Soc. Sci. Med. Vol. 47, No. 8, pp.

1121±1130, 1998
# 1998 Elsevier Science Ltd. All rights reserved
PII: S0277-9536(98)00061-6 Printed in Great Britain
0277-9536/98 $19.00 + 0.00

SOCIAL MOBILITY AND 21 YEAR MORTALITY IN A

COHORT OF SCOTTISH MEN
CAROLE L. HART,1* GEORGE DAVEY SMITH2 and DAVID BLANE3
Department of Public Health, University of Glasgow, 2 Lilybank Gardens, Glasgow, G12 8RZ, U.K.,
1

2
Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Road, Bristol,
BS8 2PR, U.K. and 3Division of Neuroscience, Imperial College of Science, Technology and Medicine,
London, W6 8RP, U.K.

AbstractÐThe objective of this prospective cohort study was to determine the contribution of changes
in social class within and between generations to mortality risk and to socioeconomic di€erentials in
mortality. In 27 workplaces in the west of Scotland. 5567 men aged 35±64 years when screened, pro-
vided information on their father's occupation, their own ®rst occupation and their occupation at
screening. Mortality risk, from broad causes of death by intergenerational and intragenerational social
mobility groups, was measured after 21 years. For all or some of the 3 routes of mobility (childhood to
screening, labour market entry to screening and childhood to labour market entry), increasing values
were seen across the 4 groups (stable non manual, upwardly mobile, downwardly mobile and stable
manual) for diastolic blood pressure, body mass index, current smoking, early school leaving, angina,
bronchitis, severe chest pain, and proportion living in deprived areas. Decreasing values were seen for
serum cholesterol, height, FEV1, exercise, never and ex-smokers, wine drinkers and car users. For
mobility between childhood and screening and between childhood and labour market entry, mortality
risk was highest for the stable manual group and lowest for the stable non manual group for all cause,
cardiovascular disease and other causes of death. The upwardly and downwardly mobile groups had in-
termediate risks. For cancer mortality, the stable manual group had the highest risk with the other
groups having lower and similar risks. For mobility between labour market entry and screening, the
highest risk was for the downwardly mobile group for all cause and cardiovascular mortality. For can-
cer mortality, the risk was higher for men in manual social classes at all occasions. Adjustment for risk
factors attenuated but could not completely explain the di€erentials in mortality risk. Overall, major
di€erences in mortality risk were seen between the stable non manual and the stable manual groups, to
which social mobility does not contribute. With the exception of the small intragenerationally down-
wardly mobile group there was little evidence that social mobility itself was associated with mortality
outcomes di€erent from those expected on the basis of lifetime socioeconomic experience. This is con-
sistent with the suggestion that the main in¯uence of socioeconomic position on mortality risk is an ac-
cumulative one, acting across the lifecourse. # 1998 Elsevier Science Ltd. All rights reserved

Key wordsÐlifecourse, mortality, social class, social mobility, Scotland, health inequalities

INTRODUCTION healthy go down (Stern, 1983; Illsley, 1986; West,

The role of inequalities in mortality across the 1991). Researchers on the other side have suggested
socioeconomic spectrum has been of interest for that the role played by social mobility is small and
many years, with the publication, in Britain, of the socioeconomic gradients are constrained (Goldblatt,
Black report stimulating fresh attempts to under- 1989; Blane et al., 1993; Power et al., 1996; Bartley
and Plewis, 1997). This is because those moving
stand the processes lying behind these health di€er-
upwards, although healthier than the group which
entials (DHSS, 1980). Researchers have considered
they have left, are less healthy than the group
the extent to which social mobility, either up or
which they join, and similarly the less healthy mov-
down the social class hierarchy, contributes to these
ing downwards are healthier than their destination
inequalities (Stern, 1983; Illsley, 1986; Goldblatt,
group. The issue is of importance because if socioe-
1989; West, 1991; Blane et al., 1993; Power et al.,
conomic di€erentials in health are generated by
1996; Bartley and Plewis, 1997) and come to di€er-
health-related social mobility rather than by the
ent conclusions about the importance of the contri-
e€ects of di€erent social environments, then policies
bution of health-related social mobility to mortality
aimed at reducing health inequalities through equal-
di€erentials. On the one side, it has been argued
ising life chances will not succeed.
that health-related social mobility can explain socio-
We have investigated the mortality experience of
economic di€erences in health and mortality since
a large cohort of working men whose social class
the healthier rise up the social spectrum and the less
was known at three stages of their life. In a pre-
vious study of this cohort, we analyzed the cumu-
*Author for correspondence. lative e€ect on mortality of social class at these
1121
1122 C. L. Hart et al.
three stages (Davey Smith et al., 1997). We allo-
cated men to one of four groups depending on the
number of occasions their social class was non man-
ual or manual (all three non manual, two non man-
ual and one manual, one non manual and two
manual and all three manual). A clear cumulative
e€ect of socioeconomic circumstances over a life-
time was found, with mortality risk positively re-
lated to the number of occasions spent in manual
social classes. In another study of this cohort, we
analyzed mortality by social class at the three stages
separately (Hart et al., 1998). We found that mor-
tality risk was similar at each lifestage with men in
social classes I and II having the lowest risk. A
widening of inequalities in adulthood suggested the
importance of the accumulation of poor socioeco-
nomic circumstances throughout life.
In the current study, we have assessed whether
the accumulation of socioeconomic risk could be
due to health-related social mobility. We looked at
three possible routes of social mobility Ð interge-
nerational mobility between childhood social class
and later adult social class, intragenerational mobi-
lity between early adult and later adult social class,
and intergenerational mobility between childhood
social class and early adult social class. In addition,
unlike other studies of social mobility, this cohort
has information on individual behavioural and
physiological measures, enabling us to see how they
contribute to the mortality experience of socially
mobile and socially stable groups.
METHODS
This analysis was based on part of a cohort of
employed people from 27 workplaces in Glasgow,
Clydebank and Grangemouth, Scotland, who were
screened between 1970 and 1973. The full sample
consisted of 6022 men and 1006 women.
Participants completed a questionnaire and
attended a physical examination. Women have been
excluded from this study due to the small number
of deaths which have occurred. Full details have
been described elsewhere (Blane et al., 1996; Davey
Smith et al., 1997).
The physical examination included measurement
of blood pressure (seated, with diastolic pressure
recorded at the disappearance of the ®fth Korotko€
sound), height, weight, lung function (with the
Garthur Vitalograph) and a 6 lead electrocardio-
gram. In addition a blood sample was taken for
measurement of serum cholesterol concentration.
The questionnaire collected information about
smoking, alcohol consumption, exercise outside
work, age leaving full time education, home
address, whether a regular car driver, angina from
the Rose questionnaire (Rose et al., 1982) and res-
piratory symptoms from the MRC questionnaire
(Medical Research Council, 1965).
Breathlessness was de®ned by a positive response
to the question ``Do you get short of breath walk-
ing with people of your own age on level ground?''.
Persistent phlegm was de®ned as usually bringing
up phlegm from the chest ®rst thing in the morning
on most days for as much as three months in winter
each year. Infective phlegm was de®ned as addition-
ally having had a period of increased cough and
phlegm lasting for three weeks or more in the past
three years. Bronchitis was de®ned as having per-
sistent and infective phlegm and being breathless
(Medical Research Council, 1965).
The forced expiratory volume in one second
(FEV1) score was calculated by obtaining the
expected FEV1 from a linear regression equation of
age and height, derived from a healthy subset of the
population who had never smoked and answered
no to questions on phlegm, breathlessness, wheezy
or whistling chest and if the weather a€ected
breathing (Blane et al., 1996; Davey Smith et al.,
1997). The FEV1 score was de®ned as the actual
FEV1 as a percentage of the expected FEV1.
A six lead electrocardiogram (ECG) was made
with the subject seated. The ECG was coded
according to the Minnesota system with any of
codes 1.1±1.3, 4.1±4.4, 5.1±5.3 and 7.1 being con-
sidered as evidence of ischaemia, encompassing
diagnoses of de®nite myocardial infarction, myocar-
dial ischaemia and left bundle branch block
(Prineas et al., 1982; Hawthorne et al., 1995).
Angina was considered present if the de®nite cri-
teria of the Rose Angina Questionnaire were met
(Hart et al., 1997). Severe chest pain was de®ned as
admitting to ever experiencing a severe pain across
the front of the chest lasting half an hour or more
(Hart et al., 1997).
The home address at the time of screening was
retrospectively postcoded, enabling deprivation cat-
egory as de®ned by Carstairs and Morris to be
ascertained (Carstairs and Morris, 1991).
Deprivation category varies from 1 (least deprived)
to 7 (most deprived) and is based on four census
variables Ð male unemployment, overcrowding, car
ownership and the proportion in social classes IV
and V.
Participants were ¯agged at the National Health
Service Central Register in Edinburgh and deaths
occurring in the 21 year follow-up period were noti-
®ed together with their cause, coded according to
ICD9. Causes of death were grouped in four ways:
all causes, cardiovascular disease (ICD9 codes 390±
459), cancer (ICD9 codes 140±208) and other
causes.
The questionnaire asked for the main occupation
of the participant's father, the participant's own
®rst regular occupation (excluding temporary work)
and the occupation at the time of screening. Social
class was coded according to the Registrar
General's Classi®cation (General Register Oce,
1966) from these three occupations. For this analy-
 Social mobility and 21 year mortality 1123

sis, social class was de®ned as either non manual
(classes I, II and III non manual) or manual (classes
III manual, IV and V). This was for clarity of pres-
entation and because of small numbers in many of
the cells in the full social mobility tables (Blane et
al., 1996). The analysis was based on 5766 men
aged between 35 and 64 years at screening. After
199 men with missing data on any of the three oc-
cupations were excluded, the analysis was based on
5567 men.
The availability of social class at three di€erent
stages of the lifecourse has enabled social mobility
to be investigated. In this analysis, we looked ®rstly
at mobility from father's social class to social class
at screening (intergenerational mobility), secondly
from social class at entry to the labour market to
social class at screening (intragenerational mobility)
and thirdly from father's social class to social class
at labour market entry (also intergenerational mobi-
lity).
Age adjusted means of continuous variables and
tests for di€erences between them were calculated
using PROC GLM of the SAS system (SAS
Institute Inc., 1989). Age adjusted proportions of
discrete variables were calculated using the direct
method with the study population as the standard.
The signi®cance of di€erences between them were
tested using PROC LOGIST.
Relative risks of mortality were calculated using
Cox's Proportional Hazards models. The baseline
category was taken as the socially stable non man-
ual group. Adjustments were ®rst made for age
only for all cause of death groups and then for age,
smoking, diastolic blood pressure, cholesterol,
FEV1 score, angina, severe chest pain, breathless-
ness, car use and ECG ischaemia for all cause and
cardiovascular disease mortality, and age, smoking,
body mass index, FEV1 score, breathlessness and
car use for cancer and other mortality. These were
the variables which were statistically signi®cant
when included in the models.
Con®dence intervals for the relative risks were
estimated by treating them as ¯oating absolute risks
(Easton et al., 1991; Prospective Studies
Collaboration, 1995). This method attributes an
estimate of variability produced by chance among
participants in the baseline category. It sub-
sequently reduces the variances associated with the
other relative risks and reduces the covariances
between them. The choice of which category to use
as the baseline is thus not as crucial. The values of
the relative risks, now called ¯oating absolute risks,
remain unchanged.
RESULTS
Father's social class and social class at screening
Taking mobility between father's social class and
social class at screening, the largest group of men
were the socially stable manual workers and the
smallest group were the downwardly mobile, who
were manual workers at screening but whose fathers
had been non manual workers (Table 1).
Statistically signi®cant di€erences between non
manual and manual social classes at screening were
seen for all the characteristics except body mass
index and ECG ischaemia. Di€erences between the
four groups were seen for most of the character-
istics, with the most favourable values seen for the

Table 1. Age adjusted characteristics according to social class at screening and father's social class
Social class at screening
non manual manual
father's social class father's social class
p value for age adjusted non manual manual non manual manual
di€erence between n = 1121 n = 1669 n = 211 n = 2566
nonmanual and manual (20%) (30%) (4%) (46%)
social class at screening (stable non manual) (moved up) (moved down) (stable manual)

Mean age 0.0001 47.6 47.9 49.6 48.6

Mean diastolic blood pressure (mmHg) 0.0001 82.5 83.4 84.6 84.7
Mean cholesterol (mmol/l) 0.0001 6.2 6.0 5.6 5.6
Mean height (cm) 0.0001 175.8 173.9 172.9 170.8
Mean body mass index (kg/m2) 0.074 24.8 25.3 25.1 25.2
Mean FEV1 score (%) 0.0001 99 97 92 90
Mean exercise per week (hours) 0.0001 6.5 6.3 6.2 5.7
% never smoked 0.0001 19.6 21.9 14.0 14.4
% current smokers 0.0001 45.8 48.7 59.2 63.4
% ex-smokers 0.0001 30.4 26.0 24.1 20.4
% wine drinkers 0.0001 12.0 6.1 5.1 2.7
% regular car drivers 0.0001 67.8 58.5 52.6 40.8
% left education at 14 or under 0.0001 10.3 37.9 54.5 77.5
% ECG ischaemia 0.89 5.6 6.2 5.3 5.8
% angina 0.005 4.3 6.2 6.6 7.2
% breathless 0.0001 3.2 5.2 8.7 7.4
% MRC bronchitis 0.0001 0.7 1.5 3.6 3.3
% severe chest pain 0.022 5.3 6.9 7.4 7.8
% deprivation category 5±7 0.0001 20.5 38.7 58.5 66.4
1124 C. L. Hart et al.

Table 2. Age adjusted characteristics according to social class at screening and at labour market entry
Social class at screening
non manual manual
social class at labour social class at labour
p value for age adjusted market entry market entry
di€erence between non manual manual non manual manual
nonmanual and manual n = 1879 n = 911 n = 281 n = 2469
social class at labour (34%) (16%) (5%) (45%)
market entry (stable non manual) (moved up) (moved down) (stable manual)

Mean age 0.0001 47.4 48.4 49.1 48.6

Mean diastolic blood pressure (mmHg) 0.0001 82.7 83.7 84.6 84.7
Mean cholesterol (mmol/l) 0.0001 6.2 5.9 5.7 5.6
Mean height (cm) 0.0001 175.3 173.4 170.9 171.0
Mean body mass index (kg/m2) 0.0001 24.9 25.4 25.0 25.3
Mean FEV1 score (%) 0.0001 99 96 92 90
Mean exercise per week (hours) 0.029 6.3 6.6 5.8 5.7
% never smoked 0.0001 21.5 19.8 14.7 14.4
% current smokers 0.0001 45.7 51.3 62.3 63.2
% ex-smokers 0.0001 29.2 24.9 19.6 20.8
% wine drinkers 0.0001 10.5 4.2 2.1 3.0
% regular car drivers 0.0001 63.5 60.0 41.2 41.5
% left education at 14 or under 0.0001 13.2 54.3 67.6 76.7
% ECG ischaemia 0.49 5.8 6.2 4.7 5.9
% angina 0.002 5.1 6.2 4.9 7.4
% breathless 0.0001 3.2 6.8 6.4 7.6
% MRC bronchitis 0.0001 0.5 2.3 2.6 3.4
% severe chest pain 0.023 5.6 7.5 9.5 7.5
% deprivation category 5±7 0.0001 25.0 44.2 67.9 65.6

socially stable non manual group, followed by the
upwardly mobile group, the downwardly mobile
group and the least favourable values seen in the
socially stable manual group. Increasing means or
proportions were seen across the four groups for di-
astolic blood pressure, current smoking, angina,
severe chest pain, early school leaving and pro-
portion living in the most deprived areas.
Decreasing means or proportions were seen for cho-
lesterol, height, FEV1, exercise, ex smokers, wine
drinkers and car drivers.
Social class at labour market entry and at screening
Statistically signi®cant di€erences between non
manual and manual social classes at labour market
entry were observed for all characteristics except for
ECG ischaemia (Table 2). Characteristics according
to social class at screening and at labour market
entry across the four groups were similar to those
seen for mobility from father's to screening social
class. Increasing values were seen for diastolic
blood pressure, current smoking, bronchitis and

Table 3. Age adjusted characteristics according to social class at labour market entry and father's social class
Social class at labour market entry
non manual manual
father's social class father's social class
p value for age adjusted non manual manual non manual manual
di€erence between n = 990 n = 1170 n = 342 n = 3065
nonmanual and manual (18%) (21%) (6%) (55%)
father's social class (stable non manual) (moved up) (moved down) (stable manual)

Mean age 0.068 47.6 47.7 48.8 48.5

Mean diastolic blood pressure (mmHg) 0.0001 82.5 83.4 83.7 84.5
Mean cholesterol (mmol/l) 0.0001 6.2 6.1 5.8 5.7
Mean height (cm) 0.0001 175.9 173.7 173.6 171.4
Mean body mass index (kg/m2) 0.0001 24.7 25.1 25.1 25.3
Mean FEV1 score (%) 0.0001 99 97 95 92
Mean exercise per week (hours) 0.002 6.4 6.1 6.7 5.9
% never smoked 0.30 19.8 21.3 15.4 15.9
% current smokers 0.0001 46.1 49.5 54.0 60.7
% ex-smokers 0.0001 30.3 25.9 26.4 21.4
% wine drinkers 0.0001 12.7 6.7 5.4 3.1
% regular car drivers 0.0001 68.3 53.7 55.9 45.4
% left education at 14 or under 0.0001 7.5 31.6 45.3 73.5
% ECG ischaemia 0.55 5.9 5.4 4.4 6.1
% angina 0.003 4.5 5.7 5.3 7.3
% breathless 0.0006 2.8 4.3 7.8 7.4
% MRC bronchitis 0.001 0.6 1.1 2.7 3.2
% severe chest pain 0.019 5.3 6.6 6.7 7.6
% deprivation category 5±7 0.0001 19.6 40.0 45.7 61.4
 Social mobility and 21 year mortality 1125

early school leaving and decreasing values were
seen for cholesterol, FEV1 and never smokers.
Father's social class and social class at labour market
entry
Analysing the data according to father's social
class and social class at labour market entry, similar
increasing or decreasing values were again seen in
the characteristics (Table 3). Increasing values were
seen for diastolic blood pressure, body mass index,
current smoking, early school leaving, bronchitis,
severe chest pain and proportion living in the most
deprived areas. Decreasing values were seen for
cholesterol, height, FEV1 and wine drinking. All
the characteristics except for age, never smokers
and ECG ischaemia showed signi®cant di€erences
between non manual and manual father's social
class. Over half of the cohort had fathers with man-
ual occupations and had manual occupations on
®rst entering the workforce.
Father's social class and social class at screening
(mortality)
In the 21 year follow up period, 1580 men died.
The greatest risk of mortality by father's and
screening social class for all causes of death studied
was seen in the socially stable manual group (all
cause 1.66, cardiovascular disease 1.83, cancer 1.38
and other 1.79) (Table 4). This excess was attenu-
ated but remained when adjustment was made for
risk factors. Mortality risks for the upwardly and
downwardly mobile groups were between those for
the stable non manual and stable manual groups
for all cause, cardiovascular disease and other
causes of death. For cancer mortality their risks
were similar to the stable non manual group. There
was no signi®cant di€erence in mortality risk
between the upwardly and downwardly mobile
groups from any cause. The stable manual group
had signi®cantly higher mortality from both the
upwardly and downwardly mobile groups for all
cause and cancer mortality. For cardiovascular dis-
ease and other mortality, the stable manual group
had signi®cantly higher risks than the upwardly,
but not downwardly mobile groups. After risk fac-
tor adjustment, this elevated risk from other mor-
tality for the stable manual group was attenuated
and became non signi®cant.
To gain an idea of the size of the contribution of
social mobility to di€erentials in mortality, the risk
of all cause mortality was calculated for manual
compared to non manual social class groups at
screening. This was the mortality risk with the
mobile groups included. The age adjusted rate was
1.40 (95% con®dence interval 1.27 to 1.55). This
was in comparison with the rate without the mobile
groups from Table 4 (1.66).
Social class at labour market entry and at screening
(mortality)
The highest mortality risks were seen for the
downwardly mobile group from labour market
entry to screening social class for all cause (1.70),
cardiovascular disease (1.96) and other mortality
(1.60) (Table 5). These risks were attenuated but
remained substantial and statistically signi®cantly
di€erent from the stable non-manual group for all
cause and cardiovascular disease mortality when

Table 4. 21 year mortality according to social class at screening and father's social class
Social class at screening
non manual manual
father's social class father's social class
non manual* manual non manual manual
(stable non manual) (moved up) (moved down) (stable manual)

All cause
Number of deaths 234 424 59 863
bd
Age adjusted ¯oating absolute risk 1 (0.88±1.14) 1.26ad (1.14±1.38) 1.21d (0.94±1.57) 1.66abc (1.55±1.77)
Risk factor$ adjusted ¯oating absolute risk bd
1 (0.88±1.14) ad
1.18 (1.08±1.30) d
1.03 (0.80±1.34) abc
1.43 (1.33±1.53)
Cardiovascular disease
Number of deaths 107 234 33 440
Age adjusted ¯oating absolute risk 1bcd (0.83±1.21) 1.51ad (1.33±1.72) 1.47a (1.05±2.07) 1.83ab (1.67±2.01)
Risk factor$ adjusted ¯oating absolute risk bd
1 (0.82±1.22) ad
1.44 (1.26±1.63) 1.33 (0.94±1.88) ab
1.71 (1.55±1.88)
Cancer
Number of deaths 89 125 15 272
Age adjusted ¯oating absolute risk 1d (0.81±1.23) 0.98d (0.82±1.16) 0.82d (0.49±1.36) 1.38abc (1.23±1.56)
Risk factor% adjusted ¯oating absolute risk 1 (0.81±1.24) d
0.95 (0.80±1.13) d
0.74 (0.44±1.23) bc
1.23 (1.09±1.40)
Other
Number of deaths 38 65 11 151
Age adjusted ¯oating absolute risk 1d (0.73±1.38) 1.19d (0.93±1.52) 1.41 (0.78±2.55) 1.79ab (1.53±2.10)
Risk factor% adjusted ¯oating absolute risk 1 (0.72±1.39) 1.10 (0.86±1.40) 1.08 (0.60±1.95) 1.33 (1.12±1.56)
a
p < 0.05 compared to stable non-manual group.
b
p < 0.05 compared to upwardly mobile group.
c
p < 0.05 compared to downwardly mobile group.
d
p < 0.05 compared to stable manual group.
*baseline category.
$
adjusted for age, diastolic blood pressure, cholesterol, smoking, angina, ecg ischaemia, severe chest pain, fev1 score, breathlessness and
car user.
%
adjusted for age, smoking, body mass index, fev1 score, breathlessness and car user.
1126 C. L. Hart et al.

Table 5. 21 year mortality according to social class at screening and at labour market entry
Social class at screening
non manual manual
social class at labour market entry social class at labour market entry
non manual* manual non manual manual
(stable non manual) (moved up) (moved down) (stable manual)

All cause
Number of deaths 411 247 107 815
Age adjusted ¯oating absolute risk 1bcd (0.91±1.10) 1.21acd (1.07±1.37) 1.70ab (1.41±2.06) 1.48ab (1.38±1.59)
Risk factor$ adjusted ¯oating absolute risk cd
1 (0.90±1.11) c
1.13 (1.00±1.28) ab
1.50 (1.24±1.82) a
1.29 (1.20±1.39)
Cardiovascular disease
Number of deaths 198 143 60 413
Age adjusted ¯oating absolute risk 1bcd (0.87±1.15) 1.45ac (1.23±1.71) 1.96ab (1.52±2.53) 1.55a (1.41±1.71)
Risk factor$ adjusted ¯oating absolute risk bcd
1 (0.86±1.16) ac
1.36 (1.15±1.60) ab
1.87 (1.45±2.41) a
1.45 (1.31±1.60)
Cancer
Number of deaths 139 75 29 258
Age adjusted ¯oating absolute risk 1d (0.85±1.18) 1.09 (0.87±1.37) 1.38 (0.96±1.98) 1.39a (1.24±1.58)
Risk factor% adjusted ¯oating absolute risk d
1 (0.84±1.19) 1.05 (0.83±1.31) 1.23 (0.85±1.77) a
1.25 (1.11±1.42)
Other
Number of deaths 74 29 18 144
Age adjusted ¯oating absolute risk 1d (0.80±1.26) 0.79cd (0.55±1.14) 1.60b (1.01±2.55) 1.46ab (1.24±1.72)
Risk factor% adjusted ¯oating absolute risk 1 (0.79±1.27) d
0.72 (0.50±1.04) 1.19 (0.75±1.89) b
1.09 (0.93±1.29)
a
p < 0.05 compared to stable non-manual group.
b
p < 0.05 compared to upwardly mobile group.
c
p < 0.05 compared to downwardly mobile group.
d
p < 0.05 compared to stable manual group.
*baseline category.
$
adjusted for age, diastolic blood pressure, cholesterol, smoking, angina, ecg ischaemia, severe chest pain, fev1 score, breathlessness and
car user.
%
adjusted for age, smoking, body mass index, fev1 score, breathlessness and car user.

adjustment was made for risk factors. The down-
wardly mobile group also had signi®cantly higher
mortality than the upwardly mobile group for all
cause, cardiovascular disease and other mortality.
For other mortality, risk factor adjustment attenu-
ated the risk which became non signi®cant. The
mortality risk of the stable manual group was not
signi®cantly di€erent from that of the downwardly
mobile group for any cause of death.
The mortality risk with the mobile groups was
1.40 and without the mobile groups was 1.48
(Table 5) for all cause mortality.
More detailed analysis of the intragenerationally
downwardly mobile group (n = 281) with social
class expanded to four groupings (I and II, III non
manual, III manual, IV and V) showed that most
had been in social class III non manual at labour
market entry, with 155 moving to social class III
manual jobs and 98 moving to jobs classi®ed as
social classes IV or V. The highest all cause mor-
tality risk was for those whose jobs at screening
were social classes IV or V, followed by those
classi®ed as social class III manual (Relative
risks = 2.32 (95% con®dence interval 1.64±3.27),
and 2.22 (1.63±3.03) respectively, compared to the
baseline category of stable social classes I or II.
Full table available on request.) Amongst the 28
men moving downwards from social classes I and
II, there were only 7 deaths and the relative risks
were not signi®cantly di€erent from the stable non-
manual group.
Father's social class and social class at labour market
entry (mortality)
The stable manual group between father's social
class and social class at labour market entry had
the highest mortality risks for all cause (1.57), car-
diovascular disease (1.82), cancer (1.32) and other
(1.48) mortality (Table 6). The next highest risk was
seen in the upwardly mobile group. After adjust-
ment for risk factors similar results were seen,
except for other mortality where the highest risk
was seen in the upwardly mobile group. None of
the groups had statistically signi®cant di€erences in
other mortality after such adjustments. For cancer
mortality, the stable manual group had a higher
risk than the other three groups which had similar
risks. There was no signi®cant di€erence between
mortality from any cause between the upwardly and
downwardly mobile groups. The stable manual
group had signi®cantly higher mortality than both
the upwardly and downwardly mobile groups for
all cause mortality. After risk factor adjustment, the
di€erence between the stable manual and the
upwardly mobile group become non-signi®cant.
The relative rate of all cause mortality was calcu-
lated for manual compared to non manual social
class at labour market entry (1.29 (95% ci 1.16 to
1.43)), the risk including the mobile groups. This
was in comparison with the rate without the mobile
groups from Table 6 (1.57).
Tests for interaction were performed for all
causes of death to see if the downwardly mobile in
each analysis had a higher risk than would be
expected by the combined in¯uence of the two
 Social mobility and 21 year mortality 1127

Table 6. 21 year mortality according to social class at labour market entry and father's social class
Social class at labour market entry
non manual manual
father's social class father's social class
non manual* manual non manual manual
(stable non manual) (moved up) (moved down) (stable manual)

All cause
Number of deaths 205 313 88 974
Age adjusted ¯oating absolute risk 1bd (0.87±1.15) 1.36ad (1.22±1.52) 1.18d (0.96±1.46) 1.57abc (1.48±1.68)
bd a d ac
Risk factor$ adjusted ¯oating absolute risk 1 (0.87±1.15) 1.28 (1.15±1.43) 1.06 (0.86±1.31) 1.36 (1.27±1.45)
Cardiovascular disease
Number of deaths 92 166 48 508
Age adjusted ¯oating absolute risk 1bcd (0.82±1.23) 1.60a (1.38±1.87) 1.43a (1.08±1.90) 1.82a (1.67±1.99)
Risk factor$ adjusted ¯oating absolute risk bd
1 (0.81±1.23) a
1.56 (1.34±1.82) 1.35 (1.01±1.79) a
1.67 (1.53±1.83)
Cancer
Number of deaths 77 91 27 306
Age adjusted ¯oating absolute risk 1d (0.80±1.25) 1.05 (0.86±1.30) 0.97 (0.66±1.41) 1.32a (1.18±1.48)
Risk factor% adjusted ¯oating absolute risk 1 (0.80±1.26) 1.01 (0.83±1.25) 0.91 (0.62±1.32) 1.20 (1.06±1.34)
Other
Number of deaths 36 56 13 160
Age adjusted ¯oating absolute risk 1d (0.72±1.39) 1.38 (1.07±1.80) 1.00 (0.58±1.72) 1.48a (1.27±1.72)
Risk factor% adjusted ¯oating absolute risk 1 (0.72±1.40) 1.24 (0.95±1.61) 0.83 (0.48±1.43) 1.13 (0.96±1.32)
a
p < 0.05 compared to stable non-manual group.
b
p < 0.05 compared to upwardly mobile group.
c
p < 0.05 compared to downwardly mobile group.
d
p < 0.05 compared to stable manual group.
*baseline category.
$
adjusted for age, diastolic blood pressure, cholesterol, smoking, angina, ecg ischaemia, severe chest pain, fev1 score, breathlessness and
car user.
%
adjusted for age, smoking, body mass index, fev1 score, breathlessness and car user.

social class locations. Interactions were only signi®-
cant for all cause (p = 0.012 before, p = 0.03 after
risk factor adjustment) and cardiovascular disease
mortality (p = 0.0006 before, p = 0.002 after risk
factor adjustment) by social class at labour market
entry and at screening.
DISCUSSION
Interpretation of mortality results
We have shown that overall, mortality experience
was worse for the stable manual group, best for the
stable non manual group and in between these two
for the upwardly and downwardly mobile groups.
This is compatible with results from the cumulative
social class approach, analyzed in a previous study
of mortality in this cohort where men were allo-
cated to one of four groups depending on the num-
ber of occasions their social class was non manual
or manual (Davey Smith et al., 1997). It was found
that the lowest mortality risk was for the group
with non manual social classes on the three oc-
casions, the next highest risk was for the group in
non manual social classes on two occasions, the
next for the group with one non manual social class
location and the highest risk was for the group in
manual social classes on all three occasions. In the
present analyses, the upwardly mobile group had a
lower mortality risk than the manual group from
where they originated and a higher mortality risk
than the non manual group which they joined.
They had a greater mortality risk than the non
manual group because of their more disadvantaged
past. Likewise, the downwardly mobile group had a
higher mortality risk than the non manual group
from where they originated and a lower mortality
risk than the manual group which they joined. This
lower risk appears to be due to their more advan-
taged earlier life experiences.
The exception to these ®ndings was for intragen-
erational mobility from social class at labour mar-
ket entry to social class at screening. Here we found
that the downwardly mobile group had a markedly
elevated risk for all cause and cardiovascular dis-
ease mortality, even after adjustment for risk fac-
tors. This e€ect remained when men with pre-
existing illness at screening were excluded from the
analysis (results not shown). Further adjustment for
education and height did not a€ect the results and
adjustment for father's social class had a small
e€ect. Most of this group had jobs at labour mar-
ket entry classi®ed as III non manual and by the
time of screening were in jobs classi®ed as III man-
ual. Inspection of a sample of the original question-
naires revealed that ®rst jobs tended to be grocer's
boys or oce boys and their jobs at screening were
skilled manual jobs, such as electricians, which
would have required an apprenticeship. It is poss-
ible that men in this group were discriminated
against in some way when applying for apprentice-
ships (West, 1991) and took the oce boy or gro-
cer's boy jobs whilst waiting to be accepted on an
apprenticeship scheme. This discrimination could be
related to their health status in childhood. Two
thirds of the group left education at 14 or under, so
may have been too young to join an apprenticeship
scheme. It must be questioned as to the validity of
coding grocer's or oce boys as III non manual
1128 C. L. Hart et al.
(although this was the correct coding in the classi®-
cation of occupations used) as these jobs were likely
to have been somewhat manual in their nature (eg
delivering groceries, taking messages) and with rela-
tively poor pay, prospects and work conditions.
Men in this group were older, shorter, had less ex
smokers, wine drinkers and car drivers and more
men living in deprived areas than the other three
groups (Table 2). Curiously this group had more
men with severe chest pain, but less with angina or
ECG ischaemia. Despite this, the group of men had
signi®cantly higher mortality rates which could not
be completely explained. This may suggest some
form of health-related selection (West, 1991). The
small proportion of the population in this group,
and the marked mortality di€erentials between the
stable manual and non-manual social groups, indi-
cates that the contribution of this health-related
selection to overall mortality di€erentials is modest.
Our calculations have shown that all cause mor-
tality di€erentials were greater without mobility
than with mobility (for example, 1.66 vs 1.40 for
father's to screening social class), showing that
social mobility constrains mortality di€erentials.
Other studies
In a study of men in Finland, Lynch et al.
found that the relatively small downwardly mobile
group whose childhood socioeconomic status
(based on several socioeconomic measures) had
been high but whose adult socioeconomic status
(based on income) was low had the highest all
cause mortality risk, but the risk was not signi®-
cantly di€erent from other groups with low adult
income (Lynch et al., 1994). A study of middle
aged men in Sweden looked at mobility from
father's social group at age 50 to son's social
group at the same age, based on occupation. The
downwardly mobile had a poorer subjective state
of health and reported more symptoms at the age
of 60, and were slightly more at risk from myocar-
dial infarction, although after 12 years showed no
di€erence in mortality risk (FaresjoÈ et al., 1994).
More recently, results were reported after 22 years
of follow up on this cohort (FaresjoÈ et al., 1997).
With a di€erent de®nition of downward mobility
(men with high education but low social position),
an increased risk for all cause and coronary heart
disease mortality was found.
Many researchers have concluded that social
mobility has little e€ect on social class di€erentials
in mortality or morbidity (Goldblatt, 1989; Blane et
al., 1993; Power et al., 1996; Rahkonen et al., 1997;
Bartley and Plewis, 1997). Our study, and that of
Lynch (Lynch et al., 1994) showed that the main
di€erentials were seen between the stable non man-
ual and manual groups for whom selection (at least
between non manual and manual classes) had not
occurred.
It is possible that socioeconomic in¯uences on
particular causes of death may have di€erent criti-
cal periods (Davey Smith et al., 1997). The present
study showed di€erent results for cancer mortality
risk by form of social mobility. In all three sets of
analyses, the highest risk was for the stable manual
groups. In the intragenerational mobility analysis,
the downwardly mobile group had a similar high
risk to that of the stable manual group, which may
suggest the particular importance of social class in
adulthood in determining overall cancer mortality
risk. This e€ect persisted after adjustment for risk
factors, including smoking. However, the down-
wardly mobile group from father's to screening
social class did not support this idea, since they did
not have an elevated cancer mortality risk. This
was, however, based on a small number of deaths
from cancer.
A study of men in the British Regional Heart
Study found that non fatal myocardial infarction or
self reported physician diagnosed ischaemic heart
disease was higher for men with manual fathers
than for men with non manual fathers, even after
adjustment for adult social class and other risk fac-
tors (Wannamethee et al., 1996). Their results when
considering father's and adult social class together
were similar to our results for cardiovascular dis-
ease mortality. Unlike overall cancer mortality, car-
diovascular disease mortality appears to be sensitive
to early life in¯uences (Davey Smith et al., 1997;
Wannamethee et al., 1996).
Our cohort was recruited from workplaces,
although at a time of relatively low unemployment.
The mortality di€erentials were similar to those
seen in Scottish men in 1981 (Davey Smith et al.,
1997). Social mobility patterns were broadly con-
sistent with studies of the same time in Scotland
(Payne, 1987) and in England and Wales
(Goldthorpe, 1980). Men under 35 years of age
when screened were excluded due to the frequent
job changes seen in younger workers (Marshall et
al., 1988). The few men over 64 were also excluded
due to the small numbers. Women were not
included in this analysis because only 176 had died
in 21 years (Blane et al., 1996). Our results on mor-
tality in 21 years could therefore probably be gener-
alised to the wider British population.
Characteristics at screening
The characteristics presented in Tables 1±3
showed broadly similar patterns, re¯ecting the ac-
cumulation of characteristics depending on socioe-
conomic experience throughout the lifecourse. Men
with non manual social classes on two occasions
had, for example the highest FEV1 score, whereas
men with manual social classes on two occasions
had the lowest FEV1 score. The upwardly mobile
had the second highest FEV1 score and the down-
wardly mobile the third highest. This was consist-
ent for the three mobility groupings. Similar
 Social mobility and 21 year mortality
patterns were seen for the other characteristics,
although for some variables the trends were not as
consistent. Apart from cholesterol where the higher
mortality risk groups had lower cholesterol values,
the other characteristics were less favourable in the
groups with the highest mortality risk. In a
Swedish study, height was linked to upward mobi-
lity from childhood to adult socioeconomic status
(NystroÈm Peck, 1992). This was seen in our study,
where the tallest men were in the stable non man-
ual group, the next tallest were in the upwardly
mobile group, followed by the downwardly mobile
group, with the shortest men being in the stable
manual group (Table 1).
Adjusting the mortality analyses for risk factors
attenuated the elevated mortality risks results for all
causes of death, particularly for other causes, but
did not alter the main ®ndings. Thus these
measured characteristics do not account for all the
di€erences in mortality risk between social class
groupings.
CONCLUSIONS
The major di€erentials in mortality were seen
between the stable non manual and the stable man-
ual groups, with mobility making only a minor con-
tribution and constraining socioeconomic
di€erentials in mortality. Apart from the intragen-
erational mobility ®ndings, the mortality risks seen
in this study were consistent with a cumulative life-
course approach. Future studies should consider the
importance of obtaining detailed information on
socioeconomic position throughout the lifecourse.
The factors relating to the high mortality of the
small intergenerationally downwardly mobile group
require investigation in studies which obtain data
on reasons for occupational change and how factors
associated with this can contribute to mortality
risk.
AcknowledgementsÐFunding was provided by a grant
from the NHS Management Executive, Cardiovascular
Disease and Stroke Research and Development Initiative.
Victor M. Hawthorne was responsible for the original
screening study and we thank him for his continuing inter-
est. Charles Gillis, David Hole and Pauline MacKinnon
are also thanked for their support.
REFERENCES
Bartley, M. and Plewis, I. (1997) Does health-selective
mobility account for socioeconomic di€erences in
health? Evidence from England and Wales, 1971 to
1991. Journal of Health and Social Behaviour, 38:376±
86.
Blane, D., Davey Smith, G. and Bartley, M. (1993) Social
selection: what does it contribute to social class di€er-
ences in health? Sociology of Health and Illness 15, 1±
15.
Blane, D., Hart, C. L., Davey Smith, G., Gillis, C. R.,
Hole, D. J. and Hawthorne, V. M. (1996) The associ-
ation of cardiovascular disease risk factors with socio-
1129
economic position during childhood and during adult-
hood. British Medical Journal 313, 1434±1438.
Carstairs, V. and Morris, R. (1991) Deprivation and
Health in Scotland. Aberdeen University Press,
Aberdeen.
Davey Smith, G., Hart, C., Blane, D., Gillis, C. and
Hawthorne, V. (1997) Lifetime socioeconomic position
and mortality: prospective observational study. British
Medical Journal 314, 547±552.
Department of Health and Social Security (1980)
Inequalities in health: report of a research working group.
DHSS, London.
Easton, D. F., Peto, J. and Babiker, A. G. (1991) Floating
Absolute Risk: an alternative to relative risk in survival
and case-control analysis avoiding an arbitrary reference
group. Statistics in Medicine 10, 1025±1035.
FaresjoÈ, T., SvaÈrdsudd, K. and Tibblin, G. (1994) Social
Mobility and Health in a Prospective Study of Middle-
Aged Men. Scandanavian Journal of Social Medicine 22,
86±89.
FaresjoÈ, T., SvaÈrdsudd, K. and Tibblin, G. (1997) The
concept of status incongruence revisited: a 22 year fol-
low-up of mortality for middle-aged men. Scandanavian
Journal of Social Medicine 25, 28±32.
General Register Oce (1966) Classi®cation of
Occupations 1966. HMSO, London.
Goldblatt, P. (1989) Mortality by social class, 1971±1985.
Population Trends 56, 6±15.
Goldthorpe, J. H. (1980) Social mobility and class structure
in modern Britain. Clarendon, Oxford.
Hart, C. L., Watt, G. C. M., Davey Smith, G., Gillis, C.
R. and Hawthorne, V. M. (1997) Pre-existing ischaemic
heart disease and ischaemic heart disease mortality in
women compared with men. International Journal of
Epidemiology 26, 508±515.
Hart, C. L., Davey Smith, G., Blane, D. (1998)
Inequalities in mortality by social class measured at
three stages of the lifecourse. American Journal of Public
Health, 88, 471±474.
Hawthorne, V. M., Watt, G. C. M., Hart, C. L., Hole, D.
J., Davey Smith, G. and Gillis, C. R. (1995)
Cardiorespiratory disease in men and women in
Scotland: baseline characteristics of the Renfrew/Paisley
(Midspan) study population. Scottish Medical Journal
40, 102±107.
Illsley, R. (1986) Occupational class, selection and inequal-
ities in health. Quarterly Journal of Social A€airs 2,
151±165.
Lynch, J. W., Kaplan, G. A., Cohen, R. D., Kauhanen,
J., Wilson, T. W. and Smith, N. L. et al. (1994)
Childhood and adult socioeconomic status as predictors
of mortality in Finland. Lancet 343, 524±527.
Marshall, G., Rose, D., Newby, H., Vogler, C. (1988)
Social class in modern Britain. Unwin Hyman, London.
Medical Research Council (1965) De®nition and classi®-
cation of chronic bronchitis for epidemiological pur-
poses. Lancet i, 775±779.
NystroÈm, A. M. (1992) Childhood environment, interge-
nerational mobility, and adult health Ð evidence from
Swedish data. Journal of Epidemiology and Community
Health 46, 71±74.
Payne, G. (1987) Mobility and change in modern society.
Macmillan, London.
Power, C., Matthews, S. and Manor, O. (1996)
Inequalities in self rated health in the 1958 birth cohort:
lifetime social circumstances or social mobility? British
Medical Journal 313, 449±453.
Prineas R. J., Crow R. S., Blackburn H. (1982) The
Minnesota manual of electrocardiographic ®ndings. John
Wright, Bristol.
Prospective Studies Collaboration (1995) Cholesterol, dias-
tolic blood pressure, and stroke: 13000 strokes in
1130 C. L. Hart et al.
450000 people in 45 prospective cohorts. Lancet 346,
1647±1653.
Rahkonen, O., Arber, S. and Lahelma, E. (1997) Health-
related social mobility: a comparison of currently
employed men and women in Britain and Finland.
Scandanavian Journal of Social Medicine 25, 83±92.
Rose, G. A., Blackburn, H., Gillum, R. F., Prineas, R. J.
(1982) Cardiovascular Survey Methods. WHO, Geneva.
SAS Institute Inc. (1989) SAS/STAT User's Guide,
Version 6, 4th edn., Vol. 2. SAS Institute Inc., Cary,
NC.
Stern, J. (1983) Social mobility and the interpretation of
social class mortality di€erentials. Journal of Social
Policy 12, 27±49.
Wannamethee, S. G., Whincup, P. H., Shaper, G. and
Walker, M. (1996) In¯uence of fathers' social class on
cardiovascular disease in middle-aged men. Lancet 348,
1259±1263.
West, P. (1991) Rethinking the health selection expla-
nation for health inequalities. Social Science and
Medicine 32, 373±384.