as follows:
'accepting asymmetry
in
fluidity patterns
is
incompatible
with
making
th e
distinction between structural and
exch ange
mobility, according
to th e model
proposed by
Sobel et al.
(1985).'
T o wh ich we would th en
reply:
so much th e worse for th e model. It h as
th e
evidently unsatisfactory property
th at its
applicability depends upon
a
purely contingent
fact: i.e. th at
fluidity
is
symmetrical; and,
in
any
event,
we would
regard
th e distinction between
structural and
exch ange mobility
as
by
now
outmoded and
unnecessary (cf.
Erikson and
Goldth orpe,
1992:
58-9).
At th e same
time, th ough ,
wh at must be
stressed is th at it remains
entirely possible
to
work with an
asymmetric
model of
fluidity,
such
as our core
model,
and still to make estimates
of structural effects on
(absolute) mobility
rates.
Indeed,
we believe th at th is can most
usefully
be done
by treating
th ese
effects, following
Sobel, Hout,
and Duncan
(SHD),
as
reflecting
ch anges
between
origin
and destination
distributions th at raise or lower th e odds of
mobility
to a
given
destination
by
th e same
factor across all
origins
alike. T h is valuable
suggestion is,
one sh ould
note, quite
detach able
from th e SHD model itself. All th at is
required
is a
recognition
th at
h eterogeneity
in th e
marginal
distributions of th e
mobility
table
may
derive in
part
from th e
pattern
of
fluidity itself
or,
in oth er
words,
th at th e latter
may
not be
symmetrical.
It is of interest h ere to note th at in th e
paper
proposing
th eir
model,
SHD do at one
point
sh ow some awareness th at
'greater generality'
migh t
be ach ieved in th e
way
in
question:
th at
is, by
not
requiring
th at
marginal h eterogeneity
be
completely
accounted for
by
structural effects
(Sobel
et
al.,
1985:
365, n.
2). T h ey
th en dismiss
th is idea on th e
grounds
th at if
symmetry
in
relative rates does not
h old,
structural effect
parameters
will not be
(uniquely)
estimable. But
th is would
appear
to be a
misconception.
At all
events,
with th e kind of
topological
model th at
we
develop,
th e
objection
does not
apply.
Structural effects can be
quite satisfactorily
estimated, regardless
of th e
asymmetry
in
fluidity
th at is
specified.7
Finally, th en,
we must
point
out th at wh en
Hout and and Hauser write
(p. 240)
th at
'Erikson and
Goldth orpe give
almost no
attention to structural
mobility',
th is is true
only
in th at 'structural
mobility',
like
'exch ange
mobility',
is not a
concept
th at we h ave use for.
But,
as readers
may
see for
th emselves,
we do
discuss structural effects on
mobility quite
extensively (1992:
ch . 6
esp.
but cf. also ch s.
3,
7, 9, 10), including
in th e manner th at Hout and
Hauser favour.
Indeed,
th e
findings
th at
th ey
report
in th is
respect (pp. 249-51)
are ones th at
we h ave in all salient features
anticipated
(Erikson
and
Goldth orpe,
1992:
204-7).
In
sum,
th e
implication
contained in Hout and Hauser's
paper
th at we are unable to treat structural
aspects
of
mobility
because our model of core
fluidity
is not
symmetrical
must be
flatly
rejected.
Cross- Validation
Hout and Hauser
point
to th e
desirability
of
validating
our model
against
fresh
data,
rath er
th an
by comparing
its fit with th at of oth er
models
against
CASMIN data. Wh ile we would
not
accept
th at th is
desirability
stems from th e
tailoring
of th e model to th e CASMIN
tables,
to
attempt
cross-validation in th e
way
th at Hout
and Hauser
propose
is still an
obviously
287
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
T ABLE 1 Results of fitting
th e model
of
core social
fluidity
to
mobility
tables
for
men
aged
20-64 in
England and
Wales
G2 df
p G2(Sa N
CASMIN (1972) 68.3 28 0.00 1.7 37
9434
BGES (1979, '83, '87) 30.2 28 0.36 2.2 29 3
012
Effect
parameters
HI1 HI2 IN 1 IN2 IN3 SE AF 1 AF2
CASMIN -0.16 -0.35 0.47 0.71 0.77 - 1.22 -0.76
0.44
BGESb -0.12 -0.30 0.60 0.41 1.79 -0.77 -0.22
0.48
Structural sh ift effects
(Class I+ IIset at 0)
III IVa + b IVc V + VI VIIa VIIb
CASMIN -0.70 -1.19 -3.16 -1.25 -1.20
-2.51
BGES -0.78 -0.91 - 2.82 - 1.34 - 1.30 -2.01
Notes:
aG2(S)
is G2 standardized for a
sample
size of N= 1
991,
th e smallest in th e CASMIN data-set. For furth er
details
of th is
statistic,
see Erikson and
Goldth orpe (1992: 88-90).
bT h e
parameters
for HI1 and AF1 do not attain
significance.
appropriate procedure.
T h eir own
exercise,
on
th e basis of
mobility
data for Britain taken from
Marsh all et al.
(1988), is,
as
th ey recognize,
inconclusive, primarily
because th e
sample
N is
too small.
Fortunately, th ough ,
th ere are oth er
data
available,
somewh at more
adequate
to th e
purpose
in h and:
namely,
th ose from th e British
General Election
Surveys
of
1979, 1983,
and
1987.
In T able 1 we sh ow th e results of
fitting
our
model of core social
fluidity
to a
mobility
table
for men
aged
20-64 in
England
and Wales th at
is derived from th e
pooled
data of th ese th ree
studies. Results
previously reported
for th e
corresponding
CASMIN
(1972)
table are also
given
for
purposes
of
comparison.
It can be seen
th at th e fit of th e model to th e 1979-87 data
is
very satisfactory,
and th at estimates of effect
parameters,
including
'structural sh ift'
para-
meters,
are for th e most
part
rath er close to
th ose for 1972. In so far as differences do sh ow
up, th ey
are ones th at are
scarcely implausible
or indeed th at
point
to
ch anges
over time th at
can be
independently documented,
such as a
relative decline in th e industrial
working
class
and a relative
growth
in th e
petty bourgeoisie.
In th e case of th e
affinity effects,
about wh ich
Hout and Hauser
express particular doubts,
it
is true th at AFl-th e
disaffinity
term
applying
to
mobility
between th e service class and th at
of
agricultural workers-becomes
insignificant
(th ough
with its
expected sign)
for 1979-87.
However, it sh ould be noted th at even with a
total N of over 3
000,
th e numbers associated
with Class
VIIb,
as one of
origin
or
destination,
are still
very
small. T h e AF2
term,
wh ich
applies
to far more
populous
cells of th e
mobility table,
is little altered.8
We would th en claim
th at,
for
England
at
least,
our model stands
up
well to a validation
exercise of th e kind th at Hout and Hauser
propose,
and th at th is result serves furth er to
undermine
any suggestion
th at th e success of th e
model in
reproducing mobility
data is ach ieved
by tailoring.9
Performance
Hout and Hauser and S0rensen alike aim
to sh ow th at models th at
th ey
advance
perform
better th an wh at
th ey
refer to as eith er
th e 'CASMIN' or
'Erikson-Goldth orpe'
model.
However,
as we would h ave
th ough t
abundantly clear,
our model of core social
fluidity
is one
specifically
tied to th e seven-
category
version of our class sch ema
(Erikson
and
Goldth orpe,
1992:
140).
And
since, th en,
neith er Hout and Hauser nor
S0rensen
work with th is version of th e
sch ema,
it must
be
pointed
out th at
th ey
are
simply
not in a
position
to make th e
comparisons
th at
th ey
claim and th at take
up
so much of th eir
papers.
288
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
S0rensen,
as we h ave
noted,
is restricted
by
th e GLT data-set to
only
a
six-category version
of th e sch ema: th e classes of farmers and of
farm workers
(IVc
and
VIIb)
are
collapsed.
As h e th en
acknowledges-th ough
in a
rath er
piecemeal way-th is
means th at th e
model h e
subsequently deploys
as th e 'Erikson-
Goldth orpe'
model is in fact
quite substantially
different from our model.
First,
of
course,
th ere
can be no
separate
effect for
immobility
with in
th e class of farmers
(IN3); secondly,
th e sector
effect
(SE)
is
completely lost;
and
th irdly,
th e
disaffinity
effect
(AF1)
cannot be retained in its
original
form. In view of
th is,
and of th e
importance
th at th e IN3 and SE effects in
particular prove
to
h ave,
we find S0rensen's
argument (p. 270)
th at h e can still
provide
a
'fair test' of our model difficult to
accept. But,
as we h ave
already observed,
h is standards of
fairness and ours are rath er far
apart.10
With
regard
to Hout and
Hauser,
th ere
is,
fortunately,
more th at can
usefully
be said. In
fitting mobility
tables constructed on th e twelve-
category
version of th e class
sch ema,
Hout and
Hauser
simply apply
our
model, designed
for
seven-class
tables,
with out
any
modification. We
do not understand wh at
justification
th ere can
be for
following
th is
procedure.
If we h ad
th ough t
it
possible
to obtain
adequate
cross-
national
comparability
at th e level of th e full
sch ema,
we would
surely
h ave
developed
a
model of core
fluidity
th at reflected th is
greater
degree
of differentiation.
Correspondingly,
if
th e aim is to assess th e
performance
of th e
'CASMIN model' in relation to twelve-class
tables
(accepting
th ese now
simply
for th e
purposes
of th e
argument),
th is must be done
on th e basis of a new version of th e model th at
is
appropriately
extended
according
to th e
th eoretical ideas
underlying
th e
original (see
Erikson and
Goldth orpe, 1987a;
1992: ch .
4).
Such a new version
could,
for
example,
be
proposed
on th e lines set out below.
(T h e
actual
levels matrices for th e model are
provided
in th e
Appendix
to th is
paper.)
Hierarch y.
T h e h ierarch ical divisions made
with in th e class
sch ema,
and understood as
reflecting
differences in
job
rewards and
requirements,
are increased with th e twelve-class
version from th ree to
five.1'
T h ere are th us
now four rath er th an two
h ierarch y effects,
HI1-HI4,
each
referring
to th e
successively
h igh er
barriers
confronting mobility as
transitions over
one, two, th ree, or four
h ierarch ical divisions are involved.
1
2
3
4
5
original
I+11
III, IVa+b, IVc+ d (D),
V + VI
IVc+d(O), VIIa, VIIb
extended
I
II, IVa
IIla, IVb, V
IIIb, VI, IVc
IVd, VIIa, VIIb
Inh eritance. T h e number of inh eritance
effects,
th at is effects
referring
to
propensities
for class
immobility,
remains at
th ree;
but with th e
twelve-class sch ema
greater
discrimination can
be sh own
regarding
th e
possibilities
for th e inter-
generational transmission of
property
in th e
placing
of classes on th e second level.
original extended
IN1 all all
IN2 I+
II, IVa +
b, IVc+d I, IVa, IVc, IVd
IN3 IVc+d IVc
Sector. Wh ile th e effect for sectoral barriers to
mobility
between
agricultural
and non-
agricultural
classes
(SE1)
is
retained,
th e twelve-
class sch ema allows for a furth er such effect
(SE2)
to be
introduced, referring
to barriers to
mobility
between classes in th e small business
sector
(IVa, IVb, IVc, IVd)
and th e rest.
Affinities.
T h e affinities included in th e
original
model are
essentially
retained in th e extended
model, alth ough ,
with th e twelve-class
sch ema,
modifications can be made th at
bring
th eir
implementation into rath er closer accord with
th e th eoretical ideas th at underlie th em.'2 One
additional
affinity
is
included, relating
to th e
propensity
for
mobility
between classes IVc and
IVd,
th ose of
'large'
and 'small'
farmers,
wh ich
migh t
be
expected
to be
h igh er
th an
provided
for
by
oth er effects on account
simply
of fluctuation in th e economic for-
tunes of farm families from one
generation
to
anoth er.
289
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
AF1
(disaffinity)
AF2
(status affinity)
(capital
affinity)
(agriculture
outflow
affinity)
original
I+ II-VIIb
I+ I1-III
V + VVIVIa
I+ IIVa + b
IVa + bIVc + d
IVc+d,
VIIb-VIIa
extended
AF1
I, II-I-Vd,
VIIb
AF2
I, II, IIIa, IIIb
(all transitions)
V, VI, VIIa
(all transitions)
AF3
I, IVa,
IVc
(all transitions)
AF4
IVc, IVd, VIIb-VIIa
AF5 IVc,-IVd
In T able 2 we sh ow in th e first row th e results
of
fitting
our extended model to th e same collec-
tion of national
mobility
tables th at Hout and
Hauser derive from th e CASMIN data-set. In th e
event,
th e
parameter
for th e HI1 term
proved
to
be
insignificant,
so th is effect was omitted and
th e model
re-applied,
with th e results th at are
sh own in th e second row. In th e th ird row we
reproduce
th e results
reported by
Hout and
Hauser for th eir own
preferred
model
(i.e.
Model
13 in th eir T able
4).
It can be seen th at our extended
model
performs
better th an
th eirs, according
to
th e bic statistic th at
th ey adopt
as th eir criterion
for model
ch oice,
as also
according
to A.
In th is
way, th en,
most of th e
argument
in
th e central
part
of Hout and Hauser's
paper
is
empirically
undermined. It is
apparent
th at th e
'superiority'
th at
th ey
claim for th eir alternative
model
specification
is ach ieved
only by keeping
th e 'CASMIN model' unaltered wh ile
moving
to th e full version of th e class sch ema as th e basis
for th e
mobility
tables
analysed.
T h e
supposed
superiority
vanish es once our model of core
fluidity
is
extended, following
its
underlying
th eoretical
rationale,
so as to
apply
to twelve-
class rath er th an seven-class tables. Moreover,
if we turn to th e
parameter
estimates for our
extended model
given
in th e lower
panel
of
T able
2,
doubt is also th rown on Hout and
Hauser's more
specific
contention th at in our
analyses
we understate th e
importance
of
h ierarch y
effects on
fluidity patterns
as
against
th e effects of inh eritance and sector. We must
stress th at we do not believe th at
any great
substantive
importance
sh ould be attach ed to
th ese estimates because of th e deficiencies of th e
data wh en th e full version of th e class sch ema
is utilized. None th e
less,
th e indication is
th at,
if we did h ave
good quality
twelve-class tables
available for
analysis,
th e relative
importance
of th e effects in
question
would not in fact turn
out to be all th at different from wh at we h ave
found in
considering
seven-class tables.13
Modelling h ierarch y effects
Hout and Hauser are
greatly
concerned to sh ow
th at
part
at least of th e reason
wh y
we
underestimate
h ierarch y
effects is because we
model th em
wrongly.
As well as
working
with
discrete h ierarch ical levels rath er th an with
T ABLE 2 Results
of fitting
selected models to seven CASMIN
mobility
tables for men
aged
25-64
(N=37857)
G2 df rG2a A
bic
Extended
(12-class)
core model 1 978 833 88.7 7.4 -6 803
with HIl omitted 1979 834 88.7 7.4 -6813
Hout and Hauser's Model 13 2455 830 86.0 8.0 -6294
Effect
parameters
under extended core model
(HI1 omitted)
HI2 HI3 HI4 IN1 IN2 IN3 SE1 SE2
-0.25 -0.42 -0.44 0.58 0.59 1.34
-
1.01 -0.30
AF1 AF2 AF3 AF4 AF5
-0.35 0.33 0.20 0.70 1.00
Note: arG2 is th e
percentage
reduction ach ieved in th e G2 for th e model of conditional
independence.
290
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
T ABLE 3 Results
of fitting
th e extended core model to seven CASMIN
mobility
tables
for
men
aged 25-64, with
different specifications of h ierarch y effects (N=37857)
G2 df rG2 A bic
Extended
(12-class)
core model 1979 834 88.7 7.4
-6813
HI2, HI3, and HI4
replaced by
linear-by-linear 'prestige'
2 150 836 87.7 7.7 - 6663
HI2,
HI3 and HI4
replaced by
linear-by-linear 'status' 2 149 836 87.7 7.5 -6 664
HI2,
HI3 and HI4
replaced by
'status' and
'prestige'
2014 835 88.5 7.4 -6789
status or
prestige continua,
we also
adopt
a
'social-distance' rath er th an a
'linear-by-linear'
specification
of
h ierarch y
effects wh ich
is,
in
th eir
view,
a
conceptual
error.
We would
point
out
again
h ere th at our
concern is with class
mobility
and
th at,
in
th is
perspective, h ierarch y
effects are
quite
appropriately
seen as
relating
to h ierarch ical
'distances' between class
categories;
wh ereas th e
kind of
modelling
favoured
by
Hout and Hauser
would
appear
more
apt
wh ere interest centres
specifically
on
mobility
with in a
prestige
or
status order.
But,
in
any event,
as Hout
and Hauser
righ tly observe,
th e best
way
of
attempting
to
adjudicate
between such
differing
conceptual approach es
is to see wh at
h appens
wh en
th ey
are
applied
to data.
In T able 3 we sh ow th e results of
fitting
our
extended model of core
fluidity
to th e same set
of tables used
by
Hout and
Hauser,
but with
th e effective
h ierarch y
terms in th e model
(HI2,
HI3,
and
HI4) being replaced, first, by
Hout
and Hauser's
linear-by-linear occupational
prestige term; second, by
th eir
linear-by-linear
socio-economic status
term;
and
th ird, by
both
of th ese terms. It can be seen
th at,
wh ere eith er
prestige
or status alone
replaces
our
h ierarch y
effects,
th e bic value returned is
clearly
inferior
to th at sh own in T able 2 for our extended model
in its unaltered form: and th at th e latter value
is still not reach ed wh en status and
prestige
are
entered
togeth er.14
Again, th en,
it is th e case th at Hout and
Hauser's
argument
comes
apart
once th e
'CASMIN model' is
represented
in a form
proper
to th e
mobility
tables to wh ich it is
applied.
With in th e context of our extended
model,
th ere is no indication th at th e meth od
of
modelling h ierarch y
effects th at
th ey
advocate
would
give any advantage.
Had we th e data th at
would allow th e serious
comparative analysis
of
twelve-class
tables,
we could
expect
to do at least
as well
using
th e meth od th at is consistent with
our basic
conceptual approach .
In th is
way
more-
over,
we would be sure of
avoiding
th e rath er
bizarre results th at Hout and Hauser obtain
wh en
th ey
consider cross-national variation
under a model in wh ich both status and
prestige
are included. We
are,
it
appears,
asked to believe
th at in some nations differences in
prestige
between classes
(net
of all oth er
effects) actually
facilitates
mobility
between th em: th e
greater
th e
difference in
prestige,
th e
greater
th e
propensity
for
mobility
to occur!
SUBST ANT IVE RESULT S
In th is section we turn to th e substantive results
of th e
analyses
th at our critics
present,
aimed
primarily
at
evaluating
th e
FJH-h ypoth esis,
and
th at
th ey
would
regard
as
improvements
on
our own efforts. T o start
with , th ough ,
we summarize th e
findings
we h ave
already
reported
in th is
respect (see
furth er Erikson and
Goldth orpe,
1992: ch . 5 and 374-9
esp.),
since
th ese do not
always
seem to be
fully
understood.
In
testing
th e FJH
h ypoth esis strictly
construed,
th at
is,
as
represented by
th e common
social
fluidity (CmSF) model,
we
find,
like
almost all
previous investigators,
th at it cannot
be
uph eld.
T h ere are
significant
variations in
relative rates of class
mobility
across th e
291
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
national societies we consider.
However, as
actually
stated
by Feath erman, Jones, and
Hauser
(1975: 337-9),
th e
h ypoth esis
claims
only
a 'basic'
similarity
in relative rates.
T h us,
we
argue, rejecting
th e
h ypoth esis
on th e
grounds
th at th e CmSF model does not fit is
scarcely
an
adequate
test
and,
more
seriously,
risks
th rowing
out th e
sociological baby
with th e
statistical bath -water. If a substantial cross-
national
commonality
in social
fluidity patterns
were to
exist,
th is would
represent
a
major
empirical regularity th at, presumably,
com-
parative macro-sociologists
would wish to know
about. As we h ave
already remarked,
our model
of core social
fluidity
was th en
developed
specifically
in order to make
possible
some
assessment of th e
FJH-h ypoth esis
in its less
strict,
but still
h igh ly consequential
form. In th e
ligh t
of
applications
of th e model to
good
quality class-mobility
data
for,
in
all,
fifteen
nations,
th e main conclusions th at we reach are
th e
following.
T h ere is clear
support
for th e FJH
h ypoth esis,
in its initial verbal
formulation,
in two
respects.
(a)
Our model does
identify
a
large commonality
in
patterns
of social
fluidity
with in national class
structures.
If,
for
example,
one were to
suppose
a counterfactual world in wh ich our model
specified exactly
th e
pattern
of
fluidity pre-
vailing
in all nations
alike, th en,
in such a
world,
absolute rates of class
mobility-total, inflow,
and outflow-would
only rarely diverge
from
th ose
obtaining
in th e real world to a
degree
th at
sociologists
would find
worth y
of comment
(Erikson
and
Goldth orpe,
1992:
210-13).
In
oth er
words,
cross-national
commonality
in
relative
mobility
rates
may
be reckoned as
substantial,
in th at variation at th is
'genotypical'
level is of almost
negligible
effect in
accounting
for variation at th e
'ph enotypical'
level of
actually
observed rates.15
(b)
T h e cross-national differences in social
fluidity
th at do sh ow
up,
in th e form of
deviations from our core
model, appear
for th e
most
part
to be better understood as
specific,
h istorically-formed
variations on th e th eme
defined
by
th is model th an as
representing
variation of a more
systematic
kind. T h at is to
say,
no
great scope
would seem to exist for
explaining
th e extent and nature of differences
in
fluidity by
reference to oth er
generalizable
attributes of
societies,
so th at
(cf. Przeworski
and
T eune, 1970)
th e names of variables migh t
be substituted for th ose of nations. In
being
couch ed in terms of
commonality
rath er th an
variation in relative
rates,
th e FJH
h ypoth esis
th us
points
to wh at would
appear
to be th e
proper
focus of
macro-sociological
attention.
At th e same
time, th ough ,
th e need for two
qualifications
to th e FJH
h ypoth esis
is also
suggested.
(a)
A recurrent source of cross-national
variation in social
fluidity
is found to lie in
political
intervention. T h e FJH
h ypoth esis,
at
least in its
original formulation, applies
to
societies in wh ich th ere is a market
economy
and
some kind of
family system.
But th e effects of
market and
family,
th e two
great pillars
of civil
society,
in
creating
and
perpetuating
differential
mobility
ch ances can
evidently
be modified
by
political
action taken via a modern state
apparatus.
Societies in wh ich such action is
initiated and sustained can th us be
regarded
as
more
likely
to sh ow deviations from 'core'
fluidity
th an are oth ers-even
th ough
it
may
not
be
possible
to associate distinctive
types
of
fluidity pattern
with th e
particular
institutional
or
ideological
features of different
political
regimes.
(b)
T h ere is also evidence to
suggest (Erikson
and
Goldth orpe,
1992:
379-89)
th at one
specific
and
quite
small
(th ough th eoretically interesting)
part
of cross-national variation in
fluidity
could
h ave a
systematic
basis: th at
is,
evidence th at
th e lower th e
degree
of economic
inequality
among
individuals and
families,
th e
greater
th e
overall level of
fluidity
with in a
society.
However,
confirmation of th is
possibility (for
wh ich several
previous investigators
h ave also
argued)
will
not,
in our
view,
be
possible
until
th e construction of
adequate
data-sets on
comparative
economic
inequality
h as
progressed
furth er, as,
for
example,
under th e
auspices
of
th e
Luxembourg
Income
Study.
So,
our conclusions are not
altogeth er
straigh tforward,
but
th ey are,
we
believe,
reasonably
clear in th eir relation to our
empirical
findings.
Wh at of th ose reach ed
by
our critics?
S0rensen
clearly
wish es to
reject
th e FJH-
h ypoth esis
out of h and. His
grounds
for so
292
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
doing
are th at wh en h e
supplements (wh at h e
represents as)
our model of core
fluidity
with
Goodman's II* association
model,
wh ich
freely
scales th e six class
categories
h e
uses,
h e ach ieves
a
significant improvement
in fit to h is
mobility
tables for 23 nations. He th erefore feels
justified
in furth er
exploiting
Goodman's model in order
to
identify inductively
a set of
differing 'mobility
profiles' among
th ese
nations,
defined in terms
of
category scaling
or wh at h e refers to as 'class
cleavages'.
Our fundamental
objection
to S0rensen's
analyses,
as we h ave
already
made
clear,
concerns th e data on wh ich h e draws. T o
repeat,
th e demonstrable lack of cross-national com-
rability
inh erent in th ese data renders th em a
quite inappropriate
basis on wh ich to seek to
test th e
FJH-h ypoth esis,
at all
events,
if
stringent
quality
controls are not
implemented.
In
turn,
we would
suspect
th at S0rensen's
'mobility
profiles'
are in
large part
artefacts of th e
degree
of
non-comparability
in h is
data,
and before h e
seeks
explanations
for th em on th e lines h e
indicates at th e end of h is
paper,
h e
too,
we
suggest,
sh ould
engage
in a cross-validation
exercise.
T h ere are several additional critical
observations th at we would wish to make on th e
results S0rensen
reports,
and wh ich would
apply
even if h is
analyses
h ad a more secure
empirical
basis.
First,
we do not find it at all
surprising
th at
adding
Goodman's II* model to th e so-
called
'Erikson-Goldth orpe'
model sh ould
give
an
improvement
in fit. T h is would no doubt
h ave occurred even if S0rensen h ad worked with
our actual core model and with th e CASMIN
data-set: th at
is,
th e Goodman model would
simply pick up
th e differences in social
fluidity
th at we treat via our national variant models.
But
sh owing
th is
simply
confirms
again
th at th e
FJH
h ypoth esis
does not h old stricto sensu: it
says noth ing
to th e
question
of h ow tenable th e
h ypoth esis
is in th e modified form in wh ich we
sough t
to test it. As we h ave earlier
suggested,
S0rensen
quite
fails to see th e
importance
of
investigating
wh eth er or not a substantial cross-
national
commonality
in
fluidity patterns
prevails
as well as some amount of variation.
Furth ermore,
th e II* model is itself ill-suited
to such a task. S0rensen
argues (p. 268)
th at a
'minimal
requirement'
of th e FJH
h ypoth esis
is th at 'class
cleavages'
in
fluidity patterns
sh ould be located at th e same
places
in all
nations. But th ere is
noth ing
in th e FJH
h ypoth esis
th at
requires
such
cleavages
to fall
in
just
one dimension. Such
unidimensionality
is, h owever, imposed by
th e
scaling
of th e II*
model,
and often with much
consequential
distortion th at S0rensen fails to
recognize.
Consider,
for
example,
h is results for
Poland,
wh ich derive from
good quality
data from
just
one
enquiry.
Poland
emerges (Figure 4)
as an
instance of th e 'lower wh ite-collar elite'
mobility
profile.
T h is sounds
improbable
and it is
improbable. Smallest-space analysis
of th e
Polish table indicates th at
cleavages,
or
distances,
between classes in terms of th e
propensity
for
mobility
between th em cannot
be
captured
in less th an two dimensions. In th e
ligh t
of such
analysis,
it is moreover
apparent
th at th e
emergence
of th e lower wh ite-collar
class at th e
top
of th e scale created
by
model
II* h as
noth ing
wh atever to do with its elite
position
in
post-war
Polish
society,
as
anyone
familiar with th at
society
could
readily
h ave told
S0rensen. It h as to
do, rath er,
with th e
large
distance between th e lower wh ite-collar class and
th e farm
class,
in a dimension th at cannot be
interpreted
as h ierarch ical or 'vertical' in
any
sense.
16
Finally, questions may
be raised as to h ow
securely
S0rensen's results are based in h is
procedures
for model ch oice. T o start
with ,
it
sh ould be noted with
regard
to T able 4 th at th e
mobility profiles
th at
figure
in h is
preferred
model
4,
and on wh ich all h is
subsequent
analysis rests, are,
as h e
explains,
arrived at
inductively:
in
fact,
from results obtained under
h is model 2
(p.
272 and T able
5, p. 275). T h us,
not
only
does model 4 'not fare as well' as
S0rensen would like
(p. 274),
but th e
grounds
for
favouring
it remain unclear: th e
significance
tests
reported
for contrasts between th is model
and model 5 can h ave little
meaning.
In
fact,
S0rensen
appeals
h ere to 'all' conventional
criteria. But
assuming th at,
as seems to be th e
case,
h e counts th e bic statistic
among th ese,
anoth er
major problem
th at h e h as
simply
failed to observe
may
th en be
pointed
out.
Remarkably, given
h is concern with th e FJH
293
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
h ypoth esis,
S0rensen does not fit to h is data th e
model th at h as served
as,
at all
events, th e
regular starting-point
for
testing
th is
h ypoth esis:
th at
is,
th e CmSF model. Had h e done
so, h e
would h ave found th at it returns a
clearly
lower
bic
(-19 1019)
th an h is
preferred
model 4 or
any
oth er h e considers.17 In
sum,
on th e basis
of data of
quite inadequate quality, S0rensen
rejects
th e FJH
h ypoth esis
in
any version,
wh ereas, according
to a criterion of model
ch oice th at h e
regularly invokes,
h e sh ould in
fact
accept
it in its strictest form.
Here,
at
least,
two
wrongs certainly
do not make a
righ t.
T urning
now to th e results
presented by
Hout
and
Hauser,
we find a similar
uncertainty,
and
one
deriving
from similar sources: th at
is,
data
of insufficient
comparability
and th e use of bic.
We h ave
already
underlined th e
unsuitability
for
comparative analyses
of th e CASMIN
mobility
tables in th eir twelve-class
form; h ere, th erefore,
we focus
ch iefly
on th e latter
problem.
T h e model th at Hout and Hauser
prefer
for
purposes
of
comparison
with
(wh at th ey
take
to
be)
th e 'CASMIN model'
(th at is,
th eir model
13)
is one wh ich
provides
for no cross-national
variation in
fluidity patterns.
In oth er
words,
it too underwrites th e FJH
h ypoth esis
stricto
sensu.
Moreover,
wh en Hout and Hauser
come to search for such variation
by adding
appropriate
terms to model
13,
as in th eir
models 14 and
15, th ey
find
th at, according
to
bic,
wh ich h as h ith erto
guided
th eir model
ch oice,
th ese latter models are inferior. 'If we
relied
upon
bic as th e criterion for
including
cross-national interactions'
th ey
write
(p. 259),
'and wanted to
improve
th e fit of model
13,
we
would not add
any
terms to th e model.' But bic
is th eir ch osen
criterion, as,
for
example,
in th eir
rejection
of models with
asymmetries.
And wh at
th e
foregoing
curious sentence
ough t
th en to
mean is th at
th ey
do
accept
model 13 un-
modified: a
distinctly embarrassing position,
to
be
sure,
for
analysts
wh o
regard
th e FJH-
h ypoth esis
as
'long
obsolete'
(p. 263).
It is
relevant h ere to
point
out th at th e bic statistic
was introduced
by Raftery precisely
in order to
provide
'a consistent model selection
procedure'
and 'an automatic
way
of
making
th e often
difficult and
subjective
trade-off between L2
and df wh ich is inh erent in th e conventional
LRT model selection
procedure.' (1986a: 146,
our
emph ases).
In oth er words, th ose wh o ride
with bic are not free to
get
on and off at will.
However,
Hout and Hauser follow th e
sentence
quoted by telling
us th at because of
th eir
'strong prior
interest' in cross-national
variation, th ey
feel
justified
in
deleting only
th e
two weakest sets of cross-national interaction
effects in models 14 and 15 so as to
produce
model
16,
wh ich allows for variation of a more
limited kind.
But, according
to
bic,
model 16
still
gives
a
poorer
fit th an model 13; and since
th e
very point
of bic is to avoid
'subjective'
judgements
in model
ch oice,
Hout and Hauser's
interests, strong
or
oth erwise,
must be
regarded
as
strictly
irrelevant. If
th ey
are to be
consistent,
th ey
are stuck with model
13,
and in turn
with th e FJH
h ypoth esis.
Some
stirrings
of conscience do indeed
arise,
and it is
acknowledged
th at 'model 13 is still
preferable
to model
16',
and caution is
urged
in th e
interpretation
of th e
parameters
for cross-
national variation included in model 16. None
th e
less,
Hout and Hauser
proceed
to discuss
th ese
parameters
and to draw substantive
conclusions from th em. And it is of course at
th is
juncture
th at our own reservations about
results under model 16 would enter in: not
regarding
its inferior bic
(we
h ave no belief in
bic or
any
oth er 'automatic' criterion of model
ch oice)
but
regarding, rath er,
th e extent to wh ich
th e variation th at Hout and Hauser claim
to detect will derive
simply
from lack of
comparability
in th e tables
th ey analyse.18
Hout and Hauser's
paper
th en ends in a
notably
inconclusive
way. Despite h aving
described th e FJH
h ypoth esis
as
'long obsolete',
th ey
state
(p. 263)
th at th eir
results,
from model
13,
'strength en
th e case for basic
similarity'
in
fluidity patterns-but
almost
immediately go
on
to add th at th e notion of basic
similarity
is not
one on wh ich
th ey
'wish to lean too
h eavily'.
Scientific caution in th e face of a difficult issue
is indeed
very proper.
But it is still
pertinent
to
ask h ow far Hout and Hauser's difficulties h ere
are of th eir own
making.
As
Bish op Berkeley
once remarked: 'First we h ave raised a
dust,
and
th en we
complain
we cannot see'.
In
sum,
we
can,
we
believe, reasonably
claim
th at our own
results, clearly rejecting
th e FJH
294
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
h ypoth esis
in its strict version but
emph asizing
still th e extent of cross-national
commonality
in
fluidity patterns
and th e
nationally specific
nature of much of th e
apparent variation, are
more coh erent th an th ose of our critics and also
h ave a more secure
grounding
in data and in
modelling
alike.
T HE AMERICAN DREAM
In th is final section of our
paper
we turn from
responding
to our critics to
try
to elaborate th e
conceptual presuppositions
and related research
programme
from
wh ich ,
we
believe,
th eir
critiques
derive. We do
so,
we would
stress,
not
with th e idea of th us furth er
undermining
th eir
arguments. Rath er,
our aims are to make more
sense of th e somewh at
disparate,
and often
quite
tech nical,
issues on wh ich
th ey
h ave ch osen to
concentrate,
and to move attention on to wh at
we would see as
underlying questions
of a more
interesting,
because more
sociological,
kind.
In a memorable
paper,
Duncan
(1968: 675)
expressed
th e view th at th e work of
Cooley
and
Sorokin
provided 'conceptual
orientations'
more suited to th e
study
of
'specifically
American social stratification' th an did th at of
auth ors such as
Weber, Veblen,
or Marx.
From
Cooley
and
Sorokin,
Duncan took an
understanding
of social stratification in wh ich
th ere were two
key
elements:
first,
a
'h ierarch y
of
inequality',
wh ich could be most
appropriately
treated as one of
occupational prestige
or 'socio-
economic'
status; and, second,
th e movement
of individuals with in th is
h ierarch y
over time
and
especially
between
generations.
T h e less
th ere was of such
mobility,
th e more 'stratified'
a
society was,
and vice versa. In th is
perspective,
th en,
research on
mobility
or
immobility
was
to be
essentially equated
with research on
stratification: rates of
mobility
or
immobility
constituted 'th e stratification
process'. Duncan,
it sh ould be
noted, explicitly
set th is
emph asis
on
process against
wh at h e saw as th e current
preoccupation
in American
sociology
with
structural features of stratification.
Questions
of 'th e structure of differentiation'
and,
more
specifically,
of
mobility
in relation to class
structure and class-based action were in fact
excluded from h is
range
of concern (Duncan,
1968:
685, 694).
From th is
position,
th e
programme
of 'status
attainment' research th en
naturally followed.
T h e focus of interest sh ifted from
mobility
flows
and
propensities
with in a
particular
structural
context
(wh eth er
th at of class or status
groups),
to th e relative
importance
of ach ieved and
ascribed ch aracteristics in
determining
th e
placement
of individuals with in th e
'h ierarch y
of
inequality'; and, correspondingly, regression
tech niques
took over from tabular
analysis.
In th e course of
time, h owever,
th e status
attainment
approach
itself became
subject
to
criticism: for
example,
for its
neglect
of
structural constraints on individual attainment
and for its
tendency
to
suppose
th at th e same
'regression
rules'
governed
attainment across all
sections of national
populations
alike
(cf.
Bielby, 1981). T h us,
an awareness of th e
advantages
of a tabular
approach returned,
and
was
greatly strength ened
as th e
emergence
of
log-linear modelling h elped
resolve
many
of th e
tech nical
problems
with wh ich th is
approach
h ad
previously
been encumbered.
It
is,
we
believe, h elpful
to 'situate' our critics
with in th is h istorical
background.
Hout and
Hauser th emselves must indeed be
regarded
as
key actors,
in
particular th rough
th eir efforts
at
reconciling
and
combining
th e traditions of
status attainment research and
mobility
table
analysis (e.g. Hauser, 1978; 1984a,
Feath erman
and
Hauser, 1978; Hout, 1984; 1989,
Sobel et
al., 1985). However,
our
argument
is
th at,
such efforts
notwith standing,
th e
'conceptual
orientations' favoured
by
Duncan in 1968
are still th ose dominant
among
American
sociologists currently engaged
in th e
study
of
stratification;
and
furth er,
th at wh ile Duncan
h imself
adopted
an
explicitly 'exceptionalist'
stance,
a
leading
concern of th ese
sociologists
today
is to take th ese orientations as th e basis
not
only
for
American,
but also for
comparative
inquiries.
More
specifically,
we would maintain th at
wh ile th e
analysis
of
mobility
tables is com-
monly practised,
it tends to be concentrated on
one main issue: th at of 'h ow destinations
depend
on
origin': precisely
th e issue
posed by
Duncan
in a later
paper, suggesting
models for such
295
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
tables th at
'emph asize
th e "ach ievement" rath er
th an th e
"mobility" point
of view h ith erto
taken'
(1979: 793).19
Wh at is in fact
pursued in
tabular
analysis
is th e
representation
of th e
dependence
of destinations on
origin th rough
a limited number of
parameters
and ones th at
relate,
as far as
possible,
to 'vertical' effects.
In
comparative work,
th ese
parameters
th en
become th e
dependent
variables of
primary
interest,
and
explanations
for th e variance
th ey
display cross-nationally (and
also
perh aps
over
time)
can be
sough t
in terms of
independent
variables
pertaining
to
generalizable
attributes
of societies and th us
capable
of
'replacing
th e
names of nations'
according
to th e Przeworski-
T eune
proposal (cf. Grusky
and
Hauser, 1984).
T h is, th en,
is th e 'American Dream' to wh ich
we refer: th e
possibility
th at th e
mobility regimes
of modern societies can be
parsimoniously
modelled with in a
primarily
h ierarch ical
perspective,
and th at cross-national variation in
th e
strength
of th e effects th us identified can
in turn be accounted for via multi-variate
analyses
in wh ich nations are th e units. We do
th erefore
appreciate
th at for th ose wh o sh are
in th is dream th e CASMIN
project
is
likely
to
be
disturbing
in at least two
respects: first,
because it reflects an
agenda
for
mobility
research of a
significantly
different
kind;
and
second,
because th e results it h as
produced
would
suggest
th at th is American
Dream-just
like th e better-known
one-may
well remain a
largely
unfulfilled
aspiration. Recognition
of
th is clash of research
programmes does,
we
believe, h elp
make more
intelligible
th an would
oth erwise be th e case th e
particular pre-
occupations
th at our critics
display.
Most
obviously,
one can in th is
way
appreciate
th eir concern to re-establish some
conception
of
h ierarch y
as th e main dimension
with in wh ich
mobility
is to be
analysed.
In th e
class structural
perspective
on
mobility
th at we
ourselves
adopt, h ierarch y
need not be
specially
privileged.
Class structures do of course h ave
important
h ierarch ical
aspects, reflecting
th e
differentiation of levels and kinds of resources
among
class
positions. But at th e same time
much class
mobility
h as to be
recognized
wh ich
is not
usefully
ch aracterized in
any
'vertical'
sense,
even
th ough clearly consequential
socially-in entailing major ch anges
in th e
relations in wh ich individuals are involved-
and
perh aps politically
also
(cf. Erikson and
Goldth orpe,
1992:
30-5). However, to th e extent
th at such a
perspective
is maintained, th e
relevance of th e 'ach ievement point of view' is
evidently
undermined. And
h ere, we would
suggest,
lies th e source of Hout and Hauser's
leading
concern to
argue
th at our core model
underestimates
h ierarch y
effects relative to
oth ers it
comprises and, th rough
th eir
reanalysis
of our
data,
to redress th e balance in th is
respect.
In
turn,
we would in th is
way
also seek
to understand th eir
strange assumption
th at th e
criterion for
validating
th e
categories
of our class
sch ema sh ould be
prestige
or status
h omogeneity
and, furth er,
wh at we earlier found most
puzzling
of all: th eir
eagerness
to base th eir re-
analysis
on th e full version of th e class
sch ema,
despite
our
repeated warnings
of th e unwisdom
of such a
procedure. By
so
doing,
th e effect
of th e
prestige
or status differentiation of
classes in
sh aping fluidity patterns may
be
allowed to sh ow
up
to a maximum
extent,
so
th at,
as Hout and Hauser in fact
put it, th ey
can
'improve
estimates of th e size of h ierarch ical
effects and cross-national differences'.
Finally,
we
may note,
Hout and Hauser's desire to
reassert Duncan's 'ach ievement
point
of view'
becomes most overt-and makes sense of
wh at would oth erwise seem much ado about
noth ing-in
th eir
length y
discussion of th e
merits of th e
linear-by-linear
as
against
th e social
distance
specification
of
h ierarch y
effects. T h e
former
specification
is
preferred
because it
'accords more
closely
with th e
prevalent
con-
ceptualization
of
stratification processes'
(pp. 251-2,
our
emph asis) wh ich ,
it h as th en to
be understood
(p. 253),
is one focused not on
th e issue of h ow
mobility
decreases as th e
distance between
categories
increases but 'more
fundamentally'
on 'h ow destinations
depend
on
origins'
.20
Similarly,
it
may
be
pointed
out th at in
S0rensen's
analyses
th e central notion of 'class
cleavages'
is not used in th e
essentially
relational
sense of th e
th eorists,
such as
Parkin,
wh om
S0rensen invokes. It h as in fact no oth er
reference th an to
quite
abstract 'distances'
between classes derived from th e
propensities
296
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
for
mobility among
th em. In oth er
words, even
wh ile
using language
th at
migh t suggest
a class
structural
perspective,
S0rensen too is in effect
seeking
to
reimpose
th e idea of a
single
'h ierarch y
of
inequality' by
wh ich
fluidity
patterns
are
crucially sh aped.
Oth er
auth ors,
also
working
with versions of Goodman's model
II* and th e same derived dimension as th at of
S0rensen's 'class
cleavages',
h ave indeed
quite
explicitly given
th is a h ierarch ical
interpretation
in terms of socio-economic status
(see e.g.
Hauser, 1984a;
Ganzeboom et
al., 1989);
and
it would at all events h ave been less
misleading
h ad S0rensen done likewise.
T h e second main
way
in wh ich
recognition
of th e discordance between th e CASMIN
project
and th e 'American Dream' can illuminate th e
reactions of our critics is in
regard
to
modelling
strategy.
For
us,
as we h ave
indicated,
th e aim
was first of all to see if it were
possible
to
identify
th e cross-national
commonality
in
fluidity patterns implied by
th e
FJH-h ypoth esis
in its less strict
form;
and
th en,
if
so,
to model
national variations on th e 'th eme' th us
defined,
wh ile
leaving open, initially
at
least,
th e
question
of wh eth er or not such variation
migh t
be
treated as
systematic.
In oth er
words,
we h ad
no
prior
commitment to th e
viability
of th e
Przeworksi-T eune
programme
in
comparative
mobility
research but
emph asized, rath er,
th e
dangers
of its
premature adoption
and th e need
for
sensitivity
to h istorical
specificities
and
singularities (cf.
Erikson and
Goldth orpe,
1992:
60-3).
Such an
approach -and
th e results th at
h ave in fact followed from it-are th en
scarcely
congenial
to th e idea th at it is on
variation,
rath er th an
commonality,
in
fluidity patterns
th at
macro-sociological
interest sh ould
focus,
and th at th is variation
can,
in its
essentials,
be
both
parsimoniously expressed
and
systematic-
ally
accounted for.
Hence,
we would
suggest,
comes our critics'
particular dislike of th ose features of our model
of core
fluidity
th at
clearly
do not make for
parsimony:
th at
is,
its
asymmetries
and also th e
affinity
terms th at are intended to
capture
effects on
mobility deriving
from
specific
linkages
or discontinuities between classes and
operate
th us to offset or to reinforce effects
of a more
generalized
kind
(Erikson
and
Goldth orpe,
1992:
128).
And in th is
way too we
may
th en understand th e concern of Hout and
Hauser,
in
re-analysing
our
data, to
'simplify
th e model'
(p. 240),
and of S0rensen to
reject
th e
very
idea of a core model in favour of
seeking inductively
to establish variation in
fluidity
in one dimension wh ich
'h opefully
...
will be
systematic
to a certain extent'
(p. 278).
Finally
from th is
standpoint,
we
may
th row
ligh t
on one furth er matter th at could oth erwise
remain
perplexing:
th at
is,
th e
predilection
of
our critics for th e bic statistic as th e criterion
of model
ch oice, despite,
as we h ave
seen, th e
embarrassment th at it turns out to cause
th em. A
difficulty long
encountered
by h igh ly
parsimonious
models of
mobility
framed with in
a h ierarch ical
perspective
was th at
th ey
failed
to fit th e data
according
to conventional
standards
(see
furth er Erikson and
Goldth orpe,
1989).
In th is
situation, th erefore,
th e intro-
duction of bic
(Raftery, 1986a, 1986b)
came to
devotees of such models as a kind of deliverance.
On account of its rath er
gross
bias in favour of
parsimony
as
against fit,
bic could be relied
upon
to
select,
and
legitimate,
models of
mobility
with
relatively
few
parameters,
even
wh en
by any
oth er criteria
th ey
were far from
reproducing
th e data to wh ich
th ey
were
applied.
T h e attraction of bic to th ose
pursuing
th e
'American Dream' is not th erefore difficult to
appreciate,
and it was in fact
rapidly accepted
among
th em.
However,
wh at
th ey appear
not
to h ave foreseen is th at once th eir
larger
purposes
were taken into
account,
bic would
be
likely
to
prove
a
dangerously two-edged
instrument. Its bias in favour of
parsimony
means th at it can indeed be
readily
used in
attempts
to
'simplify'
models th at aim to
capture
th e more detailed
aspects
of
fluidity patterns-
as,
for
example, by
Hout and Hauser wh en
th ey
seek to eliminate
asymmetries
from our core
model. But th e
point
th en is th at if bic is
insensitive to th e loss of fit th at is th us
brough t
about,
it will likewise tend to be insensitive
to th at wh ich results
wh en,
in
comparative
analyses,
models
embodying
th e FJH
h ypoth esis
stricto sensu are
imposed,
as Hout and Hauser
find to th eir
apparent discomfiture and as
S0rensen
manages
to overlook. T h e
message
from
bic,
used as th e means of a 'consistent
297
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
model selection
procedure',
would
appear
to be
th at cross-national variation in social
fluidity,
far from
being
th e
proper
focus of macro-
sociological attention,
is best
disregarded.21
We h ave
not,
we must
stress, any
wish to
argue
th at th e 'American Dream' sh ould be
abandoned. It
represents
a research
programme
th at must be left to take its
course; judgements
as to wh eth er it is
progressive
or
degenerative
in nature will be formed in th e fullness of
time. Here we h ave
only two, concluding,
observations to make.
First,
it
appears
to us
th at,
so far at
least,
its
proponents
are
working
h ard to find a world th at fits th eir
models;
we
would
prefer
to set our own efforts in th e
reverse,
more
modest,
but
perh aps ultimately
more
rewarding,
direction.
Secondly,
we th ink
th at debate in th e field of stratification and
mobility
research would be clarified and made
more
productive
if th ose
pursuing
th e 'American
Dream' were to
recognize
th at th eirs is not th e
only
research
programme
on
offer,
and th at it
is not in itself a fault for oth er
investigators
to
h ave different interests and
priorities
and
also,
perh aps,
different
conceptions
of wh at macro-
sociology
can and sh ould aim to ach ieve.
NOT ES
1.
Correspondingly,
we would see th e
tendency
for kudos
in th e field to attach
primarily
to
virtuosity
in
modelling,
rath er th an to work on
data,
as one th at creates an
unfortunate incentive
structure,
and
especially
so far
as
younger sociologists
are concerned.
2. T h e class
labels,
based on a combination of roman
numerals and
letters,
enable
ch anges
introduced from
th e earliest seven-class version of th e sch ema
(Goldth orpe, 1980/1987)
to be traced. A
listing
of th e
classes in th e twelve-class version is
given
in th e
Appendix.
3. We are aware th at th e
aggregation
of classes
may
itself
cause
problems
of
comparability
if th e classes involved
differ
markedly
in th eir relative sizes
among
nations
(or
over
time).
We were
especially
concerned with th is
problem
with
regard
to
farmers, wh ich
explains wh y
we
placed
Class IVc + d at a lower h ierarch ical level as a
class of
origin
th an as a class of destination. We would
certainly regard
th is
procedure
as a
pis
aller
(cf.
Erikson
and
Goldth orpe,
1992:
46, 124),
but it is still better th an
supposing
a distinction-between farmers with and
with out
employees-th at simply
does not exist in th e
data. As is evident from th e documentation of th e
CASMIN IMS
Superfile (Erikson
et
al., 1988) th e
distinction between
'large'
and 'small' farmers refers
to
employees
in some nations but in oth ers to amount
and/or
type
of land owned. Moreover, some of Hout
and Hauser's criticisms regarding our treatment of
farmers seem not well
grounded. T h us, for example,
wh en
th ey investigate
th e relative
position
of farmers
among
fath ers and sons via association models (Figure
2), th ey impose
a unidimensional structure on
mobility
propensities,
and fail to
recognize th at th e
presence of
a
strong
sector effect could well h ide a 'vertical' sh ift
(cf.
our
analogous
criticism of S0rensen in th e text
above, p. 317).
T h e
strongest pragmatic argument
for
implementing
a distinction at th e twelve-class level,
despite
th e loss of
comparibility entailed, arises in fact
with Classes Iand II.
4.
Japan could, h owever, h ave been included. Wh y
Hout and Hauser did not
attempt
to use th e full
CASMIN data-set is unclear (as is th eir reason for
eliminating
th e 20-24
age-group
from all th eir
analyses).
5. We are in th is
respect
indebted to Gordon Marsh all for
making available to us th e results of an exercise in wh ich
respondents
to a recent British
survey
were coded to
different versions of th e class sch ema both via th e GLT
procedure
and via our own
algorith m
based on OPCS
codes. T h e
discrepancies
are substantial. Even at th e
level of th e
six-category
version of th e sch ema
employed
by S0rensen,
23
per
cent of all
respondents
are
differently
allocated. We
do, of
course, appreciate
th e
attractiveness of
being
able to
implement th e class
sch ema in a
systematic,
cross-national
way th rough
th e
recoding
of national to ISCO
occupational categories;
and we are also aware th at some of th e
problems
th at
arise with th e initial GLT
procedure
can be overcome
by
th e use of
supplementary employment
status codes.
None th e
less,
we still doubt th at an
adequate
approximation
to th e class sch ema can be arrived at in
th is
way, ch iefly
because th e ISCO
categories are, from
our
point
of
view, too
loosely
defined.
6. Note furth er th at th ese oth er nations include six
-Australia, Czech oslovakia, Italy, Japan,
th e
Neth erlands,
and th e USA-th at were not even included
in th e
original
MDSCAL
analysis.
7. We are
grateful
to Rich ard Breen and Ruud
Luijkx
for
drawing
our attention to th e note in Sobel et al.
(1985)
to wh ich we h ave
referred, and also to th e former not
only
for
allowing
h imself to be
persuaded
th at th e
note was mistaken but furth er for th e
following
demonstration of
wh y
th is is so. T h e ratio of two sh ift
parameters is
equal
to th e ratio of two
expected
frequencies multiplied by
th e inverse ratio of th e
corresponding
cell interaction
parameters,
i.e.
aj= FijX Iji
ai
Fji Iij
wh ere F is th e fitted cell
value,
a is th e sh ift
effect, and
Iis th e cell interaction
parameter
for th e
ijth
cell. T h e
ratios
among
th e fitted values are
clearly
invariant
under alternative
parameterizations, and if th is is also
298
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
th e case with th e ratios
among cell-interaction
parameters (as in fact it
is),
th en it must also be so
with th e ratios
among
sh ift
parameters.
8. We h ave also fitted our core model to
mobility tables
derived from th e
1979, 1983,
and 1987 studies taken
separately.
In each case an excellent fit is
again
ach ieved, but with much smaller Ns (644, 1161, and
1207, respectively)
we encounter similar
problems in
estimating parameters
to th ose faced
by
Hout and
Hauser. Even
so,
in
only
one instance does a
(non-
significant) parameter
sh ow th e
wrong sign,
and in
general
effects do not differ
widely
from th ose
estimated from th e
pooled data.
9. We are
currently engaged
in
attempts
at cross-
validating
our model in th e case of oth er nations.
10. T h e
question
h as to be raised
wh y S0rensen did not
ch oose to work wtih th e same
mobility
tables th at we
h ave
analysed,
as we indeed
suggested
to h im. He
would h ave been restricted to 15 rath er th an 23
nations,
but h e would h ave avoided
major problems
of data
quality
and would th en h ave been able to make
valid
comparisons
between results
deriving
from h is
modelling approach
and our own.
11. As
before, we would seek to validate th ese divisions
by reference to national and international scales of
occupational prestige
or socio-economic status.
T h us,
for
example,
on th e new international scale of
occupational
status devised
by Ganzeboom,
De
Graaf,
and T reiman
(1992),
and used
by
Hout and
Hauser,
we estimate th e
average
score of
positions falling
into
our five divisions as
being
with in th e
following ranges:
I:
70-5; II,
IVa:
50-5; IIIa, IVb,
V:
45-50; IIIb, VI,
IVc:
35-40; IVd, VIIa,
VIIb: 25-30.
12. Hout and Hauser
(p. 240)
claim th at th e
affinity
terms in our model h ave no clear th eoretical rationale
(cf.
also
Hout,
1989:
148).
It is difficult to know h ow
to
respond
to such
ch arges
wh en th eir auth ors refuse
to take account of th e rationale th at is in fact
provided
(Erikson
and
Goldth orpe,
1987a:
67-9;
1992:
128-30).
Note th at in th e extended model we
specify
as
separate
effects
(AF2, AF3, AF4)
th e th ree different kinds of
(positive) affinity
wh ich we h ave
distinguish ed
but
wh ich in th e
original
model were subsumed under a
single
term
(AF2).
13. T h e same conclusion
h olds,
it sh ould be
noted,
if we
estimate
parameters for our twelve-class core model
in th e same
way
as we did for th e
original seven-class
version: th at
is, by fitting
it not to a set of actual
national
mobility tables,
but rath er to counterfactual
tables
sh owing
wh at national tables would be like if
our
preferred empirical representation of core
fluidity-th e odds ratios under th e model of
common social
fluidity
fitted to th e tables for th e
'central' nations of
England
and France-were
generally
to
prevail (cf. Erikson and
Goldth orpe,
1992:
131, 133-4).
T h ese results are available on
request.
14. We would not wish to claim th at our
way
of
modelling
h ierarch ical effects will
always prove superior
to a
linear-by-linear specification. We would rath er
underwrite th e conclusion reach ed earlier by
Hauser
(1984b: 390)
th at 'variations in th e functional
form of th e vertical term do not dramatically affect
fit'.
15. In
evaluating
th ese
results, we would
again point out
th at th e
FJH-h ypoth esis deserves some ch arity in th at
any non-comparability remaining in our data will count
against
it
and, furth er, in th at our core model may not
in fact be th e best
possible representation of th e
commonality
in
fluidity patterns th at exists. We would
in th is
respect agree
with th e observations made
by
Hout and
Hauser, at least as
regards th e
strength of
th e sector effect
(cf. Erikson and
Goldth orpe, 1992:
174,
n.
22).
16. T h e metric for th e
smallest-space analysis was
provided
by G2s for fits of fifteen
log-linear models to th e six-
class Polish
table, one for each
pair of rows and
columns. Each model
proposes
th at
independence of
class of
origin
and of destination
prevails in cells
given
a
weigh t
of
I(rath er
th an
0)
with in a
weigh t
matrix.
In each
case, cells
receiving
a
weigh t
of 0 are th ose
not
involving
eith er th e
given
row or
column, or
indicating mobility
between
th em,
or on th e
diagonal.
Full details of th e
analysis
are available on
request,
and will be included in a
forth coming paper intended
to extend Erikson and
Goldth orpe (1989)
into a more
compreh ensive critique
of th e use of association models
in
mobility-table analysis.
17. We h ave
difficulty
in
reproducing S0rensen's results
exactly,
but h e would
certainly
also receive a lower bic
for th e CmSF model th an for
any
of th e models h e
suggests.
18. T h is
applies especially
to th e differences in
prestige
and
status effects estimated under model 16 and sh own in
T able
6, including
th e
h igh ly implausible
results in th e
case of
prestige, previously
referred to.
19. T h is
paper
can in fact be
regarded
as
signalling
th e
start of th e
process th rough
wh ich
exponents
of th e
status-attainment
approach
h ave
sough t
to take over
th e
advantages
offered
by
tabular
analysis wh ile,
h owever, reorienting
th is to th eir
particular pre-
occupations.
20. As we h ave
argued,
th e
linear-by-linear specification
may
well be
th ough t
th e more
appropriate
to Hout and
Hauser's own
concerns,
but th is
gives
th em no
grounds
for
claiming
a
general superiority
for it.
Indeed,
th e
results we h ave
reported
in T able
2, above
suggest (and
it is
by
no means
sociologically implausible)
th at
h ierarch y
effects are not in fact
linear;
and th is
finding
is corroborated
by
th e more reliable results th at we
h ave
previously presented (Erikson
and
Goldth orpe,
1992: 136-7 and n.
14).
21. Difficulties
arising
in an
analogous way
h ave been
experienced by analysts using
bic to
legitimate
th e
ch oice of models
th rough
wh ich
th ey
th en seek to claim
trends towards
increasing
social
fluidity
over time
(e.g. Hout, 1988; Ganzeboom et
al., 1989).
It h as
not
proved difficult to
propose
oth er models im-
plying constant
fluidity
wh ich return a
preferable
bic
(see Erikson and
Goldth orpe,
1992:
100-1, 325-6).
299
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
T HE CASMIN PROJECT AND T HE AMERICAN DREAM
ACKNOWLEDGEMENT S
We are indebted to Axel van den
Berg
and
Harry
Ganzeboom for
h elpful
comments on an earlier version of
th is
paper.
REFERENCES
Bielby
W T .
(1981):
'Models of status
attainment',
Research in Social
Stratification
and
Mobility,
1: 3-26.
Duncan 0 D.
(1968):
'Social stratification and
mobility:
Problems in th e measurement of trend' in Sh eldon E
B,
Moore W E.
(eds.),
Indicators
of
Social
Ch ange,
New
York: Russell
Sage Foundation,
675-719.
(1979):
'How destination
depends
on
origin
in th e
occupational mobility table',
American
Sociological
Review,
84: 793-801.
Erikson
R, Goldth orpe
J H.
(1987a): 'Commonality
and
variation in social
fluidity
in industrial nations. Part
I: Amodel for
evaluating
th e "FJH
h ypoth esis"',
European Sociological Review,
3: 54-77.
-
(1987b): 'Commonality
and variation in social
fluidity
in industrial nations. Part II: T h e model of core
social
fluidity applied', European Sociological Review,
3: 145-66.
--
-
(1989):
'Is social
mobility
best measured
by
association
models?',
ISAResearch Committee on
Social
Stratification,
Stanford.
- -
(1992):
T h e Constant Flux: A
Study of
Class
Mobility
in Industrial
Societies,
Oxford: Clarendon
Press.
- -
Konig W, Liittinger P,
Muller W.
(1988):
CASMIN International
Mobility Superfile:
Docu-
mentation, Mannh eim: Institut fur Sozialwissen-
sch aften, University
of Mannh eim.
Feath erman D
L,
Hauser R M
(1978): Opportunity
and
Ch ange,
New York: Academic Press.
-
Jones
FL, Hauser R M.
(1975): 'Assumptions
of
Social
Mobility
Research in th e US: T h e case of
occupational status',
Social Science
Research ,
4:
329-60.
Ganzeboom
H,
de Graaf
P,
T reiman D J.
(1992):
'An
international scale of
occupational status',
Social
Science
Research ,
21: 1-56.
-, Luijkx R,
T reiman DJ.
(1989): 'Intergenerational
class
mobility
in
comparative perspective',
Research in
Social
Stratification
and
Mobility,
8: 3-55.
Goldth orpe
J H.
(with
Catriona
Llewellyn
and Clive
Payne)
(1980),
2nd
edn., 1987):
Social
Mobility
and Class
Structure in Modern
Britain,
Oxford: Clarendon Press.
-
(1985a):
'Soziale Mobilitat und
Klassenbildung:
Zur
Erneuerung
einer T radition
soziologisch er Forsch ung'
in Strasser
H, Goldth orpe
J H.
(eds.),
Die
Analyse
Sozialer
Ungleich h eit, Opladen: Westdeutsch er
Verlag,
174-204.
-
(1985b):
'On economic
development
and social
mobility',
British Journal
of Sociology,
36: 549-73.
Grusky
D
B, Hauser R M
(1984): 'Comparative social
mobility
revisited: Models of
convergence
and
divergence
in 16
countries', American
Sociological
Review,
49: 19-38.
Hauser R M.
(1978):
'Astructural model of th e
mobility
table', Social
Forces, 56: 919-53.
(1984a):
'Vertical class
mobility
in
England, France,
and
Sweden', Acta
Sociologica,
27: 87-109.
(1984b): 'Corrigenda',
Acta
Sociologica,
27: 387-90.
Hout M.
(1984): 'Status, autonomy
and
training
in
occupational mobility',
American Journal
of Sociology,
89: 1379-1409.
-
(1988):
'More
universalism, less structural
mobility:
T h e American
occupational
structure in th e
1980s',
American Journal
of Sociology,
93: 1358-1400.
(1989): Following
in Fath er's
Footsteps:
Social
Mobility
in
Ireland, Cambridge,
Mass.: Harvard
University
Press.
Konig W, Luttinger P, Muller W.
(1987):
'A
comparative
analysis
of th e
development
and structure of educational
systems-meth odological
foundations and th e con-
struction of a
comparative
educational
scale',
Mannh eim: Casmin
Working Paper
no. 12.
Marsh all
G, Newby H,
Rose
D, Vogler
C.
(1988):
Social
Class in Modern
Britain, London: Hutch inson.
Przeworski
A,
T eune H.
(1970):
T h e
Logic of Comparative
Social
Inquiry,
New York:
Wiley.
Raftery
AE.
(1986a): 'Ch oosing
models for cross-
classifications', American
Sociological Review,
51:
145-6.
-
(1986b):
'Anote on
Bayes
factors for
log-linear
contingency
table models with
vague prior information',
Journal
of
th e
Royal
Statistical
Society,
Series
B,
48:
249-50.
Sobel
ME,
Hout
M, Duncan OD
(1985): 'Exch ange,
structure and
symmetry
in
occupational mobility',
American Journal
of Sociology,
91: 359-72.
AUT HORS' ADDRESSES
Robert
Erikson, Swedish Institute for Social
Research ,
Stockh olm
University,
S-106
91, Stockh olm, Sweden.
Joh n H.
Goldth orpe,
Nuffield
College Oxford,
OX1
1NF,
Great Britain.
300
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW 301
APPENDIX
Extended
twelve-category
CASMIN class sch ema
I
High er
service class
II Lower service class
IIIa Routine non-manual
employees, h igh er grade
IIIb Routine non-manual
employees,
lower
grade
IVa Small
proprietors
etc. with
employees
IVb Small
proprietors
etc. with out
employees
IVc Farmers
IVd Small-h olders
V
Lower-grade tech nicians; supervisors
of manual workers
VI Skilled manual workers
(not
in
primary production)
VIIa Semi- and unskilled manual workers
(not
in
primary
production)
VIIb Manual workers in
primary production
Design matrices used in th e extended 12-class core model
HI1
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 2 2 2 2 2 2 2 2 2 2
II 2 1 2 2 2 1 2 2 2 2 2 2
IIIA 2 2 1 2 2 1 2 1 2 2 2
IIIB 2 2 2 1 2 1 2 2 1 2 2
IVA 2 1 2 2 2 2 2 2 2 2 2 2
IVB 2 2 1 2 2 2 2 1 2 2 2
IVC 2 2 2 1 2 2 1 2 2 1 2 2
IVD 2 2 2 2 2 2 2 1 2 2 1 1
V 2 2 1 2 2 1 2 1 2 2 2
VI 2 2 2 1 2 2 1 2 2 1 2 2
VIIA 2 2 2 2 2 2 2 1 2 2 1 1
VIIB 2 2 2 2 2 2 2 1 2 2 1 1
HI2
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1 2 2 1 2 2 2 2 2 2 2
II
2 1 1 2 2 1 2 2 2
IIIA 2 1 1 1 1 2 1 1 2 2
IIIB 2 2 1 2 1 1 1 1 1 1 1
IVA 1 1 1 2 1 1 2 2 1 2 2 2
IVB 2 1 1 1 1 1 2 1 1 2 2
IVC 2 2 1 1 2 1 1 1 1 1 1 1
IVD 2 2 2 1 2 2 1 1 2 1 1 1
V 2 1 1 1 1 2 1 1 2 2
VI 2 2 1 1 2 1 1 1 1 1 1
VIIA 2 2 2 1 2 2 1 1 2 1 1 1
VIIB 2 2 2 1 2 2 1 1 2 1 1 1
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
302 T HE CASMIN PROJECT AND T HE AMERICAN DREAM
HI3
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1 1 2 1 1 2 2 1 2 2 2
II I 1 1 1 1 1 2 1 1 2 2
IIIA 1 1 1 1 1 1 1 1 1 1
IIIB 2 1 1 1 1 1 1 1 1 1
IVA1 1 1 1 1 1 1 2 1 1 2
2
IVB 1 1 1 1 1 1 1 1
V 1 1 1 1 1 1 1 1 1 1 1 1
VI 2 1 1 1 1 1 1 1 1 1 1 1
VIIA 2 2 1 1 2 1 1 1 1 1 1 1
IVC 2 1 1 1 1 1 1 1 1 1 1
IVD 2 2 1 1 2 1 1 1 1 1 1 1
VIIB 2 2 1 1 2 1 1 1 1 1 1 1
HI4
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1
I
1 1 1 1 2 1 1 2 2
II 1 1 1 1 1 1 1 1 1 1 1
IIIA 1 1 1 1 1 1 1 1 1 1 1
IIIB 1 1 1 1 1 1 1 1 1 1
IVA 1 1 1 1 1 1 1 1 1 1
IVB 1 1 1 1 1 1 1 1 1 1 1
V 1 1 1 1 1 1 1 1 1 1
VI 1 1 1 1 1 1 1 1 1 1
VIIA 2 1 1 1
1 1 1 1 1 1
IVC 1 1 1 1 1 1 1 1 1 1
IVD 2 1 1 1 1 1 1 1 1 1 1
VIIB 2 1 1 1 1 1 1 1 1 1 1
IN1
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 2 1 1 1 1 1 1 1
II 1 2 1 1 1 1 1 1 1 1 1
IIIA1 1 2 1 1 1 1 1 1 1 1 1
IIB 1 1 2 1 1 1 1 1 1 1 1
IVA
1 2 1 1 1 1
1 1 1
IVB 1 1 2 1 1 1 1 1 1
IVC
1 1 1 2 1
1 1
1 1
IVD 1 1 1 1 2 1 1 1 1
V11 1 1
V 1 1 1 1 1 1 2 1 1 1
VI 1 1 1 1
1 1 1 1 2 1
VIIA 1 1 1 1 1 1 1 2 1
VIIB
1 1 1 1
2
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW 303
IN2
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 2 1 1 1 1 1 1 1 1 1 1
IIIA
1 1 1 1
1 1 1
1
1 1
1
IIIB
1 1
1 1
1 1
1
1
1
1
IIIBVA
1
2
1
1 1
1 1 1
1
IVA
1 1 1 1 2
1 1
1 1 1
1 1
IVB 1 1 1 1 1 1 1 1 1 1 1 1
V 1 1 1 1 1 1 1 1 1 1 1 1
VI 1 1 1 1 1 1 1 1 1 1 1 1
VIIA 1 1 1 1 1 1 1 1 1 1 1 1
IVC 1 1 1 1 1 2 1 1 1 1 1
IVD 1 1 1 1 1 1 2 1 1 1
1
VIIB 1 1 1 1 1 1 1 1 1 1 1
IN3
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1
I 1 1 1 1 1 1 1 1
II 1 1 1 1 1 1 1 1 1 1 1
IIIA 1 11 1 1 1 1 1 1 1 1
IIIB 1 1 1 1 1 1 1 1 1 1 1
IVA 1 1 1 1 1 1 1 1 1 1 1 1
IVB 1 1 1 1 1 1 1 1 1
VIIA 1 1 1 1 1 1 1 1 1 1
1
IVC 1 1 1 1 1 2 1 1 1 1 1
IVD 1 1 1 1 1 1 1 1 1
VIIB 1 1 1 1
11 1 1 1 1
SE1
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1 1 1 1 1 2 1 11 2
II 1 1 1 1 1 1 2 2 1 1 1 2
IIIA 1 1 1 1 1 1 2 2 1 1 1 2
IIIB 1 1 1 1
2 2 1 1 1 2
IVA 1 1 1 1 1 1 2 2 1 1 1 2
IVB
1
2 2
1 1
1 2
IVC 2 2 2 2 2 2 1 1 2 2 2 1
IVD 2 2 2 2 2 2 1 1 2 2 2 1
V
1 1 1 1 1
1
1 2
VI 1 1 1 1 1
1
1 2
VIIA 1 1 1 1 1 1 2 2 1 11 2
VIIB 2 2 2 2 2 2 1 1 2 2 2 1
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
304 T HE CASMIN PROJECT AND T HE AMERICAN DREAM
SE2
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA
VIIB
0
I 1 1 1 11 2 2 2 2 1 1 1
1
II 1 1 1 1 2 2 2 2 1 1 1
1
IIIA1 1 1 1 2 2 2 2 1 11 1
IIIB1 1 1 1 2 2 2 2 1 1 1 1
IVA 2 2 2 2 1 1 1 1 2 2 2 2
IVB 2 2 2 2 1 1 1 1 2 2 2 2
V 1 1 1 2 2 2 2 1 1 1 1
VI 1 1 1 1 2 2 2 2 1 1 1 1
VIIA 1 1 1 1 2 2 2 2 1 1 1 1
IVC 2 2 2 2 1 1 1 1 2 2 2 2
IVD 2 2 2 2 1 1 1 1 2 2 2 2
VIIB 1 1 1 1 2 2 2 2 1 1 1 1
AF1
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1 1 1 1 1 1 2 1 1 1 2
II1 1 1 1 1 1 1 2 1 1 1 2
IIIA 1 1 1 1 1 1 1 1 1 1 1
IIIB I 1 1 1
I
1 1 1 1 1 1 1
IVA 1 1 1 1 1 1 1 1 1 1
IVB 1 1 1 1 1 1 1 1 1 1
IVC 1 1 1 1 1 1 1 1 1 1
IVD 2 2 1 1 1 1 1 1 1 1 1
V 1 1 1 1 1 1 1 1 1 1 1
VI 1 1 1 1 1 1 1 1 1 1
VIIA 1 1 1 1 1 1 1 1 1 1
VIIB 2 2 1 1 1 1 1
1 1 1 1
AF2
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 2 2 2 1 1 1 1 1 1
II 2 1 2 2 1 1 1 1 1 1
IIIA 2 2 1 2 1 1 11 1
1 1
IIIB 2 2 2 1 1 1 1 1 1 1 1
IVA 1 1 1 1 1 1 1 1 1 1 1 1
IVB 1 1 1 1 1 1 1 1 1 1 1 1
IVC 1 1 1 1 1 1 1 1 1 1 1 1
IVD 1 1 1 1 1 1 1 1 1 1 1 1
V 1 1 1 1 1 1 1 1 1 2 21
VI 1 1 1 1 1 1 1 2 1 2 1
VIIA 1 1 1 1 1 1
2 2 1 1
VIIB 1 1 1 1 1 1 1 1 1
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions
EUROPEAN SOCIOLOGICAL REVIEW
AF3
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1 1 1 2 1 2 1 1 1 1 1
II 1 1 I 1 1 1 1 1 1 1
IIIA 1 1 1 1 I 1 1 1 1 1 1
IIIB 1 1 1 1 1 1 1 1 1 1 1 1
IVA 2 1 1 1 1 1 2 1 1 1 1 1
IVB 1 1 1 1 1 1 1 I 1
IVC 2 1 1 1 2 1 1 1 1 1 1 1
IVD 1 1 1 1 1 1 1 1 1 1 1
V I I I 1 1 1 1 1 1 1
VI 1 1 1 1 1 I 1 1 1 1 1 1
VIIA 1 1 1 1 1 1 1 1 1 I 1
VIIB 1 1 1 1 1 1 1 1 1 1 1 1
AF4
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1 1 1 1 1 1 1 1 1 1
II 1 1 1 1 1 1 1 1 1 1 1
IIIA 1 1 1 1 1 1 1 1 1 1
IIIB 1 1 1 1 1 1 1 1 1 1
IVA 1 1 1 1 1 I 1 1 1 1 1 1
IVB 1 1 1 1 1 1 1 1 1 1 1 1
IVC 1 1 1 I I 1 1 1 1 1 2 1
IVD 1 1 1 1 1 1 1 1 1 1 2 1
V 1 1 1 1 1 1 1 1 1 1 1 1
VI 1 1 1 1 1 1 1 1 1 1
VIIA 1 1 1 1 11 1 1 11 1
VIIB 1 1 1 1 1 1 1 1 2 1
AF5
D I II IIIA IIIB IVA IVB IVC IVD V VI VIIA VIIB
0
I 1
I
1 1 1 1 1 1 1 1 1
II 1 1 1 1 1 1 1 I 1 1 1 1
IIIA 1 1 1 1 1 1 1 1 1
IIIB 1 1 1 1 1 1 1 1 1 1 1 1
IVA 1 1 1 1 1 1 11 1
IVB 1 1 1 1 1 1 1 1 1 1 1 1
IVC 1 1 11 1 11 1 2 1 1 1 1
IVD 1 11 1 1 2 1 1 1 1 1
V 1 1 11
I 1
1
1 1 1 1
VI 1 1 1 1 1 1 1 1 1 1 1
VIIA 1 1 11 1 1 1 1 1 1 1
VIIB 1 1 1 1 1 1 1 1 1 1 1
305
This content downloaded from 200.89.69.93 on Thu, 16 May 2013 12:28:45 PM
All use subject to JSTOR Terms and Conditions

Das könnte Ihnen auch gefallen