Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Social Science & Medicine 293 (2022) 114676
Available online 22 December 2021
0277-9536/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
The consequences of early menopause and menopause symptoms for labour
market participation
Alex Bryson
a
,
*
, Gabriella Conti
a
,
b
, Rebecca Hardy
a
, Darina Peycheva
a
, Alice Sullivan
a
a
UCL Social Research Institute, UK
b
UCL’s Department of Economics, UK
ARTICLE INFO
JEL classication:
I1
J01
Keywords:
Menopause
Early menopause
Menopausal symptoms
Vasomotor symptoms
Employment
Full-time employment
Birth cohort
ABSTRACT
Using a difference-in-difference estimator we identify the causal impact of early menopause and menopause
symptoms on the time women spend in employment through to their mid-50s. We nd the onset of early natural
menopause (before age 45) reduces months spent in employment by 9 percentage points once women enter their
50s compared with women who do not experience early menopause. Early menopause is not associated with a
difference in full-time employment rates. The number of menopause symptoms women face at age 50 is asso-
ciated with lower employment rates: each additional symptom lowers employment rates and full-time
employment rates by around half a percentage point. But not all symptoms have the same effects. Vasomotor
symptoms tend not to be associated with lower employment rates, whereas the employment of women who suffer
psychological problems due to menopause is adversely affected. Every additional psychological problem asso-
ciated with menopause reduces employment and full-time employment rates by 1–2 percentage points, rising to
2–4 percentage points when those symptoms are reported as particularly bothersome.
1. Introduction
The onset of menopause can lead to a range of health problems which
can be debilitating. Symptoms can be numerous and can persist for some
years. They can include physical health difculties, such as vasomotor
problems (hot ushes, and night sweats), psychological issues (such as
tearfulness, irritability, anxiety and depression) or a combination of
symptoms. The early onset of menopause (before age 45) can lead to
additional health complications. Although there is a growing social
science and medical literature on the antecedents to early menopause,
and increased recognition of the needs of those suffering debilitating
symptoms, there is little research into the employment and career con-
sequences of menopause for women. Indeed, the issue is still taboo in
some quarters, leading Financial Times journalist Janina Conboye to
argue that “in most workplaces it remains an uncomfortable topic”
(Conboye, 2021). Nevertheless, some maintain that the menopause re-
sults in millions of days lost through absenteeism and that, when faced
with employers’ lack of understanding and poorly developed policies
and practices to assist women facing menopausal issues, many women
choose to leave employment rather than tackle their employers on the
issue.
The small literature on the impact of menopause on labour market
participation, reviewed in Section Two, suggests that it can have sub-
stantial impacts on women’s absenteeism, their career progression, and
their treatment by colleagues and supervisors. However, in a recent
systematic review of the literature Brewis et al. (2017: 67) conclude:
“there is no work in the evidence base that estimates the cost of the
menopause transition for women’s economic participation in the UK”.
We contribute to this literature by providing those estimates. We do
so by identifying the causal effect of early menopause and menopausal
symptoms on the time women spend in employment and full-time
employment through to age 55 for a birth cohort of women born in a
single week in 1958. Using a difference-in-difference estimation strategy
described in Section Three we nd the early onset of menopause (before
age 45) reduces months spent in employment by 9 percentage points
once women enter their 50s compared with women who do not expe-
rience early menopause. Early menopause is not associated with a dif-
ference in full-time employment rates. The number of menopause
symptoms women face as they approach age 50 is associated with lower
employment rates: each additional symptom lowers employment rates
and full-time employment rates by around half a percentage point. These
effects are larger for symptoms which women say “bother me a lot”. For
* Corresponding author. UCL Social Research Institute, 20 Bedford Way, London, WC1H 0AL, UK.
E-mail address: a.bryson@ucl.ac.uk (A. Bryson).
Contents lists available at ScienceDirect
Social Science & Medicine
journal homepage: www.elsevier.com/locate/socscimed
https://doi.org/10.1016/j.socscimed.2021.114676
Received 4 August 2021; Received in revised form 6 December 2021; Accepted 20 December 2021
Social Science & Medicine 293 (2022) 114676
2
each bothersome symptom employment rates fall by around 2 percent-
age points, and full-time employment rates fall by a little over half a
percentage point. But not all symptoms have the same effects. Vaso-
motor symptoms tend not to be associated with lower employment rates,
whereas the employment of women who suffer psychological problems
due to menopause is affected. Every additional psychological problem
associated with menopause reduces employment and full-time
employment rates by 1–2 percentage points, rising to 2–4 percentage
points when those symptoms are reported as particularly bothersome.
The remainder of the paper is set out as follows. Section Two reviews
the previous literature on the effects of menopause on labour market
outcomes. Section Three describes the data and estimation strategy used
in the paper. Section Four presents ndings and Section Five concludes.
2. Previous literature
Menopause is the point in a woman’s life when she stops menstru-
ating and has been 12 months without a menstrual period. This
commonly occurs around the age of 50, though the transition (peri-
menopause) begins earlier and symptoms can begin during the peri-
menopause and may continue for some years after menopause. Those
symptoms can be many and varied, and include sweats, joint aches and
pains, hot ushes, night sweats, trouble sleeping, fatigue, palpitations,
dizziness, severe headaches and migraines, irritability and mood swings,
anxiety and depression, tearfulness, panic, forgetfulness and poor con-
centration. There is some debate as to which symptoms are specically
related to menopause rather than other age-related transitions (Mishra
and Kuh, 2012; Kuh et al., 1997). Women can experience a variety of
these symptoms in combination and experience them as more or less
‘bothersome’. We shall present data on their incidence and how ‘both-
ersome’ the symptoms were in our sample of women born in 1958 in
Section Three.
In a recent critical review of the literature Atkinson et al. (2021)
argue that symptoms experienced around menopausal transition can
affect women’s experience of work but, at the same time, work can
exacerbate a woman’s symptoms. They cite studies suggesting that
women’s concerns about the reactions of colleagues and supervisors
make it difcult to discuss menopausal symptoms at work: “where
women disclose, they may be brushed aside, made fun of, criticised,
bullied or become subject to performance management and ongoing
capability monitoring” (op. cit. p.52). The authors cite three UK
employment tribunal cases in which women have successfully argued
that they have been subject to unfair dismissal and direct sex discrimi-
nation based on their menopausal status. However, it appears more
common for women simply to quit their jobs when confronted by such
attitudes. In most cases, women choose not to disclose their
menopause-related health problems (Grifths et al., 2013).
In the absence of support, women often over-compensate by
“working extremely hard to hide their self-perceived shortcomings
resulting from their menopausal symptoms” (Kopenhager and Guidozzi,
2015: 373). This often results in emotional exhaustion and ‘burnout’
(Converso et al., 2019). In a large American survey of women in midlife,
those reporting menopausal symptoms reported signicantly lower
levels of health-related quality of life, work impairment and greater
healthcare utilization than observationally equivalent women who did
not report menopausal symptoms (Whiteley et al., 2013).
Not all studies nd menopause impacts women at work. For
example, in a small study of midlife women in the UK Hardy et al. (2018)
found menopausal status was not associated with women’s self-reported
work performance and absence. Using panel data from the American
National Longitudinal Survey of Young Women Mvundura (2007) nds
few differences in labour force participation during menopause transi-
tion, and no effect of early menopause. But in reviewing the literature
Brewis et al. (2017: 24–29) indicate that most studies nd a negative
relationship between menopausal symptoms and performance at work,
and that these are more pronounced with more severe symptoms.
A sizeable proportion of perimenopausal and menopausal women
not in receipt of hormone therapy report moderate or severe menopausal
symptoms (Reed et al., 2009). In a large American study those who
sought medical care for their symptoms, as identied through medical
health care insurance claims, had signicantly higher sickness absence
and lower hourly and annual productivity relative to a matched sample
of women not making such claims (Kleinman et al., 2013). A much
smaller survey of Dutch women in midlife indicated that those with
severe menopausal symptoms were more likely to have serious problems
dealing with the physical and mental demands of their work, risking
prolonged sickness absence (Geukes et al., 2016). These studies lend
support to the idea that those suffering particularly severe symptoms
face particularly adverse employment consequences.
The most effective treatment for serious adverse symptoms of
menopause is Hormone Replacement Therapy (HRT). There are con-
cerns regarding adverse health consequences of HRT including a po-
tential elevated risk of ovarian cancer (Collaborative Group on
Epidemiological Studies of Ovarian Cancer (2015). Nevertheless, using
data from the American Medical Expenditure Panel Survey Daysal and
Orsini (2014) nd HRT use increases short-term employment of women
aged 40 to 55 by 2.4 percent. The short-term employment effects are
estimated to have been much larger among a group of women who were
induced to stop taking HRT due to adverse publicity about the long-term
health consequences of HRT. (Concerned about the potential endoge-
neity of HRT use due to women with greater motivation to work taking
up HRT they instrument for its use using the panel structure of their
data. Specically, they use the age of women at the time public health
warnings were issued following the publicity surrounding ndings from
the Women’s Health Initiative Study (WHIS) in 2002 and its impact on
subsequent HRT take-up).
In their review of the literature for the Department of Education
Brewis et al. (2017: 64) suggest that bothersome physical and psycho-
logical menopause symptoms may affect both the extensive and inten-
sive margins of women’s labour supply. At the extensive margin, loss of
employment results in lower earnings and employment benets, as well
as accrued pension rights, together with the lost psychological benets
of working such as self-esteem.
The menstrual cycle prior to menopause also impacts women’s
earnings and career progression. In their study of personnel records from
a single Italian bank Ichino and Moretti (2009) show the absence taking
of women under age 45 follows a 28-day cycle that is not apparent
among men nor among women aged 45 and over. This is consistent with
a causal impact of the menstrual cycle on absence taking. The increase in
absenteeism due to the menstrual cycle feeds through to lower earnings,
accounting for 2 of the 13.5 percentage point gender wage gap among
these workers. It also explains part of the gender gap in the probability of
promotion to management.
3. Data and estimation
3.1. Data
Our data are the National Child Development Survey (NCDS) which
is a birth cohort survey following all those born in one week in England,
Scotland and Wales in 1958 (http://www.cls.ucl.ac.uk/ncds). We use
data collected at birth and in subsequent follow up surveys at ages 7, 11,
16, 23, 33, 42, 44, 46, 50 and 55 including work histories from which we
construct participation in employment and full-time employment pre-
and post-menopausal onset. Because ours are panel data containing
multiple observations on individuals over time we cluster standard er-
rors at the level of the individual in our estimation.
Our dependent variables are time in employment and time in full-
time employment derived from work histories in which survey re-
spondents report the month and year in which employment spells star-
ted and ended. We derive cohort members’ main employment activity
each month for the 35 years from age 20 to age 55 and then use these to
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
3
establish what the respondent’s main activity was over the course of the
year. Periods of employment include all paid work (as an employee or
self-employed), whether the respondent knows hours of work or not.
Full-time employment includes only those spells as an employee or self-
employed where the respondent says she works at least 30 h per week.
Of the 4897 women in the sample 583 had missing work history data
in at least 5 years. These observations were dropped from the estimation
sample. We reran estimates including these cases to see how sensitive
results were to their omission. Doing so meant imputing missing years
using the averages across those years in which their work history data
were not missing. Results were robust to their inclusion.
Our data permit us to identify women who had surgery removing
their ovaries or womb and those who had hormone replacement therapy
(HRT) before the end of menstruation. These can affect the experience of
menopause and its timing, as well as inuencing labour market pros-
pects. We remove these women (unweighted N =909) from our baseline
estimation sample but check the sensitivity of our results to their in-
clusion. We refer to these robustness checks in the text.
Among the estimation sample (those with fewer than 5 years missing
work history data who had not had surgery or HRT, N =3405) the mean
employment rate between the ages of 20 and 55 was 72 percent (mini-
mum of 59 percent, maximum of 83 percent) while the mean full-time
employment rate was considerably lower at 47 percent (minimum of
39 percent and a maximum of 59 percent).
To illustrate trends in women’s employment over the life-course
Fig. 1 presents employment rates (blue line) and full-time employment
rates (red line) by age between ages 20 and 55 for those in our esti-
mation sample. Women’s employment rates rise from their late 20s,
peaking in their late 40s, after which point they fall. Full-time
employment follows a very different pattern. Women’s relatively high
full-time employment rates fall precipitously during their 20s, reaching
a low point in their early 30s before rising steadily in their 30s and 40s
before dipping once again in their 50s.
3.1.1. Menopause treatment variables
Information on menstrual periods was collected at age 44/45, 50 and
55 in the NCDS. Women were asked if they had menstrual periods in the
past 12 months. In the presence of amenorrhea, they were asked if they
had had periods in the past 3 months. Women with no periods in the past
12 months were asked about their age and month at their nal menstrual
period and the reason for amenorrhea. All women were also asked about
changes in the regularity of their menstrual cycles in the last few years or
before their nal menstrual period. Natural menopause was dened as
at least 12 consecutive months of amenorrhea not induced by surgery or
other medical treatment. If this occurred before age 45 we classied
women as having gone through early menopause. The age cut off for
early menopause we use is also the one used by the National Health
Service (https://www.nhs.uk/conditions/early-menopause/) and in the
literature (Mishra et al., 2019). Five percent of our estimation sample
had done so (Table 1).
Our analyses of the impact of menopause symptoms on employment
and full-time employment are based on the number and type of meno-
pause symptoms reported in the 12 months prior to the survey interview
at age 50. Women were asked: “Some women report a number of
symptoms associated with menopause. The next few questions are about
any symptoms you may have experienced over the last 12 months”. They
were asked whether or not they had suffered any symptoms in the last 12
months and, if so, the extent to which the symptom “didn’t bother me”,
“bothered me a little” or “bothered me a lot”. Where respondents said “a
lot” we counted these as “bad” symptoms. They were prompted with 20
symptoms namely: trouble sleeping; joint aches and pains; breast
tenderness; hot ushes; palpitations; dizziness; pins and needles in
hands; skin crawling sensation; irritability; anxiety/depression; tear-
fulness; feelings of panic; forgetfulness; cold sweats/night sweats;
vaginal dryness; difculties with intercourse; more frequent passing of
urine; passing urine when didn’t mean to; painful passing of urine;
frequent severe headaches/migraine. They were also asked to identify
any others. Respondents reported up to 22 symptoms with a mean of
7.55, and up to 18 ‘bothersome’ symptoms with a mean of 1.63.
We ran factor analysis from a tetrachoric correlation matrix for the
binary variables for each symptom with varimax rotation to establish
whether there were high inter-item correlations between some symp-
toms suggesting that they loaded on a common factor. We identied two
factors with eigenvalues above one with high scale reliability co-
efcients. The rst was based on six items capturing aspects of psy-
chological health, namely anxiety/depression, tearfulness, panic,
forgetfulness, palpitations and irritability. Using these six items we
constructed an additive scale, which we termed psychological health
symptoms. The scale had a reliability coefcient (alpha) of 0.73 and ran
from 0 to 6 with a mean of 2.76. The second factor captured vasomotor-
related symptoms, namely hot ushes and night sweats, together with
trouble sleeping which is likely a result of the night-time symptoms. The
alpha scale reliability coefcient for this three-item scale was 0.64 and
had a mean of 1.76.
3.1.2. Control variables
Throughout we condition on variables derived from the rst six
waves of NCDS through to age 33. The choice of controls was informed
by parallel work using these data examining the antecedents to early
menopause (Peycheva et al., 2021) and our knowledge of factors
inuencing employment outcomes over the life-course in NCDS (Parsons
et al., 2021; Bryson et al., 2020). Initially we experimented with the full
set of risk factors linked to menopausal experiences identied by Pey-
cheva et al. (2021) in their review of the literature and their own
empirical work with NCDS (and the British Cohort Study). The nal set
of controls used in our estimation are the subset of variables which
played some role in one or more models. They include the following sets
Fig. 1. Employment and full-time employment rates over the life-course.
Table 1
Descriptive information on early menopause and menopause symptoms.
Min Max Mean St.
Dev.
Treatment:
Early menopause 0 1 .051 .219
Number of menopause symptoms 0 22 7.55 4.13
Number of ‘bothersome’ menopause symptoms 0 18 1.63 2.39
Number of psychological health menopause
symptoms
0 6 2.66 1.84
Number of ‘bothersome’ psychological health
menopause symptoms
0 6 .551 1.12
Number of vasomotor symptoms 0 3 1.67 1.13
Number of ‘bad’ vasomotor symptoms 0 3 .275 .446
Notes: (1) This baseline sample excludes those on HRT and those who had womb
or ovaries removed. (2) N observations =3405.
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
4
of variables. First, those collected at the time of the cohort member’s
birth:
- father’s social class provided by the cohort member’s mother and
coded to the Registrar-General’s social classes;
- birth weight recorded by the midwife;
- number and age of siblings.
Second, we include controls collected during the cohort member’s
childhood, namely:
- time breast-fed (collected from the mother at Sweep 1 in 1965);
- score on conduct behavioural adjustment problems at age 7 using the
Rutter parental questionnaire (Rutter et al., 1970). Our additive scale
sums the number of times the mother identies a problem that ap-
plies to their child;
- score on the Bristol Social Adjustment Guide completed by the
teacher when the child was aged 7 (Shepherd, 2013; Stott, 1987);
- count of childhood illnesses and health difculties reported by
mother at age 7;
- standardised reading and maths scores administered to cohort
members at age 11 by teachers (Shepherd, 2012);
- Body Mass Index (BMI) calculated in medical examination at age 16;
- ever smoked asked of cohort member at 16;
- had an alcoholic drink in week before cohort member interviewed at
age 16;
- frequency of physical exercise asked of cohort member at age 16.
Third, we include controls collected at later sweeps from the cohort
member in adulthood, namely:
- number of children at ages 23 and 33;
- highest qualication at age 33;
- self-assessed health, long-standing illness, number of health prob-
lems, and score on a Malaise Inventory capturing psychological
distress and depression (Rutter et al., 1970), all recorded at age 33;
- and ever taken a contraceptive pill by age 42.
We also include the number of survey sweeps responded to and the
number of years in which work history data are missing to net-out
sources of noisiness in the data. To retain sample size we retain cases
with item non-response on categorical variables by incorporating an
additional ‘missing’ category. We recode missing observations on
continuous variables to their mean values and add a dummy variable to
identify those observations (Little and Rubin, 2020). In the robustness
checks including those with HRT or surgery we also incorporate dummy
variables for these variables Descriptive statistics on these controls for
the estimation sample are presented in Appendix Tables A1 and A2.
3.2. Difference-in-difference estimation
To estimate the impact of menopause symptoms and early meno-
pause on employment we adopt a difference-in-difference strategy
which compares employment rates for affected women (the ‘treated’
women) and unaffected women (the ‘control’ group) in a pre-treatment
period with their employment rates in a post-treatment period. Let the
variable M denote menopausal treatment. In the case of early meno-
pause this is a dummy variable where 0 indicates individuals who do not
go through early menopause, the control group, and 1 indicates in-
dividuals who go through early menopause. The difference in the mean
differences between the treated and control groups in the pre- and post-
treatment periods is our estimate of the impact of early menopause on
employment participation. In the case of menopause symptoms M is a
count variable identifying the number of symptoms a woman reports, so
our estimates will capture the intensity with which menopause symp-
toms were experienced.
Formally the difference-in-difference estimator takes the following
form:
yi=∝+βMi+γti+δ(Mi.ti) +
ε
i
where y
i
is our outcome of interest – either employment or full-time
employment – for individuals i, ∝ is a constant term; β is the coef-
cient capturing the treatment group specic effect in the pre-treatment
period (to account for average permanent differences between the
treated and control groups); γ is the time trend common to control and
treatment groups; δ is the parameter capturing the treatment effect,
which is the interaction between treatment and being observed in the
post-treatment period;
ε
is a random unobserved error term containing
all determinants of yi which are omitted from our model.
To estimate effects of menopause using a difference-in-difference
strategy one needs to establish a moment in the lifecycle that clearly
pre-dates menopause (‘before’) and one that post-dates menopause
(‘after’). This is not straightforward because although there are identi-
able reproductive stages in women’s lives (pre-menopause, perimen-
opause, climacteric and post-menopause) there is debate about when
these stages stop and start, in part reecting heterogeneity in women’s
experiences (Harlow et al., 2012; Brewis et al., 2017: 17–19).
In our data menopause symptoms are captured at age 50 for the
previous year, while early menopause status is identied as the cessation
of menstruation before age 45, a variable derived from information
asked of women at surveys undertaken at ages 44/45 (biomedical
sweep), 50 and 55 (see Section 3.1.1).
We measure employment rates in the ‘before’ stage as employment
between the ages of 20 and 33. Employment rates in the ‘post’ period are
measured between age 50 and age 55. The ti variable capturing time in
the above estimator is a categorical variable dividing age into the pre-
period (before age 33), which is used as the reference category, the
period between ages 33 and 49, and the post-period between ages 50
and 55. It is the interaction between the ‘post-period’ and treatment
status that captures the difference-in-difference impact of menopause on
employment.
We cluster standard errors at the level of the individual to account for
autocorrelation and heteroskedasticity. Using a robust estimator we
found no evidence of serial correlation, something that can affect
difference-in-difference estimates (Bertrand et al., 2004).
3.2.1. Common trends assumption
The credibility of causal inference in difference-in-difference esti-
mation relies on the assumption that, in the absence of treatment, trends
over time in the dependent variable would have been common across
treated and control groups. The assumption is not testable because the
counterfactual employment trends that would have obtained in the post-
period for treated in the absence of their treatment is not observed.
Instead, it is common practice to test for common trends in the pre-
period. If this test is satised one might reasonably infer common
trends would have obtained in the post-period in the absence of treat-
ment. We do not account for potential biases that may arise when
treatment effects may vary over time (Goodman-Bacon, 2021).
We test for common trends by running employment and full-time
employment models through to age 33 incorporating the controls
described above, together with terms interacting the cohort member’s
age with subsequent treatment status. We test for the joint statistical
signicance of these interaction terms. If an F-test rejects their joint
signicance we can assert that employment trends did not depart
signicantly in the pre-period among those who go on to be treated and
those who are their controls. The p-values for those F-tests are presented
in Appendix Table A3. They are statistically non-signicant (except in
one instance when the dependent variable is full-time employment, but
even here they are only on the margins of statistical signicance). We
infer, therefore, that the common trends assumption on which causal
inference with difference-in-difference is based is not violated in our
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
5
case. Results are almost identical throughout when we rerun the models
with the age*interaction terms only having dropped the control
variables.
4. Results
4.1. Early menopause
Fig. 2 illustrates the employment patterns of women who go on to
experience early menopause versus those who do not. The employment
rates of women who experience early menopause rise until their 40s at
which point they atten before declining in their mid-40s. In contrast,
the employment rates of those who do not experience early menopause
continue to rise until their 50s. Both groups of women experience rapid
declines in their employment rates in their 50s, but the decline is much
steeper for those who had experienced early menopause. It is notable
that employment rates for women who go onto experience early
menopause are below other women in their 20s too, something we might
not have expected. Although the dip in their employment in their early
20s appears steeper than that for other women, this proves not to be the
case statistically (see Section 3.2.1).
The picture for participation in full-time employment in Fig. 3 looks
quite different. The full-time rate of those who go on to experience early
menopause lies below that of those who do not in their 20s, though the
decline in full-time employment rates over that period follows a similar
course. Full-time employment rates for those who experience early
menopause appear to track other women’s full-time employment rates
subsequently. If anything, those who experience early menopause seem
to be a little more likely to be found in full-time employment.
Table 2 presents the employment and full-time employment rates of
women who experience early menopause and those who do not for three
separate episodes in their lives, namely between the ages of 20 and 32,
33 to 49, and 50 to 55. The differences in the rates up to age 32 and from
age 50 to 55 capture the raw difference-in-difference estimates without
controlling for potential confounders. The 9-percentage point decit in
employment rates associated with early menopause is statistically sig-
nicant at a 95 percent condence level. There is no signicant differ-
ence in full-time employment rates.
Table 3 presents difference-in-difference estimates of the impact of
early menopause on employment and full-time employment rates
implementing the equation in Section 3.2. The full models are presented
in Appendix Table A4. Column 1 shows that the employment rates
among women rose signicantly after age 33 compared to earlier years
(0.113, t =19.27) then fall a little when women are in their 50s, though
they remained well above their employment rates in their 20s and early
30s (0.076, t =9.89). There is no signicant difference in the employ-
ment rates of women who go on to experience early menopause and
those who do not when they are in their 20s and early 30s (−0.006, t =
0.27), nor in their mid-30s and 40s (−0.019, t =0.70). However, the
employment rates of women in their 50s are signicantly lower among
women who had experienced early menopause than among those who
had not, when compared against differences in their employment rates
before the age of 33: the difference-in-difference is around 9 percentage
points (−0.086, t =2.35). The difference is a little smaller when the
estimation sample includes women who had surgery removing their
ovaries or womb and those who had hormone replacement therapy
(HRT) before the end of menstruation (−0.07, t =2.61). The difference-
in-difference estimate of early menopause effects on employment rates is
also a little smaller when estimated on the sample including those
women with missing work histories for 5 or more years (−0.06, t =1.63
excluding HRT/surgery cases and −0.06, t =2.17 including HRT/sur-
gery cases).
However, there is no evidence from column 2 that early menopause
affected the full-time employment rates of women (0.023, t =0.60). The
implication is that the reduction in employment rates among those
experiencing early menopause is primarily related to their reduced
likelihood of engaging in part-time employment. These results are
apparent whether one conditions on other potential confounders
(Table 3) or not (Table 2).
The point estimates and standard errors in both the employment and
full-time employment models are identical when using a robust esti-
mator to account for potential serial correlation.
Since our denition of early menopause is at least 12 months of
amenorrhea before age 45 it is arguable that we should be using age 45
and over as the cut-off dening the post-treatment employment spells. If
we do this, results are very similar to those reported above, with a
difference-in-difference estimate of employment rates of 8 percentage
points (−0.081, t =2.44) and no signicant effect on full-time
employment rates (0.028, t =0.73).
4.2. Number of menopause symptoms in the year before age 50
As noted earlier, women cited up to 22 different menopause symp-
toms they were suffering in the year prior to their survey interview at
age 50. Few (2.7 percent) suffered no symptoms, with the mean being 8
symptoms. The effects of menopause symptoms on women’s employ-
ment and full-time employment are presented in columns 1 and 2 of
Table 4 respectively.
Fig. 2. Employment Rates over the Life-course: those who experience early
menopause versus those who do not.
Fig. 3. Full-time Employment Rates over the Life-course: those who experience
early menopause versus those who do not.
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
6
Each additional symptom reduced women’s employment rates in
their 50s by half a percentage point (−0.005, t =2.85), compared to
employment rates before age 33. This was also the case with respect to
full-time employment rates (−0.005, t =2.23). (The estimated effects
were identical in the absence of conditioning variables.) Thus, a woman
experiencing the mean number of symptoms might expect a reduction in
her employment and full-time employment rates by around 4 percentage
points, compared to a similar woman with no menopausal symptoms.
The effects of symptoms were similar when extending the sample to
include those who had been on HRT or had surgery (−0.008, t =4.89 in
the case of employment rates and −0.005, t =2.48 for full-time
employment rates).
Going on to experience menopause symptoms just before age 50 was
associated with a lower employment rate and lower full-time employ-
ment rate when women were in their 20s and early 30s as indicated by
the coefcient for menopause symptoms in row 3 of Table 4. However,
as noted earlier and as indicated in Appendix Table A2 row 2, employ-
ment trends did not differ in women’s 20s and early 30s according to the
number of menopausal symptoms suffered subsequently, suggesting the
common trends assumption holds.
Table 5 presents similar estimates where women report ‘bothersome’
menopause symptoms. These affected over half (56 percent) of women
aged 50. The mean number of bothersome symptoms experienced was
1.9 symptoms. Each additional bothersome symptom lowered employ-
ment rates for women in their 50s by 1.7 percentage points (−0.017, t =
5.11) relative to those who had no bothersome symptoms, having
controlled for employment differences prior to age 33 (column 1). Ef-
fects were about half the size for full-time employment (−0.008, t =2.14
in column 2).
Once again, coefcients and standard errors for these effects were
identical in the absence of control variables. Effects of bothersome
symptoms on employment were even larger with the retention of HRT
and surgery cases in the sample (−0.022, t =8.27) but the effects for
full-time employment were nearly identical (−0.007, t =2.36).
4.3. Number of psychological health menopause symptoms in the year
before age 50
As noted earlier, our factor analysis identied a factor based on six
menopause symptoms relating to aspects of psychological health,
namely anxiety/depression, tearfulness, panic, forgetfulness, palpita-
tions and irritability. To establish the impact of these psychological
health symptoms on women’s employment we summed those six
symptoms and interacted them with women’s age to identify the dif-
ference in employment rates in their 50s between women who did and
did not suffer these symptoms by age 50, compared to differences in
these women’s employment when they were aged less than 33 years. In
identifying the effects of mental health problems brought on by meno-
pause we ignore the other menopausal symptoms women may suffer.
They are not controlled for in this analysis.
Table 6 column 1 presents the employment impact of psychological
health problems accompanying menopause. The coefcient interacting
the number of psychological health symptoms with being aged 50 or
over indicates that employment rates fell by 1.3 percentage points
(−0.013, t =3.08) for each additional psychological health symptom.
This suggests that women with the mean number of psychological health
Table 2
Employment and Full-time Employment Rates by age for Women Who Experience Early Menopause and Those Who Do Not.
Employment Rates Full-time Employment Rates
No early menopause Early menopause Difference No early menopause Early menopause Difference
20–32 years .68 .65 -.03 (1.61) .55 .53 -.02 (0.82)
33–49 years .80 .74 -.06 (2.27)** .47 .47 +.01 (0.24)
50+years .76 .64 -.12 (3.67)*** .46 .46 +.00 (0.03)
All .75 .69 -.06 (2.87)** .50 .49 -.00 (0.18)
Difference-in-difference -.09 (2.35)** +.02 (0.60)
Notes: (1) Sample N =119,175 person-year observations and N =3405 persons (2) Those using HRT and those with surgery are excluded. (3) The difference-in-
difference gures in the bottom row are simply the difference in employment and full-time employment rates in the pre-treatment (aged 20–32) and post-
treatment (aged 50+) periods for those who go through early menopause and those who do not. (4) t-statistics in parentheses (5) * =signicant at 90% con-
dence level; ** =signicant at 95% condence level; *** =signicant at 99% condence level.
Table 3
Difference-in-difference estimates of the impact of early menopause on
employment and full-time employment.
Employment Full-time employment
Age (ref: 20–32 years)
33–49 Years .113 (19.27)*** -.084 (10.70)***
50+Years .076 (9.89)*** -.091 (9.60)***
Early Menopause -.006 (0.27) -.008 (0.34)
Early*33–49 -.019 (0.70) .030 (0.91)
Early*50+-.086 (2.35)** .023 (0.60)
Constant .334 (5.44)*** .507 (7.08)***
Adj. r-sq. .089 .107
Notes: (1) Sample N =119,175 person-year observations and N =3405 persons
(2) Those using HRT and those with surgery are excluded. (3) OLS estimation
with standard errors clustered at the individual level. (4) Models contain all
control variables included in Appendix Tables A1 and A2. (5) t-statistics in pa-
rentheses (6) * =signicant at 90% condence level; ** =signicant at 95%
condence level; *** =signicant at 99% condence level.
Table 4
Difference-in-difference estimates of the impact of menopause symptoms on
employment and full-time employment.
Employment Full-time employment
Age (ref: 20–32 years)
33–49 Years .128 (10.97)*** -.083 (5.30)***
50+Years .111 (7.12)*** -.053 (2.78)***
Menopause Symptoms -.002 (2.04)** -.004 (3.24)***
N symptoms*33–49 -.002 (1.57) .000 (0.02)
N symptoms*50+-.005 (2.85)*** -.005 (2.23)**
Constant .351 (5.66)*** .555 (7.53)***
Adj. r-sq. .090 .101
Notes: (1) See Table 3 for notes.
Table 5
Difference-in-difference estimates of the impact of ‘bothersome’ menopause
symptoms on employment and full-time employment.
Employment Full-time employment
Age (ref: 20–32 years)
33–49 Years .120 (17.35)*** -.088 (9.61)***
50+Years .099 (10.97)*** -.078 (6.92)***
N Bothersome Menopause Symptoms -.004 (1.98)** -.005 (2.23)**
N bothersome symptoms*33–49 -.005 (2.02)** .004 (1.22)
N bothersome symptoms*50+-.017 (5.11)*** -.008 (2.14)**
Constant .361 (5.86)*** .528 (7.32)***
Adj. r-sq. .092 .107
Notes: (1) See Table 3 for notes.
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
7
symptoms due to menopause (2.76) had an employment rate in their 50s
which was 3.6 percentage points lower than it would have been in the
absence of those symptoms. Effects on full-time employment, presented
in column 2, are comparable (−0.015, t =2.93). Effects are identical in
the absence of controls. And they are similar when including those on
HRT and those who had undergone surgery (−0.019, t =5.02 for
employment and −0.014, t =3.23 for full-time employment).
Table 7 runs similar estimates but focuses on ‘bothersome’ psycho-
logical health symptoms which are relatively uncommon: women report
a mean of 0.61 bothersome psychological health symptoms associated
with menopause. Their employment effects are sizeable: a single
‘bothersome’ psychological health menopause symptom leads to a 3.9
percentage point drop in employment rates after age 50 relative to
women without such symptoms (−0.039, t =5.59). The effect is a little
smaller for full-time employment (−0.025, t =3.25). Results are robust
to the removal of control variables. Employment effects are a little larger
when retaining the HRT and surgery cases in the sample (−0.049, t =
8.18) but smaller for full-time employment (−0.022, t =3.34).
4.4. Number vasomotor symptoms in the year before age 50
The factor analysis identied a factor capturing vasomotor-related
symptoms, namely hot ushes and night sweats, together with trouble
sleeping which is likely a result of the night-time symptoms. It has a
mean of 1.76. One-sixth (19 percent) of women reported no vasomotor
symptoms in the year approaching their 50th birthday, while one-third
(35 percent) reported all three symptoms.
Table 8 reports estimates of the impact of vasomotor symptoms on
women’s employment in their 50s based on our preferred sample which
excludes women who had surgery removing their ovaries or womb and
those who had HRT before the end of menstruation. The number of
vasomotor symptoms reported does not affect employment (column 1)
or full-time employment (column 2) rates: both coefcients for the
vasomotor and over-50s interaction are small and statistically non-
signicant. The number of vasomotor symptoms led to a statistically
signicant reduction in employment (−0.012, t =2.02) once the model
was run on an estimation sample including women who had HRT and
those who had surgery, but the effect on full-time employment remained
statistically non-signicant (0.004, t =0.47). Furthermore, when we
include women who had HRT and those who had surgery the common
trends assumption is not supported. The F-tests reveal that the age*-
treatment interactions are jointly statistically signicant in the pre-
period through to age 32.
Two-thirds (69.6 percent) of women reported no bothersome vaso-
motor symptoms. Only six percent reported all three vasomotor symp-
toms as being bothersome. The mean number was 0.51. Table 9 presents
the effects of bothersome vasomotor symptoms on employment and full-
time employment rates. Every additional bothersome vasomotor
symptom leads to a reduction of over 2.5 percentage points (−0.026, t =
2.91) in women’s employment in their 50s but has no effect on their full-
time employment rates (−0.004, t =0.39). Results are insensitive to the
exclusion of controls. Employment effects are stronger (−0.036, t =
4.74) when the estimation sample is extended to include women on HRT
and those who had had surgery, while the effects on full-time employ-
ment remain non-signicant (−0.001, t =0.11).
5. Conclusions
Our paper is the rst to estimate the effects of early menopause and
menopausal symptoms on employment and full-time employment rates
among women. We exploit prospective birth cohort data for all women
born in a particular week in 1958 to estimate the causal effects of
menopause on employment rates using a difference-in-difference strat-
egy. This technique compares the gap in employment rates during their
Table 6
Difference-in-difference estimates of the impact of psychological health meno-
pause symptoms on employment and full-time employment.
Employment Full-time
employment
Age (ref: 20–32 years)
33–49 Years .132
(13.27)***
-.065 (4.87)***
50+Years .105 (8.06)*** -.052 (3.19)***
N Psychological Health Menopause
Symptoms
-.003 (1.31) -.008 (2.71)***
N Psychological Health Menopause
symptoms*33–49
-.008 (2.45)** -.007 (1.67)
N Psychological Health Menopause
symptoms*50+
-.013
(3.08)***
-.015 (2.93)***
Constant .346 (5.60)*** .540 (7.51)***
Adj. r-sq. .090 .109
Notes: (1) See Table 3 for notes.
Table 7
Difference-in-difference estimates of the impact of ‘bothersome’ psychological
health menopause symptoms on employment and full-time employment.
Employment Full-time
employment
Age (ref: 20–32 years)
33–49 Years .121
(19.04)***
-.079 (9.21)***
50+Years .093
(11.29)***
-.076 (7.40)***
N bothersome Psychological Health
Menopause Symptoms
-.003 (0.67) -.005 (0.97)
N bothersome Psychological Health
Menopause symptoms*33–49
-.016
(3.05)***
-.007 (1.11)
N bothersome Psychological Health
Menopause symptoms*50+
-.039
(5.59)***
-.025 (3.25)***
Constant .353 (5.73)*** .526 (7.32)***
Adj. r-sq. .092 .108
Notes: (1) See Table 3 for notes.
Table 8
Difference-in-difference estimates of the impact of vasomotor menopause
symptoms on employment and full-time employment.
Employment Full-time employment
Age (ref: 20–32 years)
33–49 Years .104 (10.18)*** -.105 (7.74)***
50+Years .084 (6.15)*** -.093 (5.57)***
N vasomotor symptoms -.004 (1.00) -.004 (0.96)
N vasomotor symptoms*33–49 .005 (0.91) .013 (1.97)*
N vasomotor symptoms*50+-.007 (1.05) .002 (0.20)
Constant .337 (5.46)*** .515 (7.16)***
Adj. r-sq. .089 .107
Notes: (1) See Table 3 for notes.
Table 9
Difference-in-difference estimates of the impact of bothersome vasomotor
menopause symptoms on employment and full-time employment.
Employment Full-time
employment
Age (ref: 20–32 years)
33–49 Years .114
(17.43)***
-.093 (10.69)***
50+Years .083 (9.77)*** -.088 (8.36)***
N bothersome vasomotor symptoms -.001 (0.17) -.003 (0.49)
N bothersome vasomotor
symptoms*33–49
-.004 (0.64) .022 (2.50)**
N bothersome vasomotor
symptoms*50+
-.026 (2.91)*** -.004 (0.39)
Constant .335 (5.46)*** .505 (7.03)***
Adj. r-sq. .089 .107
Notes: (1) See Table 3 for notes.
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
8
20s and early 30s with the employment gap in their 50s for women who
went onto experience early menopause versus those who did not. We
make similar comparisons between women according to the intensity
with which they experienced menopausal symptoms when aged 50. In
doing so we control for a rich array of variables collected at birth, in
childhood, and in early adulthood which can affect employment pros-
pects and experiences of menopause. We show employment and full-
time employment trends during their 20s and early 30s did not differ
signicantly between the ‘treated’ group – those who went on to expe-
rience early menopause or more menopausal symptoms – and their
‘control’ groups who did not experience early menopause or did not
suffer many menopausal symptoms. This provides some assurance that
their employment rates may have trended in similar fashions later in
their lives if they had not experienced menopause differently.
We nd women’s employment rates, and their full-time employment
rates fall as the number of menopausal symptoms they report rises. Ef-
fects are larger for symptoms that are reported as ‘bothersome’. The
effects are quantitatively large. For instance, a woman who experiences
the mean number of menopausal symptoms at age 50 can expect to have
an employment rate in her 50s that is 4 percentage points lower than a
woman who has no menopausal symptoms.
Different types of menopause symptom have different employment
effects. For instance, vasomotor symptoms do not affect full-time
employment rates, and they only affect employment rates where they
are considered ‘bothersome’. In contrast, psychological health problems
associated with menopause signicantly lower employment and full-
time employment rates, and effects are much larger when those symp-
toms are ‘bothersome’.
Early menopause is associated with a very large (9 percentage point)
reduction in employment rates once women reach their 50s, yet it has no
statistically signicant effect on women’s full-time employment rates. It
is unclear why early menopause should affect employment rates, but not
full-time employment rates. This issue is worthy of further investigation.
It is striking that the inclusion and exclusion of potential confounders
makes very little difference to the impact of early menopause and
menopause symptoms on employment and full-time employment rates.
Even though their inclusion increases the variance in employment rates
explained by our models (as indicated by the adjusted R-squared) the
coefcient and statistical signicance of the interaction capturing the
impact of menopause are nearly identical in all cases. Following Oster
(2019) we take coefcient stability in the face of adjustments to con-
ditioning covariates as an indication that results are unlikely to be
biased by omitted variables.
There are some limitations to this study. First, although women are
asked specically to identify health-related symptoms due to the
menopause, in some cases those symptoms may be due to other changes
women are going through at the same time which are not directly linked
to the menopause. Second, our data only collect information on symp-
toms related to menopause in the year leading up to the survey interview
at age 50. Some women may have experienced symptoms earlier which
did not persist to age 50, leading to some error in our ability to accu-
rately capture symptoms related to menopause. Some women who
experienced symptoms, but not at age 50, will be misclassied as having
no symptoms. However, assuming symptoms experienced earlier than
age 50 also have a detrimental impact on employment, this will mean
our estimates of symptoms’ effects on employment are downwardly
biased. Third, it is worth recalling that the Great Recession hit when the
women in the study reached age 50. This was a very severe recession
creating what were, at the time, unprecedented labour market problems
for many. It would be valuable to see whether our results are replicated
in more benign labour market conditions.
These negative employment effects of early menopause and meno-
pausal symptoms are cause for some concern, not only because the size
of the effects is large, but also because so many women suffer these
problems. As we have shown, the mean number of menopausal symp-
toms experienced by women in this birth cohort when aged 50 was 8,
including 2 particularly ‘bothersome’ symptoms. Five percent of women
in the estimation sample had experienced early menopause.
These employment effects of early menopause and menopause
symptoms add to the personal costs they have for women suffering from
them in terms of their physical and mental health, and potentially their
effects on women’s private lives, although we do not quantify them here.
They also have costs for society, in terms of the health care costs of
treating women’s symptoms, potential productivity losses from
women’s lost hours of work and ability to work productively. It is
conceivable that they will also affect women’s retirement decisions and
thus pension entitlements.
Having identied the size and extent of the problem government and
employers should consider steps that could be taken to ameliorate the
problems women face in their working lives due to the menopause. That
said, this is the rst study of its kind, so there is value in seeking to
replicate and extend research investigating the impact of early meno-
pause and menopausal symptoms on labour market outcomes. First, it
would be valuable to know whether the effects we identify might vary
for other cohorts of women including more recent entrants to the labour
market. Second, there would be value exploring the heterogeneity of
menopausal effects and whether there are aspects of women’s experi-
ences that may ameliorate the effects of menopause. For instance, it may
be that women are better able to manage menopause symptoms where
they have greater opportunities to manage working patterns or working
hours, as might be the case among self-employed women or employees
in workplaces with policies and practices expressly intended to assist
women affected by menopause. Third, we know very little about the
effects of early menopause and menopausal symptoms on other aspects
of women’s labour market experiences. We would have a better picture
if studies were undertaken to investigate the impacts of menopause on
women’s wellbeing at work, their job satisfaction and their earnings.
Finally, we know of no studies piloting policies or practices in the
workplace that might assist women in raising health-related problems
they may have during menopause, nor in coping with those problems.
These evaluations are needed to provide the evidence base employers
and government need so they know what actions to take to improve
women’s working lives.
Credit author statement
Conceptualization: Alex Bryson, Gabriella Conti, Rebecca Hardy,
Darina Peycheva, Alice Sullivan, Methodology: Alex Bryson, Gabriella
Conti, Formal analysis: Alex Bryson, Gabriella Conti, Darina Peycheva,
Resources: Darina Peycheva, Writing – original draft: Alex Bryson,
Writing – review & editing: Alex Bryson, Gabriella Conti, Rebecca
Hardy, Darina Peycheva, Alice Sullivan, Project administration: Alice
Sullivan, Funding acquisition: Alice Sullivan.
Acknowledgment
We acknowledge the ESRC for access to the data and the Health
Foundation for funding (grant number 546608).
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
9
Appendices.
Appendix Table A1
Descriptive Statistics for Categorical Control Variables (n =3405)
Variables: Percent
Social class of father at birth:
I and II Professional and Managerial/Technical 19.9
III Non-manual 9.8
III Manual 43.4
IV and V Partly Skilled and Unskilled 17.7
No father in household 4.4
Missing 4.7
Birth weight (kilograms):
<2.5 4.8
2.5–2.9 19.0
3–3.49 36.3
3.5–3.9 22.5
≥4 6.9
Missing 10.4
Breast fed:
No 25.6
Yes, <1 month 20.9
Yes ≥1 month 41.7
Missing 11.8
Siblings at birth:
None 22.2
Older sibs only 21.8
Younger sibs only 25.6
Older and younger sibs 21.3
Missing 9.2
BMI at age 16:
Underweight 1.1
Normal 58.5
Overweight 6.1
Very overweight 1.6
Missing 32.5
Ever smoked by age 16:
Never 54.3
Ever 23.6
Missing 22.1
Had alcoholic drink in week before age 16 interview:
No 46.2
Yes 31.8
Missing 22.0
Physical activity at age 16:
Monthly or less often 46.8
Weekly 29.9
Missing 23.4
Highest qualication age 23:
None 11.0
CSE 10.6
O-level 26.1
A-level 14.2
Degree 19.2
Post-graduate 0.3
Missing 18.5
Number of children at age 23:
None 67.2
1 24.2
2 6.8
3+1.2
Missing 0.5
Highest qualication age 33:
None 8.9
CSE 11.8
O-level 29.0
A-level 13.4
Degree 26.9
Post-graduate 1.4
Missing 8.5
Number of children at age 33:
None 22.6
1 23.1
2 34.2
3+15.6
Missing 4.4
General health at age 33:
Poor 0.9
(continued on next page)
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
10
Appendix Table A1 (continued )
Variables: Percent
Fair 8.6
Good 47.1
Excellent 33.8
Missing 9.6
Long-standing illness, disability or inrmity at age 33:
No 79.7
Yes 11.5
Missing 8.8
Malaise score at age 33:
Normal 84.4
Malaise (8+) 5.7
Missing 9.9
Ever taken contraceptive pill by age 42:
No 9.5
Yes 75.5
Missing 15.0
Appendix Table A2
Descriptive Statistics for Continuous Control Variables (n =3405)
Variables: Mean Min Max
Rutter scale 5.59 0 19
N Rutter items missing 2.72 0 23
Score on Bristol Social Adjustment Guide 5.61 0 47
If missing on BSAG 0.10 0 1
Number of illnesses/health difculties as a child 4.44 0 17
Reading score at age 11 0.19 −2.54 3.02
Reading score at age 11 missing 0.13 0 1
Maths score at age 11 0.19 −1.61 2.26
Maths score missing at age 11 0.13 0 1
Number of physical health problems ever had through to age 33 (retrospective) 1.36 0 15
Number of survey interviews conducted 5.53 2 6
Number of years work history missing 0.86 0 5
Appendix Table A3
Joint Statistical Signicance of Age*Treatment Interactions between Age 20 and Age 32
Employment Full-time Employment
Treatment:
Early menopause .511 .632
Number of menopause symptoms .321 .105
Number of ‘bothersome’ menopause symptoms .212 .099*
Number of psychological health menopause symptoms .525 .948
Number of ‘bothersome’ psychological health menopause symptoms .622 .442
Number of vasomotor symptoms .172 .102
Number of ‘bothersome’ vasomotor symptoms .795 .486
Notes: (1) Figures are p-values for Prob >F capturing joint signicance of age*treatment interactions using Stata’s testparm command. (2) All
models exclude HRT/surgery cases. (3) N observations =44,265 for 3405 individuals (4) * denotes signicance at a 90% condence level.
Appendix Table A4
Difference-in-Difference Estimates of the Impact of Early Menopause on Employment and Full-time Employment (Full Model)
Employment Full-time Employment
Age (ref: 20–32 years)
33–49 Years .113 (19.27)*** -.084 (10.70)***
50+Years .076 (9.89)*** -.091 (9.60)***
Early Menopause -.006 (0.27) -.008 (0.34)
Early*33–49 -.019 (0.70) .030 (0.91)
Early*50+-.086 (2.35)** .023 (0.60)
Social class of father at birth (ref: I and II)
III Non-manual .045 (3.44)*** .031 (1.73)*
III Manual .027 (2.74)*** .043 (3.33)***
IV and V .018 (1.38) .034 (2.12)**
No father in household .031 (1.50) .101 (3.92)***
Missing -.004 (0.15) -.024 (0.76)
Birth weight in kilograms (ref: 3-3.49 kg):
(continued on next page)
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
11
Appendix Table A4 (continued )
Employment Full-time Employment
<2.5 .006 (0.36) -.000 (0.00)
2.5–2.9 .002 (0.20) .010 (0.81)
3.5–3.9 .002 (0.26) .009 (0.77)
≥4 .008 (0.56) .017 (0.87)
Missing .020 (1.19) .039 (1.85)*
Breastfed (ref: No):
Yes, <1 month -.000 (0.04) -.004 (0.31)
Yes ≥1 month -.011 (1.22) -.007 (0.56)
Missing -.027 (0.62) .049 (0.79)
Siblings at birth (ref: None):
Older sibs only -.009 (0.82) -.005 (0.38)
Younger sibs only -.008 (0.75) .013 (0.98)
Older and younger sibs -.011 (0.93) .026 (1.83)*
Missing -.014 (0.71) -.019 (0.82)
Rutter scale .001 (0.83) -.001 (0.35)
N Rutter items missing .000 (0.18) -.002 (0.64)
BSAG score -.001 (1.27) -.002 (0.26)
If missing on BSAG .015 (0.86) .016 (0.70)
Number of illnesses/health difculties as a child -.003 (1.66) .001 (0.35)
BMI at age 16 (Ref: Normal)
Underweight -.073 (2.13)** -.070 (1.71)*
Overweight .016 (1.23) .018 (0.99)
Very overweight .007 (0.23) .075 (1.99)**
Missing -.029 (2.84)*** -.002 (0.17)
Ever smoked by age 16 (Ref: Never):
Ever .004 (0.53) .004 (0.37)
Missing -.008 (0.10) -.016 (0.23)
Had alcoholic drink in week before age 16 interview (ref: No):
Yes .003 (0.31) .017 (1.06)
Missing .028 (0.35) -.002 (0.03)
Physical activity at age 16 (ref: Monthly or less often)
Weekly .034 (4.24)*** .027 (2.55)***
Missing .006 (0.21) .067 (1.84)*
Reading score at age 11 .003 (0.48) .017 (2.05)**
Reading score missing .253 (7.41)*** .048 (1.17)
Maths score at age 11 -.002 (0.30) .003 (0.34)
Maths score missing -.248 (7.32)*** -.036 (0.88)
Highest qualication age 23 (ref: None)
CSE -.014 (0.54) -.029 (0.93)
O-level .036 (1.87)* .039 (1.68)*
A-level -.029 (1.43) -.014 (0.54)
Degree -.006 (0.34) -.027 (1.05)
Post-graduate -.145 (2.22)** -.132 (1.59)
Missing .027 (1.24) .020 (0.80)
Highest qualication age 33 (ref: None)
CSE .101 (3.69)*** .074 (2.31)**
O-level .065 (2.89)*** .052 (1.99)**
A-level .096 (4.09)*** .087 (3.07)***
Degree .102 (4.63)*** .103 (3.86)***
Post-graduate .082 (2.60)*** .075 (1.50)
Missing -.185 (4.27)*** -.374 (6.41)***
Number of children at age 23 (ref:None)
1 -.026 (2.51)** -.005 (0.36)
2 -.069 (3.82)*** -.009 (0.42)
3+-.088 (1.95)** -.019 (0.46)
Missing -.085 (1.07) -.108 (1.17)
Number of children at age 33 (ref:None)
1 -.067 (6.79)*** -.195 (13.94)***
2 -.110 (11.87)*** -.301 (23.38)***
3+-.184 (13.13)*** -.375 (21.39)***
Missing .051 (1.74)* .026 (0.69)
General health at age 33 (ref: excellent)
Poor -.103 (2.07)** -.031 (0.55)
Fair -.058 (3.46)*** -.034 (1.75)*
Good .007 (0.84) -.001 (0.06)
Missing .069 (1.87)* .065 (1.44)
Long-standing illness, disability or inrmity at age 33 (ref: No)
Yes -.042 (3.27)*** -.030 (1.91)*
Missing .183 (3.77)*** .340 (5.52)***
Malaise score at age 33 (ref: Malaise 8+)
Normal .047 (2.50)*** .042 (2.07)**
Missing -.025 (0.47) -.064 (1.32)
Number of physical health problems ever had through to age 33 .001 (0.34) -.001 (0.35)
Ever taken contraceptive pill by age 42 (ref: yes)
No -.054 (3.75)*** -.042 (2.70)***
Missing .006 (0.57) .023 (1.64)
(continued on next page)
A. Bryson et al.
Social Science & Medicine 293 (2022) 114676
12
Appendix Table A4 (continued )
Employment Full-time Employment
Number of survey interviews conducted .061 (6.90)*** .018 (1.77)*
Number of years work history missing -.017 (5.85)*** -.021 (5.96)***
Constant .334 (5.44)*** .507 (7.08)***
Adj. r-sq. .089 .107
Notes: (1) Sample N =119,175 person-year observations and N =3405 persons (2) Those using HRT and those with surgery are excluded. (3) OLS
estimation with standard errors clustered at the individual level. (4) t-statistics in parentheses (5) * =signicant at 90% condence level; ** =
signicant at 95% condence level; *** =signicant at 99% condence level.
References
Atkinson, C., Beck, V., Brewis, J., Davies, A., Duberley, J., 2021. Menopause and the
workplace: new directions in HRM research and HR practice. Hum. Resour. Manag.
J. 31, 49–64.
Bertrand, M., Duo, E., Mullainathan, S., 2004. How much should we trust differences-
in-differences estimates? Q. J. Econ. 119 (1), 249–275.
Brewis, J., Beck, V., Davies, A., Matheson, J., 2017. The Effects of Menopause Transition
on Women’s Economic Participation in the UK. Department for Education, London.
Bryson, A., Joshi, H., Wielgoszewska, B., Wilkinson, D., 2020. A short history of the
gender wage gap in Britain. Oxf. Rev. Econ. Pol. 36 (4), 836–854.
Collaborative Group on Epidemiological Studies of Ovarian Cancer, 2015. Menopausal
hormone use and ovarian cancer risk: individual participant meta-analysis of 52
epidemiological studies. Lancet 385, 1835–1842.
Converso, D., Viotti, S., Sottimano, I., Loera, B., Molinengo, G., Guidetti, G., 2019. The
relationship between menopausal symptoms and burnout. A cross-sectional study
among nurses. BMC Wom. Health 19 (148), 1–12.
Daysal, N.M., Orsini, C., 2014. The Miracle Drugs: Hormone Replacement Therapy and
Labor Market Behavior of Middle-Aged Women. IZA Discussion Paper No. 7993.
Geukes, M., van Aalst, M.P., Robroek, S.J.W., Laven, J.S.E., Oosterhof, H., 2016. The
impact of menopause on work ability in women with severe menopause symptoms.
Maturitas 90, 3–8.
Goodman-Bacon, A., 2021. Difference-in-differences with variation in treatment timing.
J. Econom. https://doi.org/10.1016/j.jeconom.2021.03.014.
Grifths, A., MacLennan, S.J., Hassard, J., 2013. Menopause and work: an electronic
survey of employees’ attitudes in the UK. Maturitas 76, 155–159.
Hardy, C., Thorne, E., Grifths, A., Hunter, M.S., 2018. Work outcomes in midlife
women: the impact of menopause, work stress and the working environment.
Women’s Midlife Health 4 (3), 2–8.
Harlow, S.D., Gass, M., Hall, J.E., Lobo, R., Maki, P., Rebar, R.W., Sherman, S., Sluss, P.
M., de Villiers, T.J., 2012. Executive summary of the Stages of Reproductive Aging
Workshop +10: addressing the unnished agenda of staging reproductive aging.
Menopause 19 (4), 387–395.
Ichino, A., Moretti, E., 2009. Biological gender differences, absenteeism and the earnings
gap. Am. Econ. J. Appl. Econ. 1 (1), 183–218.
Janina Conboye, J., 2021. Breaking the Silence Around the Menopause. Financial Times,
4th February. https://www.ft.com/content/da9132dd-530d-44bd-88a6-d309e8c6
d228?sharetype=blocked.
Kleinman, N.L., Rohrbacker, N.J., Bushmakin, A.G., Whiteley, J., Lynch, W.D., Shah, S.
N., 2013. Direct and indirect costs of women diagnosed with menopause symptoms.
J. Occup. Environ. Med. 55 (4), 465–470.
Kopenhager, T., Guidozzi, F., 2015. Working women and the menopause. Climacteric 18
(3), 372–375.
Kuh, D.L., Wadsworth, M., Hardy, R., 1997. Women’s health in midlife: the inuence of
the menopause, social factors and health in earlier life. Br. J. Obstet. Gynaecol. 104,
923–933.
Little, R.J.A., Rubin, D.B., 2020. Statistical Analysis with Missing Data, third ed. Wiley,
Hoboken, NJ.
Mishra, G.D., Kuh, D., 2012. Health symptoms during midlife in relation to menopausal
transition: British prospective cohort study. Br. Med. J. 344, e402.
Mishra, G.D., Chung, H., Canob, A., Chedrauic, P., Goulisd, D.G., Lopese, P., Mueckg, A.,
Reesh, M., Levent, M., Senturki, M., Simoncinij, T., Stevensonk, J.C., Stutel, P.,
Tuomikoskim, P., Lambrinoudakin, I., 2019. Predictors of early menopause.
Maturitas 123, 82–88.
Mvundura, M., 2007. Menopause Transition and Labor Market Outcomes. Dissertation,
Georgia State University. https://scholarworks.gsu.edu/econ_diss/38.
Oster, E., 2019. Unobservable selection and coefcient stability: theory and evidence.
J. Bus. Econ. Stat. 37 (2), 187–204.
Parsons, S., Bryson, A., Sullivan, A., 2021. “Teenage Conduct Problems: a Lifetime of
Disadvantage in the Labour Market?”, UCL QSS Working Paper No, pp. 21–22.
Peycheva, D., Sullivan, A., Hardy, R., Bryson, A., Conti, G., Ploubidis, G., 2021. Risk
Factors for Early Natural Menopause: Evidence from the 1958 and 1970 British Birth
Cohorts. https://doi.org/10.1101/2021.09.12.21263444 medRxiv pre-print.
Reed, S.D., Ludman, E.J., Newton, K.M., et al., 2009. Depressive symptoms and
menopausal burden in the midlife. Maturitas 62, 306–310.
Rutter, M., Tizard, J., Whitmore, K., 1970. Education, Health and Behaviour. Longman,
London.
Shepherd, P., 2012. Measures of Ability at Ages 7 to 16: 1958 NCDS User Guide. Centre
for Longitudinal Studies. https://cls.ucl.ac.uk/wp-content/uploads/2017/07/NCDS-
user-guide-NCDS1-3-Measures-of-ability-P-Shepherd-December-2012.pdf.
Shepherd, P., 2013. Bristol Social Adjustment Guides at 7 and 11 Years: 1958 NCDS User
Guide. Centre for Longitudinal Studies. https://cls.ucl.ac.uk/wp-content/uploads/
2017/07/NCDS-Bristol-Social-Adjustment-Guides-nal.pdf.
Stott, D.H., 1987. The Social Adjustment of Children: Manual to the Bristol Social
Adjustment Guides. Hodder and Stoughton, London.
Whiteley, J., DiBonaventura, M.D., Wagner, J.-S., Alvir, J., Shah, S., 2013. The impact of
menopausal symptoms on quality of life, productivity and economic outcomes.
J. Wom. Health 22 (11), 983–990.
A. Bryson et al.
