Gender Differences in the Incidence of Depression and
Anxiety: Econometric Evidence from the USA+
Vani K Borooah*
University of Ulster
July 2009
Abstract
Using data from the Collaborative Psychiatric Epidemiology Surveys (CPES) for the
United States for the period 2001-2003, this paper addresses a vexed question relating
to inter-gender differences in depression rates, namely how much of the observed
difference in depression rates between men and women may be explained by
differences between them in their exposure, and how much may be explained by
differences between them in their response, to depression-inducing factors. The
contribution of this paper is to propose a method for disentangling these two
influences and to apply it to US data. The central conclusion of the paper was
differences between men and women in rates of depression and anxiety were largely
to be explained by differences in their responses to depression-inducing factors: the
percentage contribution of inter-gender response differences to explaining the overall
difference in inter-gender probabilities of being depressed was 93 percent for “sad,
empty” type depression”; 92 percent for “very discouraged” type depression; and 69
percent for “loss of interest” type depression.
Keywords: Gender, depression, anxiety, decomposition.
JEL Classification: I1, I3
+
The data used in this paper are from the Collaborative Psychiatric Epidemiology Surveys (CPES)
2001-2003 [United States] provided online by the Inter-University Consortium for Political and Social
Research (ICPSR) http://www.hcp.med.harvard.edu/ncs/. I am grateful to three anonymous referees
whose comments have greatly improved the paper. However, I am solely responsible for its
deficiencies.
*
School of Economics, University of Ulster, Newtownabbey, Co. Antrim BT370QB, Northern Ireland,
United Kingdom (vk.borooah@ulster.ac.uk).
.
1. Introduction
Some economists are beginning to question a (arguably, the) fundamental
belief that underpins our subject, namely that a better economic performance by a
country is in itself, and of itself, a "good thing". (Frank, 1997, 1999; Layard, 2002,
2003).1 In response to such concerns, studies (both econometric and noneconometric) about the nature of happiness, and about the factors underlying
happiness, have mushroomed.2 Since a prominent conclusion of such studies is that
mental ill-health is a major reason for being unhappy, this study examines two
specific aspects of mental ill-health: depression and anxiety. 3
The reasons for the emphasis on depression and anxiety are three-fold. The
first is the large number of people who are affected by these two ailments: the
Psychiatric Morbidity Survey (2000) estimates that 6 million persons in the United
Kingdom suffer from depression or anxiety or both, with 14.4 million persons
suffering similar disorders in the USA. The second reason is that depression and
anxiety, by preventing many of its afflicted citizens from working, may impose large
economic costs on a country: Layard (2006) estimates that the loss in output in the
UK due to depression and anxiety is some £12 billion (or 1 per cent of the UK’s
national income) per year. Lastly, there is a significantly large gender bias to
depression and anxiety with women being much more likely than men to have these
conditions: meta-analysis of studies conducted in various countries show that women
1
For a public policy approach to the pursuit of happiness see Marks (2004).
Inter alia Blanchflower and Oswald (2000); Clark (1996, 1999, 2001); Clark and Oswald (1994);
Easterlin (1974, 1995, 2001); Frank (1985; 1997, 1999); Frey and Stuzer (2002); Hirsch (1976);
Layard (2002, 2003); Oswald (1997); Scitovsky (1976).
3
For example, Borooah (2006) in a study of Northern Ireland reported that only four percent of those
with severe mental health problems described themselves as happy and 60 percent described
themselves as unhappy; equally tellingly, only 32 percent of those whose mental health problems were
not severe described themselves as happy - the same proportion as those with severe heart problems
who regarded themselves as happy.
2
1 .
are roughly twice as likely as men to suffer depression (Nolen-Hoeksema, 1990;
Weissman et. al., 1996).
Indeed, it is the unequal distribution of depression and anxiety between men
and women that is the focus of this paper. There is a large literature on differences
between men and women in their propensities to be depressed and Culbertson (1997),
Piccinelli and Wilkinson (2000), Nolen-Hoeksema (2001) and Murakami (2002) inter
alia provide a good perspective to this body of work. In explaining why rates of
depression are higher for women than for men, Nolen-Hoeksema (2001) distinguished
between two effects, the quantification of which constitutes the fundamental purpose
of this paper.
First, she argued that, compared to men, women might be more likely to be
exposed to depression-inducing factors. So, for example: women were more likely
than men to be the victims of childhood sexual assault; they were more likely to be
trapped in the role of perpetual carers, with their lives sandwiched between caring for
their young children and their aged parents; they were more likely to be unequal
partners in heterosexual relationships with major, life-changing decisions being made
by their male partners; and they were more likely to do atypical and non-standard type
work exemplified by temporary or part-time jobs.4
Second, even when men and women were exposed to the same depressioninducing factors, women might be more likely than men to develop depression. This
might be due to gender differences in the response to such factors caused inter alia
by: biological factors;5 differences in levels of self-esteem between men and women;6
4
On this last point, see Mangan (2000).
The hypothalamic-pituitary-adrenal (HPA) axis plays a major role in regulating stress responses and,
compared to men, women are more likely to have a dysfunctional HPA responses to stress (Weiss et.
al. 1999)
5
2 .
differences between men and women in their respective propensities to introspection
and rumination.7
Given these two effects – engendered, respectively, by gender differences in
exposure, and in response, to depression-inducing factors – the need is for an
integrative model, encompassing both exposure and response effects, to explain
differences in depression rates between women and men (Nolen-Hoeksema, 2001).8
In addition, it would be useful to quantify how much of the observed difference in
depression rates between men and women could be explained by differences between
them in their exposure, and how much could be explained by differences between
them in their response, to depression-inducing factors. The central purpose of this
paper is to build such a model and offer such quantification.
2. The Data
The data used in this paper are from the Collaborative Psychiatric
Epidemiology Surveys (CPES) for the United States for the period 2001-2003. These
data, which are described in some detail in Alegria et. al. (2007), present inter alia
information on the prevalence of mental disorders and on the personal and social
circumstances of the respondents all of whom were 18 years or older. The CPES joins
together three nationally representative surveys: the National Comorbidity Survey
Replication (NCS-R); the National Survey of American Life (NSAL), and the
National Latino and Asian American Study (NLAAS); in consequence, CPES permits
analysts to approach analysis of the combined dataset as though it were a single,
nationally representative study.
6
With the consequence that conflicts in, or the ending of, relationships were more likely to produce
depression in women than in men.
7
A greater propensity to rumination in response to stress increases the risk of developing depression
(Noel-Hoeksema et. al., 1999).
8
Another possibility is that gender differences in depression rates may be the result of men responding
to stress through alternative modes such as antisocial behaviour and alcohol abuse (Kessler et. al, 1994;
Metzler et. al. 1995.
3 .
The CPES dataset is organised in different files, each relating to a particular
aspect of respondents’ lives, and from these files this paper focused on two: the
“Screening” and the “Demographic” files. The “Screening” and “Demographic”
instruments were administered to all the respondents in the survey; the instruments
pertaining to the other files were only applied to those affected by one or (more)
mental disorder. Using information from the Screening file, we defined a person as
having experienced depression if he/she answered “yes” to any of the following
questions:
(i)
Have you ever in your life had a period, lasting several days or longer,
when most of the day you felt sad, empty, or depressed?
(ii)
Have you ever in your life had a period, lasting several days or longer,
when most of the day you were very discouraged about how things were
going in your life?
(iii)
Have you ever in your life had a period, lasting several days or longer,
when you lost interest in most things you usually enjoy like work, hobbies,
and personal relationships?
Similarly, a person was defined as having experienced mild anxiety if he/she
answered the following question in the affirmative: have you ever in your life had an
attack of fear or panic when all of a sudden you felt very frightened, anxious, or
uneasy? By extension, severe anxiety was defined as answering yes to the following
question: Have you ever had an attack when all of a sudden you became very
uncomfortable, you either became short of breath, dizzy, nauseous, or your heart
pounded, or you thought that you might lose control, die, or go crazy?9
9
A problem with self-reported information is that of recall. If it is a natural instinct to suppress
memories of unpleasant events, and if the young are more likely to be susceptible to depression and
anxiety, then older persons in a sample are likely to "forget" that they were depressed or anxious when
4 .
For the NCS-R, a total of 9,282 adult interviews were completed: 7,963 with
the main respondent and 1,589 interviews with the second adult in the household; in
addition, 554 interviews were conducted with a sample of non-respondents using a
shortened version of the instrument. The final response rate was 70.9 percent for
primary respondents and 80.4 percent for secondary respondents. For the NSAL, the
overall response rate was 71.5 percent while, for the NLAAS, the response rate was
75.7 percent.10
<Table 1 about here>
Table 1 shows that 44 percent (of the 5,862 women analysed), compared to 35
percent (of the 4,227 men analysed), had felt “sad, empty, or depressed”; 44 percent
of women, compared to 38 percent of men, had felt “very discouraged”; 33 percent of
women, compared to 29 percent of men, had “lost interest in most things”; 40 percent
of women, compared to 32 percent of men, had experienced mild anxiety; and 10
percent of women, compared to 8 percent of men, had experienced severe anxiety.11
So, for every facet of depression and anxiety, women were more likely than men to
have experienced that condition with the gender gap being largest for feeling “sad,
empty, depressed” and for mild anxiety and smallest for “losing interest” and for
severe anxiety. Since the data refer to self-reported depression or anxiety, the
possibility is that gender differences in depression rates may be the result of men
responding to stress through alternative modes such as antisocial behaviour and
alcohol abuse (Kessler et. al, 1994; Metzler et. al. 1995).
they were young while, for the younger persons in the sample such memories are likely to be vivid.
Consequently, in a cross-section of people, older, compared to younger, respondents would report
lower rates of depression or anxiety purely for reasons of differences in recall.
10
Gender specific response rates were not provided.
11
The proportions were computed for persons who reported non-missing values for all the variables
used in the logistic regressions (Table 4-8): 10, 089 persons of whom 5,862 were women and 4,227
were men.
5 .
<Table 2 about here>
Table 2 shows the distribution of depression and anxiety by non-gender
attributes. The highest rates of depression and anxiety were for White persons ("sad":
50 percent), followed by Hispanics ("sad": 46 percent) and the lowest rates were for
Asians ("sad": 31 percent). People below the age of 30 had markedly higher rates of
depression and anxiety than the over 60s ("sad": respectively, 43 and 34 percent) and
those who were married or cohabiting ("sad": 35 percent) had markedly lower rates of
depression and anxiety compared to the never married or the
separated/divorced/widowed ("sad": 44 and 48 percent, respectively). Better-off
persons (those whose income-to-poverty line ratio was higher than the mean ratio)
had slightly lower rates of depression and anxiety ("sad": 39 percent) compared to
poorer persons12 ("sad": 41 percent).
Persons born in the USA were considerably more likely to have experienced
depression and anxiety, compared to non-US born persons, ("sad": 43 and 37 percent,
respectively) and persons living in the west of the USA ("sad": 36 percent) were
markedly less likely to have experienced depression and anxiety than persons living
elsewhere in the USA ("sad": above 40 percent). Rates of depression and anxiety were
impervious to education level but there was a strong link between such rates and
employment status: people who were unemployed had markedly higher rates of
depression and anxiety than those in employment ("sad": 47 and 39 percent,
respectively). This is consistent with much of the literature on the connection
between unemployment and depression (Clark and Oswald, 1994; Clark et. al., 2008).
However, the direction of causation is open to question: does unemployment cause
12
Those whose income-to-poverty line ratio was lower than the mean ratio
6 .
depression or are depressed persons more likely to be made unemployed? Böckerman
and Ilkmakunnas (2009), using panel data for Finland, suggest that persons with lower
levels of self-assessed health were more likely to become unemployed.
Compared to those in bad physical health ("sad": 54 percent), persons in good
physical health ("sad": 37 percent) - and, compared to those who had known
childhood trauma ("sad": 52 percent),13 persons who had not experienced childhood
trauma ("sad": 29 percent) - had much lower rates of depression and anxiety. Lastly,
there appeared to be a strong association between cognitive and social disability14 and
rates of depression and anxiety.
<Table 3 about here>
Table 3 shows the distribution of the attributes, noted above, between men and
women. Compared to the female part of the sample, a larger proportion of males
were: Asian (23 versus 18 percent); married (59 versus 45 percent); employed (73
versus 61 percent); lived in the West (30 versus 24 percent); and had an income-topoverty line score higher than the mean score. Also, compared to the female part of
the sample, a smaller proportion of males were: separated/divorced/widowed (16
versus 28 percent); unemployed (7 versus 9 percent); US born (54 versus 60 percent).
So, on all these counts, the gender distribution of attributes was biased towards higher
rates of depression and anxiety for women.
Conversely, compared to men: a larger proportion of women were Black (49
versus 40 percent) and a smaller proportion were Hispanic (25 versus 28 percent); and
a smaller proportion of women had experienced childhood trauma (46 versus 55
13
Childhood traumas were any of the following: fidgety childhood; frequently in trouble with adults
for six months or more during childhood or adolescence; lying, stealing as child or teenager; ran away
frequently, played truant, or stayed out late as child or teenager; had separation anxiety, for one month
or more, as a child.
14
Both of these were measured by a person’s World Health Organisation Disability Assessment Score
(WHO-DAS): the higher the score, the greater the disability.
7 .
percent). So, on all these counts, the gender distribution of attributes was biased
towards lower rates of depression and anxiety for women.
The preceding discussion raises two issues. First, what was the contribution of
each of the factors, listed in Table 2, to the likelihood of a person experiencing
depression and anxiety, after controlling for the other factors? Table 2, and the
discussion based on it, refer to the contributions in the absence of any imposed
controls. This question was answered in the context of an estimated logit model the
results from which are discussed in the next section. The second issue relates to the
aggregate contribution that differences in the distribution of the different attributes
between men and women made to gender differences in rates of depression and
anxiety. Section 4 addresses this question in the context of a decomposition model
originally developed by Blinder (1973) and Oaxaca (1973) for measuring
discrimination in the labour market.
3. Estimates from a Logistic Model of Depression and Anxiety
We estimated a logistic model for a dependent variable Yi such that Yi=1, if
the person (i=1…N) has had a condition (depression, anxiety), Yi=0, otherwise. The
model was estimated on a vector of variables, X ij being the value of the jth variable
for the ith person (j=1…J).15 A natural question to ask from the logistic model is how
the probability of having a particular condition would change in response to a change
in the value of one of the condition-affecting factors. These probabilities are termed
marginal probabilities.
For discrete variables, the marginal probabilities refer to changes in the
probabilities consequent upon a move from the residual category for that variable to
J
Pr(Yi  1)
 exp{ X ij  j }  exp{zi } for J coefficients, βj and for
1  Pr(Yi  1)
j 1
observations on J variables.
15
The logit equation is
8 .
the category in question, the values of the other variables remaining unchanged. For
continuous variables, the marginal probabilities refer to changes in the probabilities
(of having the conditions) consequent upon a unit change in the value of the variable,
the values of the other variables remaining unchanged. Tables 4-6 show the
estimated marginal probabilities from the logistic model for, respectively: “felt sad,
empty depressed”; “felt very discouraged about how things were going in life”, and
“lost interest in most things one usually enjoyed” and Tables 7 and 8 show the
estimated marginal probabilities from the logistic model for, respectively, “mild
anxiety” and “severe anxiety”.
<Tables 4-8 about here>
The marginal probabilities are first shown for the model estimated over the
entire sample (10,087 persons: 5,861 women and 4,226 women) and then for the
model estimated over, respectively the female and male subsamples. The associated
z-values are shown alongside the marginal probabilities: a z value exceeding 1.96
indicates that the coefficient was significantly different from zero at a 5% significance
level. The rows of the table with typeface in italics are those variables for which the
difference between the female and male marginal probabilities was significantly
different from zero at the 20% (or less) level of significance.
So, for example, in Table 4, looking at the estimates obtained from the entire
sample, ceteris paribus the average probability of men having felt “sad, empty,
depressed” was 9.7 percentage points lower than that for women. The results
estimated across the female and male subsamples show that there were six statistically
significant (significance level: 20% or less) differences between women and men:
between Black women and men; between Hispanic women and men; between men
and women who were separated, divorced, or widowed; between employed women
9 .
and men; between women and men who had had childhood traumas; and between
women and men with respect to their WHO-DAS cognitive scores.
Compared to the probability of a white person having felt “sad, empty,
depressed”, the corresponding probability for a black person was ceteris paribus 12.2
points lower: row “Black”, column 2 of Table 4. However, compared to the
probability of a white woman having felt “sad, empty, depressed”, the corresponding
probability for a black woman was ceteris paribus 8.8 points lower and, compared to
the probability of a white man having felt “sad, empty, depressed”, the corresponding
probability for a black man was ceteris paribus 16.1 points lower: row “Black”,
columns 3 and 5, respectively, of Table 4. The results further suggest that the
marginal probability for black women (8.8 points) was significantly different from
that for black men (16.1 points).
A persistent feature of the results reported in Tables 4-8 was the importance of
race: ceteris paribus Asians, followed by Blacks, were much less likely to have
experienced depression and anxiety than White persons (see above discussion): see
Dunlop et.al. (2003). Another important feature of the results was marital status:
ceteris paribus compared to never married persons, separated/divorced/widowed
persons were more likely (in the case of feeling "sad, empty, depressed" by 5.8
points) to have experienced depression (though not anxiety). The effect of marital
status on depression, however, varied by gender: compared to never married men,
men who were separated/divorced/widowed were significantly more likely to have
experienced all three forms of depression ("sad, empty, depressed": 9.6 points; "very
discouraged": 7.8 points; "loss on interest": 9.7 points); compared to never married
women, women who were separated/divorced/widowed were significantly more likely
to have only experienced feeling “sad, empty, depressed” (by 4.3 points) without any
10 .
significant difference in the likelihood of feeling “very discouraged” or a “loss of
interest”.
A third feature of the results was the importance of childhood trauma in
determining whether a person, male or female, would experience depression and
anxiety as an adult: compared to their counterparts who had not known childhood
trauma, men and women who had were more likely, by approximately 22 percentage
points, to have experienced all three forms of depression and, by approximately 20
points, to have experienced mild anxiety.
4. Gender Decomposition of the Probabilities of Depression and
Anxiety
The Oaxaca (1973) and Blinder (1973) method (hereafter, O-B) of
decomposing differences between groups, in their respective mean values, into a
“discrimination” and a “characteristics” component is, arguably, the most widely used
decomposition technique in economics. This method has been extended from its
original setting within regression analysis, to explaining group differences in
probabilities derived from models of discrete choice with a binary dependent variable
and estimated using logit/probit methods (Nielsen, 1998).
The O-B decomposition (and its extension) is formulated for situations in
which the sample is subdivided into two mutually exclusive and (collectively
exhaustive) groups, such as, for example, men and women.
Then, one may
decompose the difference in, for example, average wages between men and women –
or the difference between men and women in their average probabilities of being
depressed – into two parts, one due to gender differences in the coefficient vectors
and one due to gender differences in the attribute (or variable) vectors.
The attribute contribution is computed by asking what the average malefemale difference in probabilities would have been if the difference in attributes
11 .
between men and women had been evaluated using a common coefficient vector. The
critical question though is: what should be this common coefficient vector?
Typically, two separate computations of the attribute contribution are provided using,
respectively, the male and the female coefficient vectors as the common vector.
<Table 9 about here>
Column 1 of Table 9 shows the difference between men and women in the
average proportions with a particular condition. On average, compared to men,
women were: more likely by 8.9 percentage points to have felt “sad, empty,
depressed”; more likely by 6.4 points to have felt “very discouraged”; more likely, by
3.9 points to have felt a “loss of interest”; more likely by 7.7 points to have had mild
anxiety; and more likely by 2.0 points to have had severe anxiety.
Column 2 of Table 9 shows the amount of the overall gap that is due to the
attributes effect when female and male attributes are both evaluated using female
coefficients; similarly, column 4 of Table 9 shows the amount of the overall gap that
is due to the attributes effect when female and male attributes are both evaluated using
male coefficients. Three points need to be made about the attributes effect:
1.
The size of the attributes effect differs according to whether female or male
coefficients are used in the evaluation. For example, for “loss of interest”
depression, the attributes effect based on female coefficients explains 15
percent – while the attributes effect based on male coefficients explains 44
percent - of the overall difference in inter-gender rates.
2.
For some conditions, the attributes effect predicts that, if male attributes were
evaluated at female coefficients, then, compared to women, a higher
proportion of men would have had that condition. Feeling “sad, empty,
12 .
depressed”, feeling “very discouraged”, and mild anxiety are three such
conditions.
3.
In general, the contribution of the attributes effect towards explaining the
overall difference in inter-gender rates was small. The largest contributions
were recorded when male coefficients were used in the evaluation of female
and male attributes and these were 16 percent for “very discouraged”
depression and 44 percent for “loss of interest” depression.
The problem with the O-B method of decomposition is that the decomposition
is anchored either by treating men as women (column 2, Table 9) or women as men
(column 4, Table 9). More recently, Borooah and Iyer (2005) have proposed a
method of decomposition which combines both “anchors” into a single decomposition
formula.
Denote, by P W and P M , the average probabilities of having experienced a
condition, computed over all the persons in the sample, when their individual attribute
vectors (the X ki ) are all evaluated using the coefficient vectors of, respectively
women ( βW ) and men ( β M ); in other words, P W and P M are the average probabilities
of having had a condition, computed over the entire sample, when all the persons in
the sample are treated as, respectively, women and men. The difference between the
probabilities, PW  P M , represents the “response effect” because it is entirely the
consequence of differences between women and men in their (coefficient) responses
to a given vector of attributes.
Borooah and Iyer (2005) have shown that these synthetic probabilities can be
used to resolve the ambiguity of the O-B formulation since:
Y W  Y M  ( PW  P M )  the weighted average of the two attribute effects
13 .
where the two attribute effects are shown in columns 2 and 4 of Table 9, the weights
being the proportions of women and men in the sample.
Column 7 of Table 9 shows the difference between women and men in their
synthetic probabilities of experiencing a particular condition. Remembering that these
differences represent inter-gender response differences, the percentage contribution of
such differences to explaining the overall difference in probabilities was: 93 percent
for “sad, empty” depression”; 92 percent for “very discouraged” depression; 69
percent for “loss of interest” depression; 116 percent for “mild anxiety”; and 80
percent for “severe anxiety”.
5. Conclusions
The contribution of this paper to the extensive literature on gender differences
in depression rates was to apply the Oaxaca-Blinder decomposition methodology to
quantifying the contribution of differences in exposure, and differences in response,
to inter-gender differences in depression rates. Needless to say, the division of the
sample could have been by factors other than gender and, in the context of depression
and anxiety, it might be useful to examine inter-racial differences in the propensity
towards these conditions. As the data in Table 2 shows, reported rates of depression
and anxiety are considerably lower for Asians than for Whites. Is this due to the fact
that Asians are, perhaps, relatively sheltered from depression/anxiety inducing
factors? That the sample of Asians contained relatively more males? Or is it because
Asian responses to such factors (these responses perhaps engendered by a Confucian
stoicism which makes it culturally less acceptable to admit to emotional frailty than
might be the case for those with liberal, Western roots) are relatively more muted than
the responses of people of other races? Similar observations might be made about
regional divisions: does living in sunny California offer greater protection from
14 .
factors associated with depression and anxiety compared to living in the rust belts of
the Mid West? Or are people living in California better able to cope with such
factors?
This paper offered a methodology - with a long and distinguished pedigree in
economics - which is capable of providing answers to questions in which
responsibility needed to be apportioned between exposure and response. However, it
might be pertinent to conclude this paper by pointing to a limitation of this
methodology. It should be emphasised that the response effect was defined as a
residual: it was what could not be explained by differences between men and women
in their exposure to the various "depression-influencing" factors. Consequently, the
empirical results are only as good as the variables used in the logit regression; with a
different set of variables the exposure/response split might have been different.
The relevance of this observation to the present analysis is that several of the
variables used could not be better nuanced. For example, marital breakdown was a
"depression-inducing" factor but the data were silent on the circumstances
surrounding the breakdown; similarly, differences between men and women in the
nature of their employment, or in the nature of their work-home balance, could not be
elaborated upon. These examples provide arguments for marrying mental health
information to a richer set of data on individual circumstances.
15 .
References
Alegria, Margarita, Jackson, J.S., Kessler, R.C., and Takeuchi, D. (2007),
Collaborative Psychiatric Epidemiology Surveys (CPES), 2001-2003 [United States]
[Computer file]. ICPSR20240-v5, Ann Arbor, MI: Institute for Social Research,
Survey Research Center [producer], 2007. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2008-06-19.
Blanchflower, D. and Oswald, A. (2002), Well-Being Over Time in Britain
and the USA, NBER Working Papers, no. 7487, Cambridge, Mass.: National Bureau
of Economic Research.
Blinder, A.S. (1973), “Wage Discrimination: Reduced Form and Structural
Estimates”, Journal of Human Resources, vol. 8, pp. 436-455.
Böckerman, P. and Ilmakunnas, P. (2009). "Unemployment and Self-assessed
Health: Evidence from Panel Data", Health Economics, vol. 18, pp. 161-179.
Borooah, V.K. and Iyer, S. (2005, "The Decomposition of Inter-Group
Differences in a Logit Model: Extending the Oaxaca-Blinder Approach with an
Application to School Enrolment in India”, Journal of Economic and Social
Measurement , vol. 30, pp.279-93.
Borooah, V.K. (2006), “What Makes People Happy? Some Evidence From
Northern Ireland”, Journal of Happiness Studies vol. 7, pp. 427-65.
Clark, A.E. and Oswald, A. (1994), "Unhappiness and Unemployment",
Economic Journal, vol. 104, pp. 648-59.
Clark, A.E. (1996), "Job Satisfaction in Britain", British Journal of Industrial
Relations, vol. 34, pp. 189-217.
Clark, A.E. (1999), "Are Wages Habit Forming? Evidence from Micro Data"
Journal of Economic Behaviour and Organisation, vol. 39, pp. 179-200.
16 .
Clark, A.E. (2001), "What Really Matters in a Job? Hedonic Measurement Using
Quit Data", Labour Economics, vol. 8, pp. 223-242.
Clark A.E, Diener E, Georgellis Y, Lucas R. 2008, "Lags and leads in life
satisfaction: a test of the baseline hypothesis", Economic Journal vol. 118, pp. F222–
F243.
Culbertson, F.M. (1997), “Depression and Gender: An International Review”,
American Psychologist, vol. 52, pp. 25-31.
Dunlop, D.D, Song, J., Lyons, J.S., Manheim, L.M., and Chang, R.W., (2003)
"Racial/Ethnic Differences in Rates of Depression Among Preretirement Adults",
American Journal of Public Health , vol. 93, pp. 1945-1952.
Easterlin, R.A. (1974), "Does Economic Growth Improve the Human Lot? Some
Empirical Evidence", in P.A. David and M.W. Reder, Nations and Households in
Economic Growth: Essays in Honour of Moses Abramowitz, New York: Academic
Press.
Easterlin, R.A. (1987), Birth and Fortune: The Impact of Numbers on Personal
Welfare, Chicago: Chicago University Press (2nd edition).
Easterlin, R.A. (2001), "Income and Happiness: Towards a Unified Theory",
Economic Journal, vol. 111, pp. 465-484.
Frank, R.H. (1985), Choosing the Right Pond, Oxford: Oxford University Press.
Frank, R.H. (1997), "The Frame of Reference as a Public Good", Economic
Journal, vol. 107, pp. 1832-47.
Frank, R.H. (1999), Luxury Fever: Money and Happiness in an Era of Excess,
Princeton and Oxford: Princeton University Press.
Frey, B.S. and Stutzer, A. (2002), Happiness and Economics, Princeton, New
Jersey: Princeton University Press.
17 .
Hirsch, F. (1976), The Social Limits to Growth, Cambridge, Mass.: Harvard
University Press.
Kessler, R.C, McGonagle, K.A, Zhao, S. et al. (1994), “Lifetime and 12-month
prevalence of DSM-III-R psychiatric disorders in the United States. Results from the
National Comorbidity Survey”, Arch Gen Psychiatry vol. 51, pp.8-19.
Layard, R. (2002), Rethinking Public Economics: Implications of Rivalry and
Habit, Centre for Economic Performance, London: London School of Economics.
Layard, R. (2003), Happiness: Has Social Science a Clue? Lionel Robbins
Memorial Lectures 2002/3, London: London School of Economics.
Layard, R. (2006), The Depression Report: A New Deal for Depression and
Anxiety Disorders, Centre for Economic Performance, London: London School of
Economics.
Mangan, J. (2000), Workers Without Traditional Employment: an International
Study of non-Standard Work, Cheltenhan: Edward Elgar.
Meltzer H, Gill B, Petticrew M, Hinds K (1995), OCPS Surveys of Psychiatric
Morbidity in Great Britain, Report 1: The prevalence of psychiatric morbidity among
adults living in private households, London: Her Majesty's Stationery Office.
Murakumi, J. (2002), “Gender and Depression: Explaining the Different Rates
of Depression between Men and Women”, Perspectives in Psychology (the
undergraduate Psychology journal of the University of Pennsylvania), vol. 5, pp. 2734.
Nielsen, H.S. (1998), “Discrimination and Detailed Decomposition in a Logit
Model”, Economics Letters, vol. 61, pp. 115-20.
Noel-Hoeksema, S. (1990), Sex Differences in Depression, Stanford CA: Stanford
University Press.
18 .
Noel-Hoeksema, S., Larson, J., and Grayson, C. (1999), “Explaining the Gender
Difference in Depression”, Journal of Personality and Social Psychology, vol. 77, pp.
1061-72.
Noel-Hoeksema, S. (2001), “Gender Differences in Depression”, Current
Directions in Psychological Science, vol. 10, pp. 173-76.
Oaxaca, R. (1973), “Male-Female Wage Differentials in Urban Labor Markets”,
International Economic Review, vol. 14, pp. 693-709.
Oswald, A. (1997), "Happiness and Economic Performance", Economic Journal,
vol. 107, pp. 1815-31.
Piccinelli, M. and Wilkinson, G. (2000), “Gender Differences in Depression”,
British Journal of Psychiatry, vol. 177, p. 486-92.
Scitovsky, T. (1976), The Joyless Economy, New York: Oxford University Press.
Weiss, E.L., Longhurst, J.G., and Mazure, C.M. (1999), “Childhood Sexual Abuse
as a Risk Factor for Depression in Women: Psychosocial and Neurobiological
Correlates”, American Journal of Psychiatry, vol. 156, pp. 816-28.
Weissman, M.M., Bland, R.C., Canino, G.J., Faravelli,C., Greenwald, S., Hwu,
H.-G., Joyce, P.R., Karam, E.G., Lee, C.-K., Lellouch, J., Lepine, J.- P., Newman,
S.C., Rubio-Stipc, M., Wells, E.,Wickramaratne, P.J., Wittchen, H.-U., & Yeh, E.-K.
(1996), “Cross-national epidemiology of major depression and bipolar disorder”,
Journal of the American Medical Association, vol. 276 , pp. 293–299.
19 .
Table 1
Proportion of Men and Women who have been Depressed/Anxious, by type of
Condition
Percentage who for a period
Men (4,227)
Women (5,862)
All persons
lasting several days or longer:
(10,089)
Felt sad, empty, or depressed
35
44
40.5
Felt very discouraged about
38
44
41.3
how things were going in life
Lost interest in most things they
29
33
31.6
usually enjoyed like work,
hobbies, and personal
relationships
Mild Anxiety
32
40
36.7
Severe Anxiety
8
10
7.9
Notes to Table 1:
Proportions were computed for persons who reported non-missing values for all the variables
used in the logistic regressions (Table 4-8): 10,089 persons of whom 5,862 were women and
4,227 were men.
Depression is defined as a positive response to any of the following questions
(iv)
Have you ever in your life had a period, lasting several days or longer, when most
of the day you felt sad, empty, or depressed?
(v)
Have you ever in your life had a period, lasting several days or longer, when most
of the day you were very discouraged about how things were going in your life?
(vi)
Have you ever in your life had a period, lasting several days or longer, when you
lost interest in most things you usually enjoy like work, hobbies, and personal
relationships?
Mild anxiety is defined as answering yes to the following question: Have you ever in your life
had an attack of fear or panic when all of a sudden you felt very frightened, anxious, or
uneasy?
Severe anxiety is defined as answering yes to the following question: Have you ever had an
attack when all of a sudden you became very uncomfortable, you either became short of breath, dizzy,
nauseous, or your heart pounded, or you thought that you might lose control, die, or go crazy?
20 .
Table 2: Distribution of Depression and Anxiety by Attribute
Sad
Discouraged Loss of
Percentage with the condition
interest
Overall sample
41
41
32
Race, Ancestry
Asian
31
32
24
Black
40
41
33
Hispanic
46
44
33
White
50
55
41
Age
Age: 18-30
43
46
36
Age:31-45
41
42
31
Age: 46-60
42
42
33
Age >60
34
31
22
Marital Status
Married or cohabiting
35
36
27
Separated, divorced, widowed
48
46
37
Never married
44
46
37
Resources (average score = 3.57)
Income-to-needs-ratio: hhinc/poverty line > mean
39
39
29
Income-to-needs-ratio: hhinc/poverty line < mean
41
42
33
Years of Education
Education high
40
41
30
Education Medium
41
43
35
Education moderate
40
41
31
Education low
41
40
30
Employment status
Employed
39
40
30
Unemployed
47
50
40
Inactive
43
41
34
Immigration status
US born
43
46
36
Non-US born
37
35
26
US region
North East
44
45
36
Mid West
45
48
38
West
36
37
28
South
41
41
31
Health, Childhood, Society
Good physical health
37
38
28
Bad physical health
54
54
44
Childhood traumas
52
55
44
No childhood trauma
29
28
19
Church member
39
40
31
No Church membership
41
42
32
Disability Scores
WHO-DAS: cognitive score> mean =1.19
76
80
72
WHO-DAS: cognitive score< mean =1.19
37
38
28
WHO-DAS: social interaction score > mean =0.85
78
81
76
WHO-DAS: social interaction score < mean =0.85
38
39
29
21 .
Mild
anxiety
37
Severe
anxiety
8
30
35
41
45
5
9
8
12
37
37
39
31
8
7
8
11
34
40
39
7
10
7
36
37
7
9
37
38
37
35
6
8
8
9
36
40
39
7
10
10
39
34
10
6
42
41
35
34
9
10
6
8
34
47
47
27
36
37
7
14
12
5
8
8
69
34
68
35
28
7
24
7
Table 3: Distribution of Attributes between Women and Men
Women (58%)
Men (42%)
Percentage of women and men with attribute
Race, Ancestry
Asian
18
23
Black
49
40
Hispanic
25
28
White
8
9
Age
Age: 18-30
27
28
Age:31-45
36
35
Age: 46-60
23
23
Age >60
14
14
Marital Status
Married or cohabiting
45
59
Separated, divorced, widowed
28
16
Never married
27
25
Resources (average score)
Income-to-needs-ratio: hhinc/poverty line
3.16
4.15
Years of Education
Education high
21
24
Education Medium
25
23
Education moderate
29
29
Education low
25
24
Employment status
Employed
61
73
Unemployed
9
7
Inactive
30
20
Immigration status
US born
60
54
Non-US born
40
45
US region
North East
24
21
Mid West
9
8
West
24
30
South
43
41
Health, Childhood, Society
Good physical health
77
82
Childhood traumas
46
55
Church membership
36
23
Disability Scores (average)
WHO-DAS: cognitive score
1.37
0.96
WHO-DAS: social interaction score
0.94
0.61
Notes to Table 3
1. Asian: Vietnamese, Filipino, Chinese, all other Asian; Hispanic: Puerto Rican, Mexican, all other Hispanic;
Black: African American, Afro-Caribbean.
2. Education high: 16 or more years of education; Education medium: 13-15 years; Education moderate:
12 years; Education low: 0-11 years
3. Childhood traumas: yes to any of the following: fidgety childhood; frequently in trouble with adults for 6
months or more during childhood or adolescence; lying, stealing as child or teenager; run away frequently,
played truant, or stayed out late as child or teenager; separation anxiety for 1 month or more as child.
4. WHO-DAS: World Health Organisation Disability Assessment Score: higher the score, greater the
disability.
22 .
Table 4
Marginal Probabilities from “sad, empty, depressed” Logistic Equation
Entire Sample
Female Sample
Male Sample
(10,087)
(5,861)
(4,226)
Mg. prb
Z val
Mg. prb
Z val
Mg. prb
Z val
Sex
Male
Race, Ancestry (residual: White)
Asian
Black
Hispanic
Age (residual >60 years)
Age: 18-30
Age:31-45
Age: 46-60
Marital Status (residual: never married)
Married or cohabiting
Separated, divorced, widowed
Resources
Income-to-needs-ratio: hhinc/poverty line
Years of Education (residual, low education)
Education high
Education Medium
Education moderate
Employment status (residual: inactive)
Employed
Unemployed
Immigration status (residual: born abroad)
US born
US region (residual: South)
North East
Mid West
West
Health, Childhood, Society
Good physical health
Childhood traumas
Church membership
Disability Scores
WHO-DAS: cognitive score
WHO-DAS: social interaction score
-0.097
-8.98
-0.142
-0.122
-0.025
-5.8
-6.27
-1.08
-0.127
-0.088
0.014
-3.6
-3.29
0.43
-0.160
-0.161
-0.069
-4.93
-5.98
-2.2
0.083
0.089
0.095
3.72
4.38
4.6
0.079
0.073
0.083
2.73
2.75
3.13
0.092
0.113
0.111
2.63
3.57
3.42
-0.054
0.058
-3.81
3.32
-0.045
0.043
-2.37
1.97
-0.058
0.096
-2.72
3.24
0.002
1.35
0.001
0.6
0.003
1.25
0.087
0.035
0.022
4.81
2.16
1.45
0.083
0.031
0.022
3.48
1.43
1.12
0.087
0.037
0.020
3.23
1.49
0.88
-0.022
0.031
-1.5
1.39
0.005
0.050
0.26
1.78
-0.066
-0.003
-2.64
-0.09
0.001
0.1
0.013
0.74
-0.018
-0.91
0.017
0.011
-0.036
1.19
0.56
-2.13
0.017
-0.006
-0.024
0.89
-0.22
-1.04
0.017
0.038
-0.041
0.79
1.21
-1.76
-0.133
0.231
-0.026
-9.31
21.9
-1.81
-0.140
0.225
-0.039
-7.72
15.94
-2.07
-0.118
0.230
-0.002
-5.18
15.13
-0.1
0.013
0.011
7.7
5.16
0.015
0.012
6.51
4.12
0.009
0.010
4.15
3.04
Notes to Table 4
1. Depression is defined as answering yes to the following question: Have you ever in your life had a period,
lasting several days or longer, when most of the day you felt sad, empty, or depressed?
2. Asian: Vietnamese, Filipino, Chinese, all other Asian; Hispanic: Puerto Rican, Mexican, all other Hispanic;
Black: African American, Afro-Caribbean.
3. Education high: 16 or more years of education; Education medium: 13-15 years; Education moderate:
12 years; Education low: 0-11 years.
4. Childhood traumas: yes to any of the following: fidgety childhood; frequently in trouble with adults for 6
months or more during childhood or adolescence; lying, stealing as child or teenager; run away frequently,
platyed truant, or stayed out late as child or teenager; separation anxiety for 1 month or more as child.
5. WHO-DAS: World Health Organisation Disability Assessment Score: higher the score, greater the disability.
23 .
6. The rows of the table with typeface in italics are those variables for which the difference between the female
and male marginal probabilities was significantly different from zero at the 20% (or less) level of significance.
24 .
Table 5
Marginal Probabilities from “very discouraged” Logistic Equation
Entire Sample
Female Sample
(10,087)
(5,861)
Mg. prb
Z val
Mg. prb
Z val
Male Sample
(4,226)
Mg. prb
Z val
Sex
Male
-0.072
-6.50
Race, Ancestry (residual: White)
Asian
-0.180
-7.55
-0.186
-5.54
-0.179
Black
-0.156
-7.92
-0.117
-4.33
-0.197
Hispanic
-0.096
-4.30
-0.068
-2.17
-0.129
Age (residual >60 years)
Age: 18-30
0.137
6.01
0.121
4.05
0.163
Age:31-45
0.129
6.20
0.118
4.36
0.146
Age: 46-60
0.122
5.77
0.127
4.65
0.117
Marital Status (residual: never married)
Married or cohabiting
-0.044
-3.09
-0.052
-2.69
-0.030
Separated, divorced, widowed
0.044
2.49
0.028
1.27
0.078
Resources
Income-to-needs-ratio: hhinc/poverty line
0.002
1.00
0.004
1.53
0.000
Years of Education (residual, low education)
Education high
0.096
5.25
0.089
3.66
0.102
Education Medium
0.060
3.61
0.066
3.02
0.047
Education moderate
0.031
2.04
0.037
1.81
0.023
Employment status (residual: inactive)
Employed
-0.015
-1.00
0.000
-0.02
-0.045
Unemployed
0.050
2.19
0.063
2.22
0.025
Immigration status (residual: born abroad)
US born
0.034
2.56
0.048
2.67
0.014
US region (residual: South)
North East
0.035
2.45
0.033
1.73
0.040
Mid West
0.035
1.74
0.041
1.56
0.027
West
-0.012
-0.68
0.008
0.36
-0.029
Health, Childhood, Society
Good physical health
-0.135
-9.32
-0.141
-7.66
-0.122
Childhood traumas
0.249
23.72
0.246
17.48
0.250
Church membership
-0.024
-1.62
-0.060
-3.15
0.030
Disability Scores
WHO-DAS: cognitive score
0.017
8.84
0.019
7.22
0.015
WHO-DAS: social interaction score
0.010
4.51
0.011
3.7
0.009
Notes to Table 5
Depression is defined as answering yes to the following question: Have you ever in your life
had a period, lasting several days or longer, when most of the day you were very discouraged
about how things were going in your life?
25 .
-5.37
-7.15
-4.14
4.52
4.46
3.52
-1.4
2.61
-0.06
3.7
1.86
0.99
-1.76
0.68
0.71
1.81
0.85
-1.19
-5.21
16.24
1.32
5.16
2.52
Table 6
Marginal Probabilities from “loss of interest” Logistic Equation
Entire Sample
Female Sample
(10,087)
(5,861)
Mg. prb
Z val
Mg. prb
Z val
Male Sample
(4,226)
Mg. prb
Z val
Sex
Male
-0.033
-3.24
Race, Ancestry (residual: White)
Asian
-0.111
-5.18
-0.140
-4.86
-0.088
Black
-0.092
-5.19
-0.081
-3.3
-0.105
Hispanic
-0.065
-3.33
-0.054
-1.97
-0.082
Age (residual >60 years)
Age: 18-30
0.176
7.55
0.159
5.16
0.207
Age:31-45
0.146
7.03
0.141
5.12
0.159
Age: 46-60
0.154
7.08
0.152
5.34
0.163
Marital Status (residual: never married)
Married or cohabiting
-0.040
-3.07
-0.041
-2.33
-0.031
Separated, divorced, widowed
0.054
3.26
0.034
1.65
0.097
Resources
Income-to-needs-ratio: hhinc/poverty line
0.003
1.69
0.005
2.22
0.000
Years of Education (residual, low education)
Education high
0.079
4.46
0.063
2.67
0.093
Education Medium
0.072
4.54
0.063
2.99
0.080
Education moderate
0.023
1.59
0.008
0.43
0.040
Employment status (residual: inactive)
Employed
-0.059
-4.26
-0.041
-2.32
-0.091
Unemployed
0.007
0.34
0.000
-0.01
0.018
Immigration status (residual: born abroad)
US born
0.031
2.55
0.053
3.18
0.001
US region (residual: South)
North East
0.043
3.16
0.050
2.76
0.032
Mid West
0.039
2.06
0.021
0.85
0.069
West
-0.010
-0.64
0.001
0.04
-0.015
Health, Childhood, Society
Good physical health
-0.124
-8.81
-0.133
-7.33
-0.110
Childhood traumas
0.224
22.56
0.212
15.56
0.237
Church membership
-0.014
-1.04
-0.038
-2.19
0.024
Disability Scores
WHO-DAS: cognitive score
0.016
10.03
0.016
7.86
0.016
WHO-DAS: social interaction score
0.010
5.27
0.013
5.05
0.004
Notes to Table 6
Depression is defined as answering yes to the following question: Have you ever in your life
had a period, lasting several days or longer, when you lost interest in most things you usually
enjoy like work, hobbies, and personal relationships?
26 .
-2.80
-4.20
-2.99
5.64
4.91
4.75
-1.56
3.37
0.12
3.51
3.27
1.83
-3.79
0.54
0.06
1.59
2.26
-0.66
-4.92
16.74
1.16
6.20
1.80
Table 7
Marginal Probabilities from the “Mild Anxiety” Logistic Equation
Entire Sample
Female Sample
(10,087)
(5,861)
Mg. prb
Z val
Mg. prb
Z val
Male Sample
(4,226)
Mg. prb
Z val
Sex
Male
-0.089
-8.58
Race, Ancestry (residual: White)
Asian
-0.110
-4.70
-0.124
-3.78
-0.107
Black
-0.101
-5.44
-0.092
-3.6
-0.110
Hispanic
-0.016
-0.74
0.012
0.38
-0.053
Age (residual >60 years)
Age: 18-30
0.011
0.53
-0.017
-0.63
0.059
Age:31-45
0.033
1.70
0.035
1.36
0.036
Age: 46-60
0.056
2.86
0.048
1.86
0.071
Marital Status (residual: never married)
Married or cohabiting
-0.017
-1.27
-0.019
-1.02
-0.007
Separated, divorced, widowed
0.004
0.22
0.009
0.45
-0.011
Resources
Income-to-needs-ratio: hhinc/poverty line
0.001
0.80
0.001
0.54
0.001
Years of Education (residual, low education)
Education high
0.094
5.32
0.079
3.37
0.113
Education Medium
0.059
3.68
0.038
1.79
0.088
Education moderate
0.050
3.39
0.044
2.27
0.055
Employment status (residual: inactive)
Employed
-0.015
-1.10
-0.005
-0.28
-0.032
Unemployed
0.006
0.27
0.004
0.14
0.007
Immigration status (residual: born abroad)
US born
0.023
1.78
0.031
1.8
0.004
US region (residual: South)
North East
0.071
5.12
0.051
2.78
0.100
Mid West
0.044
2.25
0.023
0.9
0.078
West
0.025
1.50
0.031
1.36
0.025
Health, Childhood, Society
Good physical health
-0.086
-6.24
-0.050
-2.82
-0.141
Childhood traumas
0.198
19.31
0.188
13.42
0.210
Church membership
-0.005
-0.37
-0.009
-0.47
-0.005
Disability Scores
WHO-DAS: cognitive score
0.011
8.40
0.017
8.04
0.005
WHO-DAS: social interaction score
0.006
3.74
0.006
2.82
0.005
Notes to Table 7
Mild anxiety is defined as answering yes to the following question: Have you ever in your life
had an attack of fear or panic when all of a sudden you felt very frightened, anxious, or
uneasy?
27 .
-3.34
-4.2
-1.79
1.79
1.23
2.34
-0.36
-0.41
0.58
4.25
3.57
2.45
-1.37
0.22
0.21
4.62
2.51
1.06
-6.29
14.36
-0.21
3.03
2.48
Table 8
Marginal Probabilities from the “Severe Anxiety” Logistic Equation
Entire Sample
Female Sample
Male Sample
(10,087)
(5,861)
(4,226)
Mg. prb
Z val
Mg. prb
Z val
Mg. prb
Z val
Sex
Male
-0.010
-1.66
Race, Ancestry (residual: White)
Asian
-0.020
-1.54
-0.030
-1.65
-0.007
Black
-0.012
-1.19
-0.007
-0.45
-0.015
Hispanic
-0.010
-0.90
-0.014
-0.84
-0.004
Age (residual >60 years)
Age: 18-30
-0.020
-1.94
-0.003
-0.15
-0.034
Age:31-45
-0.022
-2.28
-0.014
-0.97
-0.027
Age: 46-60
-0.014
-1.60
0.002
0.13
-0.027
Marital Status (residual: never married)
Married or cohabiting
0.013
1.55
0.020
1.67
0.002
Separated, divorced, widowed
0.018
1.60
0.025
1.62
0.006
Resources
Income-to-needs-ratio: hhinc/poverty line
0.000
0.44
0.000
-0.16
0.001
Years of Education (residual, low education)
Education high
-0.010
-1.00
0.002
0.14
-0.023
Education Medium
0.005
0.61
0.010
0.73
-0.002
Education moderate
-0.006
-0.73
0.004
0.32
-0.017
Employment status (residual: inactive)
Employed
-0.011
-1.27
-0.004
-0.33
-0.020
Unemployed
0.010
0.74
0.016
0.87
0.001
Immigration status (residual: born abroad)
US born
0.026
3.42
0.019
1.74
0.031
US region (residual: South)
North East
0.013
1.47
0.011
0.87
0.014
Mid West
0.009
0.78
0.007
0.46
0.009
West
-0.005
-0.53
-0.005
-0.35
0.000
Health, Childhood, Society
Good physical health
-0.048
-4.76
-0.055
-3.98
-0.037
Childhood traumas
0.059
8.13
0.052
4.97
0.063
Church membership
-0.009
-1.21
-0.015
-1.44
0.000
Disability Scores
WHO-DAS: cognitive score
0.002
2.74
0.002
1.55
0.002
WHO-DAS: social interaction score
0.001
1.78
0.004
3.96
-0.003
Notes to Table 8
Severe anxiety is defined as answering yes to the following question: Have you ever had an
attack when all of a sudden you became very uncomfortable: you either became short of breath, dizzy,
nauseous, or your heart pounded, or you thought that you might lose control, die, or go crazy?
28 .
-0.41
-1.12
-0.28
-2.92
-2.23
-2.6
0.22
0.43
0.92
-2.13
-0.18
-1.85
-1.55
0.05
3.08
1.12
0.55
0.01
-2.68
6.59
0.04
3.37
-2.00
Table 9
The Decomposition of the Proportions of Men and Women who have experienced
Depression and Anxiety: all coefficients
1
2
3
4
5
6
7
Sad, empty,
0.4420.442–
0.4430.3640.4420.006
0.443depressed
0.353 =
0.443 =
0.353
0.353
0.364
0.360
0.089
-0.001
=0.09
=0.011
=0.078
=0.083
Discouraged
0.4390.4390.4420.3850.4390.005
0.4400.375
0.442 =
0.375
0.375
0.385
0.381
=0.064
-0.003
=0.067
=0.010
=0.054
=0.059
Loss of
0.3320.3320.3260.3100.3320.012
0.330interest
0.293
0.326 =
0.293
0.293
0.310
0.303
=0.039
0.006
=0.033
=0.017
=0.022
=0.027
Mild anxiety
0.3990.3990.4040.3150.399-0.013
0.4010.322
0.404 =
0.322
0.322 =
0.315
0.318
=0.077
-0.005
=0.082
-0.007
=0.084
=0.09
Severe
0.1030.1030.0970.0850.1030.004
0.100anxiety
0.083
0.097
0.083
0.083
0.085
0.084
=0.020
=0.006
=0.014
=0.002
=0.018
=0.016
Notes to Table 9:
Column 1: Observed difference. Difference between the proportions of women and men,
who have experienced the condition, P( X , βˆ )  P( X , βˆ )
W
W
M
M
Column 2: Attribute Difference. Difference between the proportion of women who have
experienced the condition and the proportion of men who would have experienced the
condition if male attributes had been evaluated using female coefficients,
P( XW , βˆ W )  P( XM , βˆ W ) .
Column 3: Coefficient difference. Difference between columns 1 and 2 (Residual),
{ P( X , βˆ )  P( X , βˆ ) }-{ P( X , βˆ )  P( X , βˆ ) }= P( X , βˆ )  P( X ,βˆ
W
W
M
M
W
W
M
W
M
W
M
M
).
Column 4: Attribute Difference. Difference between the proportion of women who would
have experienced the condition if female attributes had been evaluated using male
coefficients and the proportion of men who have experienced the
condition, P( X , βˆ )  P( X ,βˆ ) .
W
M
M
M
Column 5: Coefficient difference. Difference between columns 1 and 4 (Residual),
{ P( X , βˆ )  P( X , βˆ ) }-{ P( X , βˆ )  P( X ,βˆ ) }= P( X , βˆ )  P( X , βˆ
W
W
M
M
W
M
M
M
W
W
W
M
).
XW and XM are the attribute vectors, and  W and M are the coefficient vectors, for women and
men, respectively.
Column 6: Attribute difference obtained as the weighted average of columns 2 and 4, the
weights being the proportions of men (column 2) and women (column 4) in the sample.
Column 7: Coefficient difference obtained as the difference between men and women in their
“synthetic” probabilities of experiencing that condition.
29 .
Technical Appendix
The Decomposition of Probabilities
More formally, there are N people (indexed, i=1…N) of whom NM are men and NW
are women: k=M (men), W (women.). Define the variable Yi such that Yi=1, if the
person has had a condition (depression, anxiety), Yi=0, otherwise. Then, under a logit
model, the likelihood of a man or woman having had the condition is:
Pr(Yi  1) 
exp( Xik β k )
 F ( Xik βˆ k ), k  M , W
k k
1  exp( Xi β )
(1)
where: X ik   X ij , j  1...J  represents the vector of observations, for person i of group
k, on J variables which determine the likelihood of the person having that condition,
and βˆ k   jk , j  1...J  is the associated vector of coefficient estimates for persons
from group k.
The average probability of a man or woman having had the condition is:
Nk
Y k  P (Xik ,βˆ k )  N k 1  F ( Xik βˆ k ) k  M ,W
(2)
i 1
So that:
Y W  Y M  [ P(XiM ,βˆ W )  P(XiM ,βˆ M )]  [ P(XWi ,βˆ W )  P(XiM ,βˆ W )]
(3)
Alternatively:
Y W  Y M  [ P(XWi ,βˆ W )  P(XWi ,βˆ M )]  [ P(XWi ,βˆ M )  P(XiM ,βˆ M )]
(4)
The first term in square brackets, in equations (3) and (4), represents the
“response effect”: it is the difference in average rates (of having had a condition)
between women and men resulting from inter-gender differences in responses (as
exemplified by differences in the coefficient vectors) to a given vector of attribute
values.16 The second term in square brackets in equations (3) and (4) represents the
16
That is, from applying different coefficient vectors to a given vector of attributes
30 .
“attributes effect”: it is the difference in average rates (of having had a condition)
between women and men resulting from inter-gender differences in attributes, when
these attributes are evaluated using a common coefficient vector.
So for example, in equation (3), the difference in sample means is decomposed
by asking what the average rates for men would have been, had they been treated as
women; in equation (4), it is decomposed by asking what the average rate for women
would have been, had they been treated as men. In other words, the common
coefficient vector used in computing the attribute effect is, for equation (3), the
female vector and, for equation (4), the male vector.
31 .