Jurnal Health Behavior

JID: JEBO
ARTICLE IN PRESS [m3Gsc;December 10, 2018;13:45]

Journal of Economic Behavior and Organization xxx (xxxx) xxx
Contents lists available at ScienceDirect
Journal of Economic Behavior and Organization

journal homepage: www.elsevier.com/locate/jebo
An economic theory of depression and its impact on health

behavior and longevityR
Holger Strulik
University of Goettingen, Department of Economics, Platz der Goettinger Sieben 3, 37073 Goettingen, Germany
a r t i c l e i n f o a b s t r a c t
Article history: In this paper, I propose an economic theory of depression and its impact on health behav-
Received 11 September 2018 ior and longevity. Based on studies from happiness research, depression is conceptualized
Revised 31 October 2018
as a drastic loss of utility and value of life for unchanged fundamentals. The model is
Accepted 25 November 2018
used to explain how untreated depression leads to unhealthy behavior and adverse health
Available online xxx
outcomes: depressed individuals are predicted to save less, invest less in their health, con-
JEL classification: sume more unhealthy goods, and exercise less. As a result, they age faster and die earlier
D15 than non-depressed individuals. I calibrate the model for an average American and discus
D91 how depression enlarges the socioeconomic gradient of health and consider feedback ef-
I10 fects of depression on earnings and of physical exercise on depression as well as a variety
I12 of depression shocks. Delays in treatment for depression in young adulthood are predicted
to have significant repercussions on late-life health outcomes and longevity.
Keywords:
Depression © 2018 Elsevier B.V. All rights reserved.
Depression therapy
Health behavior
Aging
Longevity
The first thing that goes is happiness. You cannot gain pleasure from anything. That’s famously the cardinal symptom
of major depression.
(Solomon, 2015)
1. Introduction
In health economic theory, depression has received relatively little attention. This is perhaps surprising given the seri-
ousness of the disease. According to the WHO (2017), more than 300 million people worldwide are living with depression,
which makes depression the leading cause of ill health and disability. In the US, 6.7% of all adults above age 18 have had at
least one major depressive episode in the past year (NSDUH, 2016). Similar population shares are observed in other devel-
oped countries. In many developing countries, the prevalence is higher since depression is associated with poverty (Lorant
et al., 2003; Strulik, 2018b).
The most notable feature of depression is its strong influence on life satisfaction. According to the Diagnostic and Statisti-
cal Manual of Mental Disorders, the two main symptoms of depression are depressed mood and anhedonia, i.e. the reduced
ability to experience pleasure (American Psychiatric Association, 2013). Happiness studies document an independent influ-
R
I would like to thank Gustav Feichtinger, Sophia Kan, Michael Kuhn, Alexia Prskawetz, Miguel Sanchez-Romero, Aida Tanios, Timo Trimborn, and two
anonymous referees for helpful comments and discussion. Declarations of interest: none
E-mail address: holger.strulik@wiwi.uni-goettingen.de
https://doi.org/10.1016/j.jebo.2018.11.022
0167-2681/© 2018 Elsevier B.V. All rights reserved.
Please cite this article as: H. Strulik, An economic theory of depression and its impact on health behavior and longevity,
Journal of Economic Behavior and Organization, https://doi.org/10.1016/j.jebo.2018.11.022
JID: JEBO
2 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx
ence of depression on life satisfaction (Clark et al., 2008) and identify it as the single largest determinant of misery, i.e.
of being in the bottom quarter of the population in terms of life satisfaction (Layard et al., 2013). Building on this litera-
ture, I consider the empirical observation of life satisfaction from surveys as a reasonable approximation of lifetime utility
(Clark et al., 2017; Frey and Stutzer, 2002; Stevenson and Wolfers, 2008). As the World Hapiness Report (Layard et al., 2013;
Clark et al., 2017), I treat life satisfaction and happiness about life as a whole interchangeably. In the literature, the term
happiness is sometimes also used to describe the (affective) instantaneous component of subjective well-being while life
satisfaction is considered the more appropriate measure of its evaluative, long-term component (Deaton, 2008). Here, as in
any standard life cycle model, lifetime utility is conceptualized as the discounted lifetime sum over instantaneous utility
experienced at any instance. There is thus a unique and monotonous relation between the instantaneous and the evaluative
measure of happiness. Depression is modeled as a drastic and persistent drop in instantaneous utility, which – because of
its persistence – translates into a drastic drop of life satisfaction.
The exogenous drop of utility conceptualizes depression as “sadness without a cause” as suggested by philosophers
(Aristotle, 1984, p. 1501) and psychologist (Horwitz and Wakefield, 2007). This notion captures the finding from the World
Happiness Report (2017, Table 5.2) that depression and anxiety reduce life satisfaction controlling for (i.e. independently
from) income, education, physical health and other confounders (in the U.S. by 0.21 standard deviations). The available
economic theory focusses on the explained part of low utility, explained, for example, by low income, bad health, unem-
ployment etc. The theory of depression developed in this study, in contrast, focusses on a loss of life satisfaction that is not
explained by standard arguments of the utility function and which is, in this sense, exogenous.
The conceptualization of depression as unexplained sadness does not mean that the onset of depression needs to be
without a cause. Depression may have been triggered by a particular life event, as for example the loss of a job or a spouse.
Everyone experiences in the course of life the feeling of sadness as a transient state of mind, which is explained by particular
life events and circumstances. For depressed individuals, however, the sadness persists after the cause is gone or when non-
depressed individuals have adapted to the new life circumstances. As a result, sadness in depression is more severe and
lasts longer than sadness as a normal part of life (American Psychiatric Association, 2013).
In this study, we neglect any trigger of depression and consider the loss of utility as being fully exogenous because this
allows us to identify causal mechanisms. In particular, we neglect that depression may have been caused by deteriorating
health, a phenomenon, which in econometric analysis could seriously confound causal inferences. Since depression is treated
exogenously in the counterfactual analysis performed in this study, any confounding effects from endogeneity or reverse
causality (as for example effects of deteriorating health on depression) are shut down by design. This means that all elicited
changes in health behavior and health outcomes can be causally attributed to depression. Counterfactual analysis means
that we calibrate a rich life cycle model for a non-depressed average American and investigate how an exogenous drop in
life satisfaction changes the predicted health behavior and health outcomes.1
A potential response to (severe) depression is suicide. In the model developed below, suicide could be an optimal re-
sponse if the depression shock turns lifetime utility negative and death would be associated with zero utility. In order to
exclude this case, we assume that depression is not severe enough such that depressed individuals do not kill themselves,
i.e. that lifetime utility of the depressed individuals is reduced but still positive. This covers by far the majority of cases.
SAMHSA (2016c) estimates that, among adults who suffered from major depressive disorder in the past year, 4.3% attempted
suicide and (Bostwick and Pankratz, 20 0 0) estimate a rate of suicide among patients with depression between 2 and 9%.
Excluding suicide allows us to focus the study on how depression influences life cycle health behavior and through these
channels health outcomes and longevity.2
This paper proposes the first economic theory that establishes mechanisms through which depression affects health be-
havior and health outcomes (other than suicide). The life cycle model explains the impact of depression on health behavior
by treating aging and longevity as endogenous. Individuals desire a long life because the instantaneous utility function is
concave while lifetime utility is linear in instantaneous utility. In simple words, consuming this year and next year provides
more utility than consuming twice as much this year with subsequent death. In order to facilitate a long life, individu-
als are thus interested in behaving in a healthy way. The health-deficit model of Dalgaard and Strulik (2014) investigates
how decreasing marginal utility of non-depressed individuals explains the income gradient of longevity. Here, we adapt and
extend the methodology introduced by Dalgaard and Strulik (2014) in order to explain the behavioral responses to depres-
sion. Depression reduces lifetime utility (without a cause) such that depressed individuals have less incentive to behave in
a healthy way. Through this mechanism the model explains why depressed individuals invest less in their health, exercise
1
If depression would be endogenous (caused for example by deteriorating health) it could be hard or impossible to identify whether the observed
change in behavior is caused by depression or by the circumstances that also triggered depression. Since the depression is the only thing that changes in
the computational experiments, all elicited changes in behavior can be causally attributed to depression. While the utility drop from depression is treated
as exogenous from an economists perspective it may still be endogenous from the perspective of neuroscientists (e.g. Drevets et al., 2008 or biologists e.g.
Levinson, 2006). Here we follow standard economics and treat the brainchemistry determining our preferences as exogenous. For the theoretical model,
the drop in utility captures any depressive disorder. For the numerical calibration, we conceptualize it as major depressive disorder (ICD-code 296).
2
The economic theory of rational suicide based on Hamermesh and Soss (1974) largely neglects depression as a cause of suicide. Instead it focusses on
explained low utility (explained, for example, by low income, old age, or certain lifetime events). Cutler et al. (2001) briefly discuss depression as an acute
state of high pain in youth. But altogether the theoretical literature neglects the disease of depression, conceptualized as a drastic and sustained decline of
happiness without a cause. Naturally, the suicide literature does not consider repercussions of low utility on health behavior or on health outcomes (other
than suicide), which are the objects of investigation in the present study.
JID: JEBO
H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 3
less, consume more unhealthy goods, and save less for health expenditure in old age than otherwise identical non-depressed
individuals. As a result, their bodies age faster and their life ends prematurely. Depression, according to this modeling, does
not change the preferences of individuals it “only” reduces the pleasure from any utility-providing activity. With respect to
unhealthy consumption, for example, the model captures the notion that a high rate of smoking among depressed people
does not reflect a particular preference for nicotine of the depressed but their lower opportunity cost of unhealthy con-
sumption due to their reduced life satisfaction and reduced desire for a long life (Solomon, 2015).
To capture these health effects and health outcomes we need a life cycle model that treats deteriorating health and
death endogenously. An appropriate framework is the (Dalgaard and Strulik, 2014) model, which conceptualizes aging as the
accumulation of health deficits in the course of life. Death occurs when an upper bound of health deficits has been accumu-
lated. The model has a foundation in gerontology and predicts that the pace of aging increases as people get older. Through
health behavior, individuals can speed up or slow down the accumulation of deficits. The model is easily calibrated with
data because health deficits are measured by a straightforward metric, the health deficit index (or frailty index; Mitnitski
et al., 2005; Mitnitski et al., 2002a; Mitnitski et al., 2002b; Rockwood and Mitnitski, 2007; Harttgen et al., 2013). The health
deficit index computes, from a long list of potential health deficits, the relative number of deficits that a person has (see
Searle et al., 2008 for details on the construction of the index). It has been established as an accurate proxy of morbid-
ity and mortality and has been used in hundreds of studies in the medical science literature. Its use allows economists to
communicate with medical scientist using a common metric.3
The model prediction that depression is associated with less healthy lifestyles is supported by several studies. The rate
of daily cigarette use, for example, is almost twice as high among adults with major depression compared to the general
population (29.7% t vs. 16%; NSDUH, 2006). Depressed individuals are also more likely to drink heavily and to use other
addictive substances (NSDUH, 2006). Aside from increased alcohol and cigarette consumption, depression is also associated
with inactivity, eating and sleeping problems, and a lack of treatment adherence for medical problems (Schulz et al., 2002).
Consequently, many studies have documented an association between health and depression. One of the most robust find-
ings is the relation between depression and cardiovascular disease (Schulz et al., 2002). With regard to the measure of
health used in the present study, Lohman et al. (2015) document a positive association between depression and the health
deficit index in late life, and Rao et al. (2017) find that depression significantly predicts increased health deficits, disability,
and mortality among the elderly. A rare study that estimates the impact of depression on life expectancy is provided by
Zivin et al. (2012) who find that depressed male patients died about 4.8 years earlier than non-depressed patients (71.1 vs.
75.9 years).
Depression is experienced individually at different intensity and duration. While depression diagnosed as dysthymia is
characterized by a depressed mood for more than two years, major depressive disorder is experienced in shorter episodes
with a median duration of between 8 weeks (Andrews, 2001) and 30 weeks (Vos et al., 2004). In the US national comorbidity
survey, 15–54 year old individuals diagnosed with depression had experienced on average 11 prior periods of depression
(Kessler et al., 1997). The model developed below is calibrated with annual data and variations in exogenous shocks to utility
of shorter duration than one year are smoothed out by the calibration strategy. The calibrated utility drop from depression
is thus understood as the long-run average decline in life satisfaction. The benchmark calibration assumes that untreated
depression persists throughout life. This is a useful starting point of analysis since it allows a straightforward elaboration of
the channels through which depression effects health behavior and health outcomes. In Section 6 we investigate a variety
of depression shocks that occur at different ages and disappear and reoccur later in life. The main conclusion from these
computational experiments is that all qualitative results from the benchmark model are confirmed.
Depression can be treated with drugs or psychotherapy but the efficacy of treatment is imperfect and individual-specific.
For major depression, it has been estimated that pharmacological treatment as well as cognitive behavioral therapy avert
about 50% of the disease burden (Vos et al., 2004). The probability of relapse into depression, however, remains high unless
the treatment continues. Fournier et al. (2010) estimate that maintenance treatment reduces the risk of relapse by about
50%. In the model developed below, depression treatment is conceptualized as medical expenditure that dampens the sever-
ity of the drop in utility. The modeling of depression treatment implements, for the first time, the suggestion to integrate
the demand for antidepressants into consumer theory (Blanchflower and Oswald, 2016; Katolik and Oswald, 2017).
The loss of happiness is the most important but not the only symptom of depression (for the full list see American Psy-
chiatric Association, 2013. This rigorous application of Occam’s razor neglects other symptoms of depression and other po-
tential channels through which depression may affect economic behavior. There exist two complementing economic theories
of depression focussing on other channels. De Quidt and Haushofer (2016) conceptualize depression as downward shock in
the belief about the return to effort in a model where a decision-maker chooses labor effort, non-food consumption, food
consumption, and sleep. Strulik (2018b) focuses on maternal depression modeled in reduced-form as increased present-bias,
which leads to less child investments of depressed mothers.4 In an overlapping generations model it is shown that the
intergenerational transmission of depression through gene-environment interaction may generate a poverty trap. Conceptu-
ally, both studies differ from the present one by their assumption that depression changes individual preferences while here
3
Throughout this paper, aging is considered as a physiological phenomenon, defined as the intrinsic, cumulative, progressive, and deleterious loss of
function that eventually culminates in death (Arking, 2006). It should not be confused with chronological aging from one birthday to the next, which is, of
course, neither affected by depression nor by health behavior.
4
For an empirical study on the impact of depression on impatience, see Lerner et al. (2013).
JID: JEBO
we assume that preferences remain intact and focus on the life-satisfaction channel. Both studies also focus on different
outcomes. The present study is the first one that investigates how depression affects the development of health deficits,
aging, and longevity.
By considering the effects of reduced happiness and life satisfaction on health behavior the paper is also more broadly
related to studies that find an impact of happiness on economic behavior. Guven (2012) instruments happiness with un-
expected sunshine and finds that happier people save more, have a lower marginal propensity to consume, and expect a
longer life. Similarly, De Neve et al. (2013) find that high subjective wellbeing is associated with healthier eating, reduced
likelihood of smoking, more exercise, and a longer life. Surveys of the impact of happiness on decision making, health, and
longevity include Lyubomirsky et al. (2005), Isen (2008), and Diener and Chan (2011).5
The paper is organized as follows. In the next section, I set up a life-cycle model of depression, health behavior, health
outcomes, and longevity; and, using comparative statics, I derive analytical results for the use of antidepressants at the
intensive and extensive margins. The full dynamic model can only be analyzed numerically. In Section 3, I calibrate the
model for a reference American (a 20-year-old male in the year 2010) and in Section 4, I assess the quantitative impact
of depression and depression treatment on life-cycle behavior, health, and longevity. In Section 5, I consider applications
and extensions of the basic model with regard to depression and the income gradient of health and longevity, feedback
effects of depression on labor market outcomes (retirement and lifetime earnings), and the impact of delayed diagnosis
and treatment on health outcomes. In Section 6, I discuss the timing of depression shocks as well as (multiple) depression
periods separated by non-depressed periods. Section 7 concludes with suggestions for future research.
2. A theory of depression, health behavior, and health outcomes
2.1. Utility
Consider an individual who derives utility from the consumption of health-neutral goods c and unhealthy goods u and
whose utility declines in the number of health deficits D. The (sub-)utility function U(c, u, D) exhibits, as usual, declining
marginal utility from consumption, ∂ U/∂ j > 0, ∂ 2 U/∂ j2 < 0 for j = c, u. Health deficits reduce utility and, following Finkelstein
et al. (2013), we assume that health deficits reduce the marginal utility from consumption, ∂ 2 U/(∂ c∂ D) < 0. Additionally,
individuals may engage in physical exercise x to slow down the accumulation of health deficits. Physical exercise has a
utility cost (e.g. in terms of leisure), which increases in the number of health deficits present in a person. For simplicity, we
assume that the subutility function for the cost of physical exercise f(x, D) is separable from consumption utility and has
positive first derivatives and non-positive second derivatives.
As motivated in the Introduction, depression can be best conceptualized as a downward-shifter of the utility function.
The severity of the depression is measured by δ ≥ 0. In this paper, we focus on depression that is not severe enough such
that utility remains positive and depressed individuals are not induced to kill themselves. We investigate the effects of de-
pression on longevity through induced unhealthy behavior. When depression is treated, the negative utility shock becomes
less severe. A unit of treatment m is conceptualized as the use of antidepressants or therapy. The efficacy of the treatment
is denoted by η ≥ 0. Together, this means that the utility of depressed individuals is reduced by δ g(m, η) with ∂ g/∂ η < 0
and with declining returns of treatment intensity, ∂ g/∂ m < 0, ∂ 2 g/∂ m2 > 0. Moreover, we assume that the marginal return
of antidepressant expenditure declines more strongly when the treatment is effective, i.e. ∂ 2 g/∂ η∂ m > 0. Reasonable albeit
not decisive further assumptions are that g(0, η ) = 1 and lim gm→∞ = 0. The first assumption means that untreated indi-
viduals experience the full power of depression δ . The second assumption rules out that non-depressed individuals take
antidepressants to further boost their happiness. Altogether, this means that lifetime utility is given by
T
V = [U (D, c, u ) − f (x, D ) − δ g(m, η )]e−ρ t dt, (1)
0
in which t is age, T is the age at death, and ρ is the time preference rate. Here, we focus on a deterministic model in
which death occurs when an upper limit of health deficits D̄ has been accumulated. As shown in Strulik (2015a) and
Schünemann et al. (2017b), stochastic versions of the health deficit model in which the probability of death depends pos-
itively on health deficits add more realism at the price of higher complexity but provide little extra insight into life-cycle
choices and outcomes. Since the present model is already quite complex, for the sake of simplicity, we abstract from stochas-
tic death.6
Depression δ is treated as a parameter that may change occasionally and unexpectedly, which means that it is conceptu-
alized as a shock. At the onset of depression, δ changes from zero to a positive value and at the end of depression δ changes
from a positive value back to zero. In the benchmark calibration of the model we consider depression that sets in at the
beginning (at age 20) and stays until the end of life of the depressed individual. The duration of the depression shock thus
5
The model of depression focusses on the impact of reduced happiness and neglects any independent influence of positive affect on behavior.
6
We follow the standard assumption in the literature that lifetime utility is additive in (discounted) instantaneous utility. In a stochastic version of this
model, one could also discuss multiplicative utility and risk aversion with respect to the length of life as in Bommier (2006). We would expect that then
individuals would consume more in earlier periods of their life. In a deterministic framework, however, the utility function suggested by Bommier would
lead to additive utility (see Bommier, 2006, p. 1224, p. 1226) and would thus leave the qualitative results unchanged.
JID: JEBO
coincides with the lifespan of the individual, which is a convenient special case to elaborate the main mechanisms through
which depression affects health behavior. In Section 5, we consider depression shocks with different onsets and ends of
depression.
2.2. Health
The state of health is measured by a health deficit index D, defined as the share of health deficits present in a per-
son (from a long list of potential health deficits). As shown by Mitnitski et al. (2002a,b), Harttgen et al. (2013), and
Abeliansky and Strulik (2018), health deficits are accumulated in a quasi-exponential way with increasing age. As in
Dalgaard and Strulik (2014) and Schünemann et al. (2017a), we assume that the “natural” increase of health deficits can
be slowed down by health investment h and accelerated by unhealthy consumption u. Additionally, we assume that physical
exercise x slows down the accumulation of health deficits. In short, health deficits accumulate as
D˙ = μ[D − Ahγ + Buω − Exκ − a], (2)
in which μ is the “natural” force of aging. The parameters A and γ measure the efficacy of health technology, the parameters
B and ω measure the degree of unhealthy consumption, and the parameters E and κ measure the impact of physical exercise.
We assume decreasing returns of health investment and physical exercise, 0 < γ , κ < 1 while marginal health costs are
increasing in unhealthy consumption, ω > 1. The parameter a captures health deficit increase that is unrelated to aging (via
D) and health behavior (via h, u, and x). The parameter is unimportant for the qualitative results but useful to increase the
precision of the calibration.
2.3. Wealth
Individuals receive an income flow w, conceptualized as net wage income until retirement, and as pension income after-
wards. Income is used for consumption, saving at interest rate r, and health expenditure. In order to introduce a primitive
health care system, we assume that a share φ of health investments and a share φ m of depression treatments are paid out
of pocket. Let the prices of unhealthy consumption, health investment, and depression treatment be denoted by q, p, and
pm . Then, wealth k accumulates as
k˙ = w + rk − c − qu − pφ h − pm φm m. (3)
2.4. Solution
Individuals chose consumption (c and u), health expenditure h, physical exercise x, and depression treatment m in order
to maximize lifetime utility (1) subject to constraints (2) and (3), the initial states of health D(0) and wealth k(0), the
boundary conditions D(T ) = D̄ and k(T ) = k̄, and non-negativity constraints on all choice variables. However, it will turn out
that only for unhealthy consumption and depression treatment, the non-negativity constraint becomes occasionally binding
whereas the other choices are always positive. The associated current value Hamiltonian is given by
H = U (D, c, u ) − f (x, D ) − δ g(m, η ) + λD μ[D − Ahγ + Buω − Exκ − a] + λk [w + rk − c − qu − pφ h − pm φm m]. (4)
The first order conditions for a maximum are:
∂ U (D, c, u )
= λk (5)
∂c
∂ U (D, c, u )
≤ λk q − λD μBωuω−1 with = for u > 0 (6)
∂u
−λD μγ Ahγ −1 = λk φ p (7)
∂f
−λD μκ Exκ −1 = (8)
∂x
∂g
−δ ≤ λk φm pm with = for m > 0. (9)
∂m
I have aligned the first order conditions such that the left-hand side always shows the marginal benefit and the right-hand
side shows the marginal cost. Condition (5) requires that the marginal utility from consumption of health neutral goods
equals the marginal cost in terms of foregone future utility, measured by the shadow price of capital. Condition (6) requires
that the marginal utility from unhealthy consumption is not larger than the marginal cost, which consists of implied forgone
future utility λk q and the marginal impact of unhealthy consumption on health μBωuω−1 , evaluated by the shadow price of
JID: JEBO
health deficits. Notice that future health deficits are bad such that λD is negative. Condition (7) requires that the marginal
return from health investment μγ Ahγ −1 evaluated with the shadow price of health deficits equals the marginal cost in
terms of foregone marginal utility from consumption. Condition (8) requires that the marginal return from physical exercise
equals the marginal utility cost.
Finally, condition (9) requires that the marginal return of depression treatment is not larger than the marginal cost of
treatment. To observe this, recall that treatment reduces depression, i.e. that ∂ g/∂ m < 0. For the treated, marginal benefits
equal marginal costs. When marginal costs exceed marginal benefits, individuals remain untreated. Inserting (5) into (9), we
obtain the depression treatment constraint:
∂ U (D, c, u ) ∂ g(m, η )
0≤G≡ φm p m + δ . (10)
∂c ∂m
From the inspection of (10), we obtain a first result:
Proposition 1 (Intensive Margin). The treatment intensity of depression m is increasing in the level of consumption c (and thus
income) of the patient, and in the number of health deficits of the patient. It declines in the price pm of the treatment and the
out-of-pocket share φ m .
For the proof, I apply the implicit

function
theorem on (10) evaluated with equality. For example, we have for price pm ,
dm/d pm = −(φm ∂ U (D, c, u )/∂ c )/ δ∂ 2 g/∂ m2 < 0, and likewise for φ m . The comparative static for health deficits uses the
assumption ∂ 2 U/(∂ c∂ D) < 0 and the comparative static for consumption uses the assumption of declining marginal utility
from consumption ∂ 2 U/(∂ c2 ) < 0. The intuition for the result with respect to consumption (income) is that individuals who
experience little extra utility from spending an extra dollar on consumption are more inclined to spend income on depres-
sion treatment. The marginal utility from consumption is low when the level of consumption is high or when many health
deficits are present. As long as consumption goods are normal (which we assume), richer individuals will consume more. In
other words, poor individuals have other pressing needs and are therefore less inclined to spend much money on depression
treatment, in particular, if treatment prices are high and the out-of-pocket share is large (i.e. if they are uninsured).
Proposition 2 (Extensive Margin). Depression remains untreated if the treatment is sufficiently inefficient (η is low), if the out-
of-pocket share φ m is sufficiently large, or if the level of consumption (income) of the patient is sufficiently low.
The proof with respect to η shows that G is increasing in η since ∂ 2 g/∂ m∂ η > 0, which makes it less likely that
(10) binds with equality when η is low. The other comparative statics are obtained similarly. These comparative statics,
however, do not provide any information about the impact of depression on health behavior and longevity. For this, we
need the full solution of the model, which also requires the costate equations and boundary conditions to be fulfilled. The
costate equations are given by
λk r = λk ρ − λ˙ k (11)
∂ U (c, u, D ) ∂ f (x, D )
− + λD μ = λD ρ − λ˙ D . (12)
∂D ∂D
Aside from the boundary conditions on k and D, the optimal solution of the free terminal time problem also fulfills H (T ) = 0.
Inspection of (4) shows that the Hamiltonian assumes a large value, ceteris paribus, when utility is large, i.e. for individuals
who are healthy, wealthy, and not depressed. The fact that instantaneous utility is strictly concave in the level of consump-
tion while lifetime utility is linear in instantaneous utility motivates the pursuit for a long life. Individuals, unconditionally
prefer to extend their consumption over a longer time period against consuming more right now. The pursuit for a long life,
however, comes also at a cost in terms of health investments, physical exercise, and the eschewal of unhealthy consumption.
Rich individuals experience lower marginal utility from (even more) consumption and have more funds to finance a long life.
They are thus predicted to live longer. This mechanism was the main subject of investigation in Dalgaard and Strulik (2014).
The novelty here is that depression drives down the level of instantaneous utility (experienced happiness). This means that
it reduces the desire for a long life, which in turn reduces the incentive to invest time and money in health and increases
the incentive to indulge in unhealthy consumption.
In order to explore the impact of depression on health and health behavior we need the full solution of the model, i.e.
the dynamic life cycle trajectories implied by the first order conditions and boundary conditions. This can be achieved only
numerically and the first step is to assume functional forms for the sub-utility functions. For simplicity, we conceptualize
total consumption c˜ as a convex combination of the consumption of health neutral goods and unhealthy goods, c˜ = θ c + (1 −
θ )u, 0 < θ < 1. One advantage of such a simple additive sub-utility function is that it allows for a preemptively high price
at which households abstain from unhealthy consumption (see below). We assume that utility from total consumption is
isoelastic. Controlling for health deficits, utility is given by (c˜1−σ − 1 )/(1 − σ ), in which 1/σ is the elasticity of intertemporal
substitution. Following Finkelstein et al. (2013), we treat the state of health as a shifter of the utility function of consumption
such that both utility and marginal utility of consumption are negatively affected by bad health. Specifically, we assume
that health adjusted utility is given by (D/D )− · (c˜1−σ − 1 )/(1 − σ ). The parameter controls the amount by which an
additional health deficit will shift the utility function downwards. The normalizing constant D > 0 is the measure with which
individuals compare their health deficits. For the calibration, we will set D to the health deficit index at age 20.
JID: JEBO
The disutility from physical exercise is formalized as f (x, D ) = (D/D )χ τ x, with τ > 0 and χ > 1. In order to obtain a
closed form solution, we assume that utility is linear in the intensity of physical exercise x. Since χ > 1, physical exercise
becomes increasingly painful with the accumulation of health deficits. The parameter τ controls how much an individual,
in general, dislikes physical exercise.
Finally, we assume a simple exponential function for depression treatment g(m, η ) = e−ηm . This function fulfills the as-
sumption of decreasing returns. It implies that non-depressed individuals will not demand antidepressants and that non-
treated individuals experience the full power of depression. For m > 1/η it fulfils ∂ 2 /(∂ m∂ η) > 0. For the calibrated model it
will be verified that this condition is fulfilled such that Proposition 2 holds. Altogether, lifetime utility (1) is parameterized
as
∞ D − θ c + ( 1 − θ ) u 1 − σ − 1 D χ
[ ]
V = − τ x − δ e−ηm dt. (13)
0 D 1−σ D
From the first order conditions, we obtain closed form solutions for u, x, and m (see the Appendix for details on the deriva-
tion of these results). The solution for unhealthy consumption is displayed in (14).
⎧
⎨ 1 −θ Aγ hγ −1 ω1−1
θ −q Bωφ p
for q < 1−θ θ
u= (14)
⎩
0 otherwise.
For unhealthy consumption to prevail, the relative utility weight of unhealthy goods, (1 − θ )/θ , has to exceed the price q. If
unhealthy consumption exists, then condition (14) predicts that its extent is large if the resulting health damage is low (B
is low), if medical efficiency in repairing damage is large (A is large), or if the price of health goods p is low.
The solution for physical exercise is displayed in (15).
−( +χ ) −σ 1
D (c˜) θ φ pκ E 1−κ
x= . (15)
D τ μAγ hγ −1
The intensity of physical exercise is high for healthy persons (with low D) and when the health gain E is large. It is low
when τ is large, i.e. when exercising provides much disutility. Physical exercise is also high when the marginal return of
health investments γ Ahγ −1 /p is low. This outcome is intuitive because it makes sense to exercise more in order to reduce
the accumulation of health deficits when the return to further monetary health investments is low.
The solution for depression treatment is displayed in (16). The comparative statics stated in general form in
Proposition 1 and 2 are easily verified by inspection of (16).

1 δη (D/D ) c˜σ
m = max 0, log . (16)
η p m φm θ
Finally, the optimal solution is characterized by two dynamic equations for consumption and health expenditure (see
Appendix for details on the derivation):

c˜˙ 1 D˙
= r−ρ − , (17)
c˜ σ D
D χ + μAγ hγ −1 c˜σ
h˙ 1 c˜1−σ − 1
= r−μ− + χτx . (18)
h 1−γ 1−σ D pφ D
Eq. (17) shows the standard Euler equation for consumption adjusted by health deficits, Intuitively, the accumulation of
health deficits drives down the incentive to spend on consumption in old age, and thus lowers the slope of the lifetime
consumption path (which may become hump-shaped; see Strulik, 2015b). Eq. (18) shows the health-Euler equation. When
health does not show up in the utility function ( = τ = 0), health expenditure becomes steeper with rising return on wealth
and declining force of mortality (i.e. with greater incentive to save for late-life health expenditure). The presence of health
deficits in the utility function decreases the slope of the health expenditure path through its impact on the utility from
consumption and on the disutility from physical exercise.
3. Calibration
The model is calibrated, as in Schünemann et al. (2017b), for an average white 20-year-old American male in the year
2010. The calibration matches life cycle data on initial health deficits, health deficits at death, age at death, and life-cycle
health investments, unhealthy behavior (conceptualized as smoking), and physical exercise. From Mitnitski et al. (2002a),
I take the estimate for the rate of aging, μ = 0.043. I set r to 0.07 according to the long-run real interest rate (Jorda
et al., 2017) and ρ = r as in Dalgaard and Strulik (2014) and Schünemann et al. (2017b). In the year 2010, the aver-
age life expectancy of a 20-year-old American male was 57.1 years, i.e the expected age at death was 77.1 (NVSS, 2014).
From Mitnitski et al. (2002a), I infer terminal health deficits D̄ = 0.106 at the terminal age 77.1 and initial health deficits
JID: JEBO
D(0 ) = 0.0273 at initial age 20.7 Without loss of generality, the normalizing constant D is set to 0.0273 as well, implying
that at age 20 health has no impact on utility.
When the individual is between 20 and 65 years old, I set w = 27, 928, which is the average labor income for single
men in the year 2010 (BLS, 2011). For older individuals, I set w = 0.45 · 27, 928 using an average replacement rate of pre-
retirement income of 0.45 from OECD (2016). In order to assure that the savings motive is confined to that of health and
consumption expenditure, I assume that the initial and final capital stock are zero. I set the out-of-pocket expenditure share
to φ = 0.28 for all ages. This proxy somewhat overestimates the out-of-pocket share of the elderly and underestimates it
somewhat for the young (Machlin and Carper, 2014). For the benchmark run, I set φm = φ , assuming no discrimination of
payments for depression treatment. Furthermore, I normalize prices p = q = 1. This is an interesting benchmark case because
it eliminates any price channel through which poor individuals may have an incentive to consume more unhealthy goods or
spend less on general health care.
The calibration of the effect of health on utility is related to Finkelstein et al. (2013) who estimate that a one-standard-
deviation increase in the number of chronic diseases is associated with a 11% decline in the marginal utility of consumption
(with a 95% confidence band from 2.7% to 16.8%). Finkelstein et al. (2013) focus on individuals above 50 and apply a smaller
set of more severe health deficits compared to this study. Here, we calibrate the frailty index of Mitnitski et al. (2002a),
which also contains relatively mild health deficits like farsightedness and incontinence. We thus consider, for the benchmark
case, a smaller impact of health deficits by setting = 0.05. This means that an unexpected increase in health deficits from D
by one standard deviation reduces the marginal utility from consumption by 3%, a value at the lower bound of the estimate
in Finkelstein et al. (2013).
Since most of the available empirical literature on consumption of unhealthy goods is on cigarettes and tobacco, u is
conceptualized as smoking. On average, single male Americans spent $ 364 on cigarettes in the year 2010 (BLS, 2012). The
smoking intensity declines with age by about factor 2 from age 25 to age 50. Since many individuals quit smoking as they
get older, smoking prevalence declines as well, by about factor 2.5 between age 25 and age 50. I try to capture the combined
effect of declining intensity and declining prevalence for the calibrated average American whose smoking intensity thus
declines by factor 5 (i.e. by 80%) from age 25 to age 50. Empirically, there is some variation in the estimates of years of life
lost due to smoking, ranging from 2.5 years (Preston et al., 2010) to 10 years (Jha et al., 2013). Here, I match an intermediate
value of 4.3 years, estimated by a study that takes selection into smoking into account, see Darden et al. (2015).
To measure physical exercise, I use the metric of metabolic equivalents (METs) defined as the energy cost of a given
physical activity divided by energy expenditure at rest. This metric allows for the aggregation of different physical activities
like walking, playing sports, gardening, etc., and to compare them across individuals and ages. The average American in
Moore et al.’s (2012) sample spends about 1.14 MET per day (8 MET per week) on physical exercise, an equivalent of about
23 min of brisk walking per day. Moore et al. estimate that this exercise increases life expectancy by about 3 years. They
also document strongly decreasing returns from physical exercise, i.e. large gains for departure from inactivity and very
small gains for excessive exercise. Studies from the UK (Townsend et al., 2015) and Canada (Statistics Canada, 2011) suggest
that the intensity of exercise declines by 50% from age 35 to age 70. Assuming that British and Canadian men are, in this
regard, sufficiently similar to Americans, I try to match their age gradient of physical activity.
To summarize, the remaining parameters of the model, that is A, a, B, E, γ , χ , θ , σ , τ , and ω, are estimated to fit
the following stylized facts: (i) the model predicts the actual accumulation of health deficits over a lifetime (as estimated
by Mitnitski et al., 2002a; (ii) death occurs at the moment when D̄ health deficits have been accumulated at an age of
77.1 years; (iii-iv) health expenditure matches health care expenditure of American men in 2010 at the age of 35 and 70
(MEPS, 2010); (v-vi) the Reference American spends on average $ 364 per year on smoking and expenditure declines by 80%
between ages 25 and 50; (vii) that consumption of the unhealthy good costs about 4.3 years of life; (viii-ix) the individual
spends x = 1.14 METs per day on exercise, which allows him to live 3 years longer than he would in total inactivity; (x)
physical exercise declines by factor 50% between age 35 and 70. This leads to the estimates A = 0.0 0 013, a = 0.0175, B =
9.5 · 10−7 , E = 1.06 · 10−3 , γ = 0.39, χ = 1.55, κ = 0.1, θ = 0.203, σ = 0.90, τ = 0.0028, and ω = 1.20. While most of these
parameters are latent, there exist many studies on the size of σ . The estimated value accords well with studies suggesting
that the intertemporal elasticity of substitution is close to unity (Chetty, 2006).
As another plausibility check of the calibration, I compute the value of life (VOL) of the Reference American and compare
it with previous estimates. The VOL provides a monetary expression of aggregate utility experienced during life until its
end, that is, utility is converted by the unit value of an “util”, u (c). The VOL at the initial age is obtained by applying the
formula V OL = V/[∂ u(c (0 ), u(0 ), D )/∂ c (0 )]. The benchmark calibration predicts a VOL of about $ 4.7 million at age 20. In
terms of order of magnitude, this value corresponds well to (Murphy and Topel, 2006, Fig. 3) estimate of a VOL of about $
6.5 million for American men at age 20.
For the quantitative part of the paper we conceptualize depression as major depressive disorder (ICD-codes 296). Most
studies on the toll of depression on longevity compute mortality hazard rates for small samples because computation of
effects on life expectancy needs large sample sizes. Such a large-sample study is provided by Zivin et al. (2012) who use
data from the U.S. department of Veteran Affairs and the National Death Index and find that depressed individuals died, on
7
Mitnitski et al. estimate health deficit accumulation for Canadian men. Deficit accumulation within the USA and Canada appears to be similar enough
to justify it as a good approximation for the U.S. (Rockwood and Mitnitski, 2007).
JID: JEBO
average, at age 71.1 whereas average non-depressed individuals died at age 75.9. While the sample is not representative,
the leading causes of death were similar to those of the general U.S. population. It is also useful for the calibration that the
life expectancy of non-depressed individuals in the sample is close to the life expectancy of the Reference American. I thus
model depression such that it reduces the length of life by 1 − 71.1/75.9 ≈ 6%, implying an estimated age at death of 72.4
years for a depressed individual.
In the benchmark run, I assume that depression is untreated, present from the start (i.e. from age 20), and permanent.
According to American Psychiatric Association (2013) the incidence of (first appearance) major depressive disorder peaks
in the 20s such that the assumption of onset at age 20 is reasonable. As discussed in the Introduction, major depressive
disorder is experienced in episodes with a median duration of between 8 and 30 weeks. Most individuals experience several
recurring depressive episodes, potentially throughout their live. The short-run episodic nature of depression is smoothed out
by calibrating the model to annual data. The depression shock δ is thus conceptualized as the average annual loss in utility
due to depression. For the benchmark case, I obtain the estimate δ = 2.73 for the average severity of depression. A sensitivity
analysis considers different intensities of the depression shock and Section 6 considers alternative onsets and progressions
of depression.
Notice that this step of the calibration assumed that depression remained suboptimally untreated, i.e. the first order
condition (9) is not fulfilled. The most intuitive explanation for this assumption is that depression remained undiagnosed.
For example, it has been estimated that more than 50% of depression cases remain unrecognized by primary care physi-
cians (Lecrubier, 2007). This calibration strategy allows for a comparison of the same depressed person with and without
treatment.
The literature has not yet converged on a general view about the efficacy of depression treatment and its impact on
health and longevity. As a benchmark, I begin with a case where treatment is quite effective such that it reduces the impact
of depression on life expectancy by two thirds from 4.6 years to 1.5 years. This modeling of efficacy relates to the estimate
by Vos et al. (2004) that maintained drug treatment and cognitive behavioral therapy avert about 50% of the health effects
from depression (measured in disability adjusted life years) when the adherence rate is 60%. In order to pin down pm , I take
into account that, in the year 2010, expenditure for mental health contributes to about 6.1% of all health care expenditure
(SAMHSA, 2016b), that 6.7% of Americans suffered from a major depressive episode in the past year, and that one third of
this population remained untreated (SAMHSA, 2016a). Let the average expenditure for health care and mental health over
the life cycle be denoted by h̄ and m̄. I then back out pm by solving 0.061 = (0.067 · 2/3 pm m̄ )/(0.067 · 2/3 pm m̄ + ph̄ ). This
provides the estimates pm = 0.45 and η = 0.0 0 038. I provide a sensitivity analysis with respect to the efficacy of treatment.
4. Results: depression, life-cycle health behavior, and longevity
The model is solved by applying the relaxation algorithm of Trimborn et al. (2008). The predicted life cycle choices of the
non-depressed benchmark American are shown in Fig. 1 by blue (solid) lines. The targeted data points are shown by circles.
The model predicts that with advancing age, the individual develops more health deficits, spends more on health, consumes
less unhealthy goods, and exercises less. At age 77.1, the individual dies when the terminal condition for health deficits D̄
is reached. The life-cycle choices of non-depressed individuals are discussed in detail by Dalgaard and Strulik (2014) and
Schünemann et al. (2017b). The new choice variable here is physical exercise, which declines with age since exercise be-
comes increasingly painful as the individual develops more health deficits.
The life-cycle choices of the depressed individual are shown by red (dashed) lines in Fig. 1. The relative change of behav-
ior and age at death, compared to the non-depressed benchmark, are shown as case 1 in Table 1. The depressed individual
is predicted to spend much more on unhealthy consumption. Averaged over the lifetime, unhealthy consumption increases
by 68%, compared to the benchmark case. The depressed individual invests less in health at all ages (25% less over the
lifetime). The depressed individual also exercises less intensely (21% lower intensity of physical exercise). The fact that the
depressed individual cares less about the future is also expressed in less savings. The average wealth (k) accumulated over
the lifetime declines by 22%. The depressed individual has a relatively low regard for a long life because depression reduces
the value of life quite dramatically. Compared to the benchmark, the VOL declines by 20%, implying that the absence of
depression contributes 20% to life satisfaction. This prediction accords with the estimate that mental illness explains 21% of
the variation in adult life satisfaction in the US, see Clark et al. (2017).
The prediction that depression motivates more unhealthy consumption, less savings, less health care, and less physical
exercise are in line with the empirical observations cited in the Introduction. In order to elicit these behavioral changes, we
need not to appeal on preference changes of depressed individuals. Instead, the reason for the more unhealthy behavior is
the depression itself, which reduces the value of life and makes a longer life less desirable. Notice that a life-cycle model
with endogenous health and longevity is necessary in order to obtain these results. A model that treats life length as given
(and perhaps infinitely long) would require the manipulation of the utility functions U(c, u, D) and f(x, D) in order to elicit
these behavioral changes. Here, the introduction of an exogenous drop in utility and life satisfaction is sufficient to explain
the change in behavior.
The life cycle choices of the treated depressed individual are shown by green (dash-dotted) lines in Fig. 1 and summa-
rized as case 2 in Table 1. By assumption the treatment is quite effective. The lifetime trajectory for physical exercise moves
close to the one of the non-depressed individual (average exercise is only 2% below benchmark). There is also a pronounced
increase in health expenditure (to 7% below non-depressed expenditure). The treatment almost eliminates the entire impact
JID: JEBO
Fig. 1. Depression, Health Behavior, and Health Outcomes.
Table 1
Comparative dynamics: sensitivity analysis and model extensions.
Case Parameter change Remark T u/u h/h x/x k/k

1 δ = 2.73 Benchmark depression −4.6 0.68 −0.25 −0.21 −0.22
2 η = 0.0 0 038 Benchmark treatment −1.5 0.18 −0.07 −0.02 −0.05
3 η = 0.0 0 015 Lower efficacy −3.0 0.39 −0.13 −0.04 −0.08
4 φm = 0 . 1 Lower out of pocket share −0.6 0.07 −0.03 −0.01 −0.04
5 φm = 0 . 5 Higher out of pocket share −2.4 0.30 −0.11 −0.03 −0.07
6 φm = φ = 0 . 5 Higher out of pocket share −4.2 0.21 −0.66 −0.03 −0.25
7 δ = 4.1 More severe depression −7.6 −1.29 −0.36 −0.32 −0.36
8 η = 0.0 0 038 ... & benchmark treatment −1.6 0.20 −0.08 −0.02 −0.06
9 ν=1 Less labor supply −4.7 0.70 −0.25 −0.21 −0.06
10 ν=3 Even less supply −5.0 0.76 −0.26 −0.22 0.23
11 ν = 1, η = 0.0 0 038 Less supply & treatment −1.5 0.18 −0.07 −0.02 −0.05
12 ν = 3, η = 0.0 0 038 Less supply & treatment −1.5 0.18 −0.07 −0.02 −0.05
13 Treatment delay 10 years −2.9 0.30 −0.07 −0.07 −0.13
14 Treatment delay 20 years −3.8 0.46 −0.07 −0.11 −0.19
15 w = 1.5wbench Higher wage −3.1 0.87 −0.27 −0.19 −0.15
16 w = 2wbench Even higher wage −2.7 0.81 −0.21 −0.17 −0.14
17 w = 2wbench , η = 0.0 0 038 High wage & treatment −0.7 0.17 −0.06 −0.01 −0.03
18 as 17 and φm = 0.5 High out of pocket share −1.0 0.28 −0.09 −0.01 −0.04
19 = 0.53 Feedback from exercise −3.3 0.46 −0.19 −0.16 −0.16
20 = 1.1 More feedback −1.9 0.26 −0.12 −0.10 −0.11
21 τ = 0.014 Less sports aversion −1.6 0.20 −0.10 −0.08 −0.06
of depression on happiness. The value of life of the treated individual is 1% below the benchmark (compared to 20% of
the non-treated). The treatment cured the low valuation of a long life and raised the need for savings to finance increased
consumption and health care in extended old age (savings is now only 5% below benchmark).
Case 3 in Table 1 considers outcomes when depression treatment is less effective by setting η = 0.0 0 015. As a result,
treatment is not able to sufficiently restore the value of life. Depressed individuals still exhibit less healthy behavior than
the non-depressed benchmark, and treatment saves only 1.6 years of longevity; T is reduced by 3.0 years instead of 4.6
when depression remained untreated. Compared to no treatment, treatment improves health outcomes by 1/3 (instead of
2/3 as for case 2).
Case 4 and 5 consider different out-of-pocket shares for depression treatment (φ m is 0.28 in benchmark run). A larger
out-of-pocket share reduces the budget that remains for other expenditures. In particular, general health investment declines
JID: JEBO
with increasing out-of-pocket share for depression treatment. This increases the marginal productivity of health care and
induces more unhealthy consumption (see (14)). Physical exercise, in contrast, is almost restored to the non-depressed level.
This indicates that depression is still (almost) cured, as in case 3. The difference in health deficit accumulation and age of
death between these cases operates through the budget constraint. A higher out-of-pocket share induces more savings for
depression treatment in old age, which, among other things, leaves less income to be spent on current health investments.
Case 6 considers a simultaneous increase in the out-of-pocket share for both types of health expenditure. As expected, this
particularly harms investments in general health care.
Case 7 considers a more severe depression: δ increases to 4.1, which means that, if untreated, depression can be at-
tributed to a loss of 7.6 years of lifetime and a loss of 30% of VOL. Interestingly, when treated, the health outcomes and
behavioral changes are almost the same as in the case of the less severe depression, as a comparison between cases 8 and
2 shows.
5. Applications and extensions
5.1. Depression and labor supply
So far, we have ignored the additional indirect effects of depression on life-cycle outcomes through reduced productivity
and labor income. The recent study of Peng et al. (2016) suggest that the effects on income may indeed be small. Depression
is estimated to reduce the probability of employment by 2.6 percentage points and to increase annual work loss days by
1.4 days. Earlier studies suggested higher effects on labor market outcomes (e.g. Chatterji et al., 2011). In the model, a
convenient way to acknowledge labor market effects is through early retirement. Subsequently, we assume that the previous
age of retirement (R = 65) is reduced to R − νδ e−mη . In cases 9 and 10 of Table 1, we consider relatively strong effects of
depression on retirement and lifetime income by setting ν = 1 and ν = 3. This means that, in case 9, the individual receives
2.73 times (1–0.45) times w, i.e. $ 41,900 less lifetime income. In case 10, it implies that depression reduces lifetime income
by $125,800.8
Comparing results of case 9 and 10 with case 1 shows that outcomes for untreated depression are, perhaps surprisingly,
only marginally affected by the income feedback although the impact on income is large. In other words, the direct depres-
sion effect on life satisfaction is much stronger than any indirect effect through income, as suggested in related empirical
studies (Layard et al., 2013). Health behavior, in turn, responds more sensitively to depression than to income. Case 11 and
12 further confirm this assessment by showing that the treatment outcomes in case of feedback effects also deviate only
insignificantly from the benchmark treatment (Case 2).
5.2. Delayed treatment
As discussed in the Introduction, a large share of depression cases remains untreated. One reason for a delay in treat-
ment could be that individuals and their relatives fail to self-diagnose the disease and perhaps rationalize it as (justified)
sadness “with a reason”. In Table 1, case 13, we see the health outcomes when depression is treated with a delay of 10
years. The individual loses 2.9 years of lifetime, compared to 1.5 year if depression is treated immediately. The life-cycle
behavior shown in Fig. 2 explains why. Blue (solid) lines reiterate the life-cycle trajectories of the benchmark non-depressed
individual. Red (dashed) lines show the trajectories when treatment is delayed by 10 years. They coincide with those of
the untreated individual from Fig. 1 until treatment sets in. Then, the treated depressed individual behaves in a similarly
healthy manner as the non-depressed individual: he invests more in health, consumes less unhealthy goods, and engages
in more physical exercise. But it is too late to fully compensate for the faster deteriorating health during early adulthood.
Health deficits continue to accumulate faster after treatment and lead to an earlier death.
Case 14 in Table 1 and the green (dash-dotted) trajectories in Fig. 2 show the outcome when treatment is delayed by
20 years. Aggregate behavioral outcomes are now more similar to completely untreated depression than to instantly treated
depression. Significantly increased health expenditure after treatment and reduced unhealthy consumption fail to compen-
sate for the lost health during the depression phase. As a result, the loss of lifetime T is now closer to the non-treated
individual (case 1) than to the immediately treated individual (case 2). These outcomes are a manifestation of the quasi-
exponential nature of health deficit accumulation, which leads to the amplification of early-life health shocks over time.
The health capital model (Grossman, 1972), in contrast, would predict that initial differences in health shocks – or, in the
current case, the consequences of delayed treatment – depreciate as the individual ages (see Almond and Currie, 2011 and
Dalgaard et al., 2017 for a critique). The health deficit model captures the cumulative character of health deficit accumula-
tion emphasized in gerontology (e.g. Arking, 2006). The damage done in young adulthood by unhealthy consumption, little
health investment, and little physical exercise is never fully repaired after the onset of depression treatment. These outcomes
highlight the importance of diagnosing and treating the disease early.
8
These labor income effects are chosen deliberately to be higher than suggested in the cited empirical studies in order to ensure that the effect is not
underestimated. The real effect is likely to be lower than predicted by the model. According to the setup of the model, the permanent income hypothesis
holds. In this framework, it does not affect the results whether the income loss originates from reduced labor supply at the intensive margin (e.g. through
unpaid sick leave during the day or week) or at the extensive margin (through early retirement).
JID: JEBO
Fig. 2. Delayed treatment.
5.3. Income gradient
We next investigate the income gradient of depression. As shown in Dalgaard and Strulik (2014) and Strulik (2018a),
the health deficit model predicts a strong income gradient of health and longevity. The reason is that marginal utility from
consumption is relatively low for rich individuals, which results in strongly decreasing returns of income on utility, as em-
phasized by many happiness studies (e.g. Layard et al., 2013; Lecrubier, 2007). Thus, compared to poor individuals, rich
individuals desire a long life relatively more than higher instantaneous consumption. This induces them to invest more in
health and behave more healthfully.
Here, we see that the income gradient becomes even larger for depressed individuals. Case 15 considers the difference
in behavior and health outcomes between depressed and non-depressed individuals when both earn 50% more than in
the benchmark case. We see that, within the richer income group, depression leads to a greater increase in unhealthy
consumption whereas behavioral changes for health investment and physical exercise are about the same as within the
average income group (i.e. as for case 2). To understand why the impact of depression on longevity is lower for the rich
despite the relatively stronger response of unhealthy consumption, it is useful to recall that the non-depressed rich behave
more healthfully and spend less on unhealthy goods. For example, a 100% increase of smoking from 1 to 2 cigarettes per
week will likely lead to less health repercussions than a 50% increase from 10 to 15 cigarettes per day.
Case 16 considers the impact of depression among individuals who are 100% richer than the benchmark American and
shows that this further reduces the impact of depression on longevity. Summarizing, the health deficit model predicts that
depression has less severe effects on health and longevity for the rich and that the income gradient of health is steeper for
depressed than for non-depressed individuals.
Case 17 shows that treatment is more effective for the rich in terms of restoring health. The loss in longevity with
treatment is 0.7 years (compared to 1.5 years in the benchmark case 3). In terms of life years saved, however, treatment is
more effective for the poor. In the benchmark case (2 and 3) treatment saves 3.1 years of live while for the rich (case 16
and 17), treatment saves 2 years. Comparison of case 18 and 5 shows that the rich are less harmed by a high out of pocket
share of depression treatment.
5.4. Feedback effects from exercise
The benchmark model focusses on the impact of depression on health behavior. For physical exercise, however, there also
exists evidence for reverse causality: physical exercise reduces depression symptoms similarly to drugs and psychotherapy
(for a survey, see Silveira et al., 2013. In order to capture the feedback effect in an algebraically convenient way, we consider
JID: JEBO
Fig. 3. Depression shocks.
the following refinement of the treatment function to g(m, η, x ) = e−ηm − x, where ≥ 0 measures the impact of physical
exercise on reduced depression.
In order to quantify , I use the meta-study of Silveira et al. (2013), which finds that on average, physical exercise re-
duces the intensity of depression by 0.61 standard deviations. The intensity of depression is about exponentially distributed
in the U.S. population (Tomitaka et al., 2018), which implies that its mean is equal to its standard deviation. For the cali-
bration, this feature suggests a standard deviation of the intensity of depression of 2.73 since δ = 2.73 is the intensity of
depression of the average depressed American. The estimate of Silveira et al. (2013) implies that depression is reduced with
exercise by 0.61 · 2.73 = 1.66 units such that δ x̄ = 1.66. Inserting average exercise of the benchmark American (x̄ = 1.14)
and solving for provides the estimate = 0.53.
Results are shown as case 19–21 in Table 1. For the calibrated value of (case 19), we observe that depression reduces
physical exercise by less than in the benchmark case (case 1). Moreover, unhealthy consumption, health investments, and
savings, are also less strongly affected by depression compared to the benchmark case. Exercise works similar to a depres-
sion therapy of relatively low efficacy. By imposing a higher benefit from exercise, as in case 20, the predicted impact of
depression becomes larger. Moreover, for individuals with relative little aversion to physical exercise (for whom τ = 0.0014
instead of 0.0028), physical exercise could be as effective as therapy, as a comparison of case 21 and case 2 shows.
6. Depression shocks and depression periods
6.1. Depression shocks
So far, we considered that the 20 years old benchmark American is either “born” with depression or not. In order to
investigate the timing of depression, we next consider cases where depression occurs as an unexpected shock later in life.9
In Fig. 3, blue (solid) lines show the life cycle trajectory for the benchmark non-depressed individual, red (dashed) lines
show behavior when depression hits at age 40 and green (dash-dotted) lines show behavior when depression hits at age 60.
The severity of depression is always the same as for the benchmark run (δ = 2.73). Since depression comes unexpected, life
cycle behavior coincides in the three cases until the depression shock occurs. After the shock, affected individuals experience
decreased happiness and a reduced value of life and respond by desiring a long life less, i.e. by investing less in health, by
consuming more unhealthy goods, and by exercising less.
9
The shocked life cycle trajectories are obtained from matching the trajectories obtained for δ = 0 until the age at which the shock occurs with the
respective trajectories for δ > 0 from the onset of the shock until death. At the age of the shock, the final values of the state variables of the first trajectories
become the initial values of the state variables of the second trajectories (transversality condition).
JID: JEBO
Table 2
Depression shocks and depression periods.
Case Timing T u/u h/h x/x k/k

1 Benchmark −4.6 0.68 −0.25 −0.21 −0.22
2 & treatment −1.5 0.18 −0.07 −0.02 −0.05
3 δ –onset at 40 −1.3 0.29 −0.24 −0.12 −0.07
4 & treatment −0.3 0.05 −0.07 −0.01 −0.03
5 δ –onset at 60 −0.3 0.05 −0.17 −0.04 −0.02
6 & treatment −0.1 0.00 −0.06 0.00 −0.02
7 δ –period age 40–50 −0.8 0.17 −0.02 −0.04 −0.03
8 & treatment −0.2 0.04 −0.01 0.00 −0.01
9 δ –period age 30–40 −1.5 0.23 −0.03 −0.05 −0.06
10 & treatment −0.5 0.06 −0.01 −0.01 −0.01
11 δ –period age 30–50 −2.3 0.40 −0.03 −0.09 −0.08
12 & treatment −0.6 0.10 −0.01 −0.01 0.01
13 2 δ –periods 30–40 and 50–60 −1.9 0.31 −0.06 −0.08 −0.08
14 & treatment −0.6 0.08 −0.01 −0.01 −0.01
15 2 δ –periods 40–50 and 60–70 1.0 0.20 −0.10 −0.07 −0.04
16 & treatment −0.2 0.04 −0.03 −0.01 −0.01
17 2 δ –periods 30–45 and 50–65 −2.4 0.43 −0.11 −0.12 −0.09
18 & treatment −0.7 0.11 −0.02 −0.01 −0.01
The response to a late-life depression shock thus elicits qualitatively the same responses as permanent depression. The
quantitative effect of depression on health and health behavior is smaller when the depression hits later. The reason is that,
at the time of the depression shock, depressed individuals inherit the health and wealth of an until this day non-depressed,
healthier individual and thus devalue their value of life by relatively less. This is perhaps most evident for exercise. A forty
years old individual with a 20 year history of depression, experiences a high utility cost of exercise because of the relatively
bad state of health at age 40. The health deficit index at age 40 is high because the individual has already lived 20 years with
depression and thus with less health investments, less exercise, and more unhealthy consumption than a non-depressed
individual. For this reason, he exercises less than a forty years old who is freshly hit by depression of the same severity and
equipped with a healthier body (less health deficits) inherited from the past.
These results are summarized in the first part of Table 2. Case 1 and 2 reiterate the results from the benchmark model
(case 1 and 2 from Table 1). Case 3 shows results for the depression shock at 40. The elicited behavioral changes are
qualitatively the same as for benchmark depression but they are attenuated. This is in particular true for savings and exercise
since most of lifetime exercise and savings have been done before age 40. For the same reason, the impact on health
investments is almost the same as for the early shock since most health investments are made after age 40. Case 5 shows
that this reasoning applies also when depression hits even later in life, at age 60. Case 4 and 6 show that treatment effects
are also similar to the benchmark case.
6.2. Depression periods
We next consider depression periods and begin with a one-time depression shock that appears at some age and then
disappears again at some later age. As argued above, the short-run episodic nature of depression is “hidden” in the calibra-
tion of the model with annual data, i.e. δ is the average annual decrease of life satisfaction through depression. We thus
consider the remission and relapse of depression in the long run. Depression episodes of shorter duration would display a
similar pattern, albeit at a finer time scale.
In Fig. 4, blue (solid) lines reiterate the predicted lifetime trajectories of the non-depressed benchmark American. Red
(dashed) lines show the trajectories when depression occurs from age 40 to 50 (with benchmark severity, δ = 2.73). Before
depression hits the trajectories coincide. During the depression period the individual invests less in health, consumes more
unhealthy goods, exercises less, and saves less. After the depression period, the individual resumes “normal” behavior. In fact,
he invests somewhat more in health and consumes somewhat less unhealthy goods than a permanently non-depressed in-
dividual. Exercise, however, falls somewhat short of that of a non-depressed individual because the additional health deficits
accumulated during the depression period make exercising harder.
Lifetime outcomes, relative to the non-depressed individual, are summarized as case 7 in Table 2. Again, behavioral
changes and outcomes are qualitatively the same as for permanent depression and are quantitatively attenuated. Case 9
shows that the change in outcomes becomes larger when the depression period (of the same duration) is experienced earlier
in life, from age 30 to 40. Case 11 shows that, naturally, the effects from depression are stronger when the depression period
lasts longer. The effects of treatment (at benchmark efficacy, η = 0.0 0 038) are also analogous to the benchmark case.
Finally we consider multiple depression periods. In Fig. 5, blue (solid) lines again reiterate the lifetime trajectories of the
benchmark non-depressed individual. Red (dashed) lines show lifetime behavior when the individual experiences depression
(of benchmark severity) from age 30 to 40 and from age 50 to 60. By now, it comes perhaps as no surprise that, in periods
without depression, the individual behaves similar to the permanently non-depressed individual and that, in depression
JID: JEBO
Fig. 4. Depression Periods.
Fig. 5. Two depression periods.
periods, he behaves similar to a permanently depressed individual. It is interesting to see that, during non-depressed spells,
the individual behaves somewhat more healthfully in terms of health investment and unhealthy consumption than the
permanently non-depressed individual but that this “overshooting” behavior is not sufficiently strong to make up for the
loss of health during the depression period.
JID: JEBO
The lifetime outcomes are summarized as case 13 in Table 2. The lifetime responses are smaller in magnitude than those
for permanent depression but otherwise similar. The remaining cases in Table 2 show that lifetime health behavior and
health outcomes are less strongly affected if multiple depression periods happen later in life and more strongly affected if
the depression periods last longer.
7. Conclusion
This paper introduced depression into health economic theory and evaluated its impact on life-cycle health behavior and
longevity. The model captures one major symptom of depression, the reduced ability to experience pleasure. Inspired by
happiness research, depression has been conceptualized as a large drop in instantaneous utility and life satisfaction (lifetime
utility). Successful treatment partly restores original utility and life satisfaction. These elements have been integrated into a
life-cycle model of health deficit accumulation, augmented by utility-enhancing unhealthy consumption and utility-reducing
but health-enhancing physical exercise.
The model explains why, as a result of reduced life satisfaction, depressed people consume more unhealthy goods, save
less, invest less in their health, and exercise less. The reason for this predicted behavior is that health, aging, and longevity
are endogenous in the life cycle model and depressed individuals, because of the reduced life satisfaction, have less incentive
to perform healthy activities in order to prolong their life. The model generates these behavioral responses to depression
without interfering with individual preferences, i.e. without assuming that, for example, depressed people prefer unhealthy
goods more strongly. Instead, it is shown that it is sufficient for the predicted outcomes to assume that the depressed
experience less pleasure from everything, which is modeled as a sustained drop of utility for unchanged fundamentals.
The model has been calibrated to predict the life-cycle trajectories of health behavior and health outcomes for a non-
depressed Reference American. Introducing depression as a 20% drop in life satisfaction rationalizes changes in health be-
havior that reduce the length of life by about 4 years. Extensions of the model have shown that these results are obtained
largely independently from the size of (reasonable) feedback effects of depression on labor supply and lifetime income. The
timing of treatment, however, is predicted to be crucial. Delays in diagnosis or treatment lead to severe losses in health that
are not fully recovered in the treatment period. The reason for this result is that the (gerontologically founded) health deficit
model amplifies the late-life consequences of health shocks in early life. In the present context, this means that individuals
do not manage to compensate fully for the health destroyed by unhealthy behavior during untreated depression in early
adulthood.
Since this study is a first attempt to discuss depression in the context of health and longevity, there are ample possibil-
ities to further extend and refine the theory. For example, it could also be taken into account that depression has a direct
biological impact on health through, for example, blood pressure, inflammation, or immune function. Because the sustained
drop of utility captures in a general way the major symptom of depression (in contrast to the manipulation of specific pref-
erences), the theory can easily be extended in various directions to investigate other potential behavioral changes caused by
depression. It could thus inspire further research in health economic theory to better understand the mechanisms behind
behavioral changes caused by depression. For example, future applications could investigate inferior and potentially health-
damaging self-treatment of depression (by excessive alcohol consumption or illicit drug intake). This could be a particularly
interesting endeavor in the context of the income gradient of longevity and limited access to insurance-covering clinical
depression treatment. Such an analysis could also consider the addictive potential of the applied self-treatment, investigate
the repercussions of unhealthy addiction on longevity (Strulik, 2018c), and enhance our understanding of the association
between depression and addiction (Solomon, 2015).
As explained in the Introduction, the assumed exogeneity of depression is a reasonable first approximation of reality
and a useful device in order to identify causality. An extension, however, could discuss depression triggered by bad health
and investigate the interdependence of both phenomena. In particular, chronic pain seems to be associated with depression
(DePaulo and Horvitz, 2002). By integrating this feature, the proposed model of depression could establish a link between
depression, use of antidepressants, and painkiller consumption, and contribute to a better understanding of the opioid epi-
demic.
Appendix A
A1. Derivation of (14) to (18)
We first write the first order conditions (5)–(9) and the costate Eqs. (11) and (12) for the parameterized utility function.
D −
−σ
[θ c + ( 1 − θ ) u ] θ = λk (A.1)
D
D −
−σ
[θ c + ( 1 − θ ) u ] (1 − θ ) ≤ λk q − λD μBωuω−1 with = for u > 0 (A.2)
D
− λD μγ Ahγ −1 = λk φ p (A.3)
JID: JEBO
D χ
− λD μκ Exκ −1 = τ (A.4)
D
− δηe−ηm ≤ λk φm pm with = for m > 0. (A.5)
λr = λk ρ − λ˙ k (A.6)
D − −1 θ c + (1 − θ )u 1−σ − 1 D χ −1
[ ]
− −χ τ x + μλD = λD ρ λ˙ D (A.7)
D 1−σ D
Let c˜ ≡ θ c + (1 − θ )u define a weighted measure of total consumption. From (A.1) and (A.2) we obtain
1−θ
λk − qλk + λD μBωuω−1 ≤ 0.
θ
Using (A.3) to eliminate λD and and dividing by λk we obtain
1−θ φp
−q≤ Bωuω−1 .
θ γ Ahγ −1
Solving for u provides (14) in the text. Inserting (A.3) into (A.4) provides:
λk φ p χ
κ Ex κ −1 = D τ.
γ Ahγ −1 D
Inserting λk from (A.1) we get
D − ( + χ ) φp
θ c˜−σ κ E = τ x1−κ .
D γ Ahγ −1
Solving for x provides (15) in the text. Inserting (A.1) into (A.5) we obtain
δη (c˜σ (D/D )
≤ eη m .
φm p m
Solving for m provides (16) in the text. Log-differentiating (A.1) provides
D˙ c˜˙ λ˙
− −σ = k.
D c˜ λk
Eliminating λ˙ k /λk using (A.6) provides (17) in the text. Log-differentiating (A.3) and eliminating λ˙ k /λk using (A.6) provides

h˙ 1 λ˙
= ρ −r− D . (A.8)
h 1−γ λD
Using (A.3), Eq. (A.7) can be written as
− −1 1−σ D χ −1 μγ Ahγ −1
λ˙ D D c˜ −1
− =μ−ρ − +χ τx .
λD D 1−σ D λk φ p
Using (A.1) this can be expressed as
1 −σ D χ + μγ Ahγ −1 c˜σ
λ˙ D c˜ −1
− =μ−ρ − +χ τx . (A.9)
λD 1−σ D λk φ pD
Inserting (A.9) into (A.8) provides (18) in the text.
References
American Psychiatric Association, 2013. Diagnostic and Statistical Manual of Mental Disorders, fifth ed. American Psychiatric Publishing, Arlington.
Abeliansky, A., Strulik, H., 2018. How we fall apart: similarities of human aging in 10 European countries. Demography 55 (1), 341–359.
Almond, D., Currie, J., 2011. Killing me softly: the fetal origins hypothesis. J. Econ. Persp. 25 (3), 153–172.
Andrews, G., 2001. Should depression be managed as a chronic disease? Br. Med. J. 322 (7283), 419–421.
Aristotle, 1984. The Complete Works of Aristotle: The Revised Oxford Translation. Princeton University Press, Princeton.
Arking, R., 2006. The Biology of Aging: Observations and Principles. Oxford University Press, Oxford.
Blanchflower, D.G., Oswald, A.J., 2016. Antidepressants and age: a new form of evidence for u-shaped well-being through life. J. Econ. Behav. Organ. 127,
46–58.
JID: JEBO
BLS, 2011. State Average Annual Pay for 20 0 0 and 2001. United States Department of Labor, Bureau of Labor Statistics. http://www.bls.gov/news.release/
annpay.t01.htm.
BLS, 2012. Consumer Expenditure Survey. U.S. Bureau of Labor Statistics. September 2012 http://www.bls.gov/cex/csxcross.htm#y1011.
Bommier, A., 2006. Uncertain lifetime and intertemporal choice: risk aversion as a rationale for time discounting. Int. Econ. Rev. 47 (4), 1223–1246.
Bostwick, J.M., Pankratz, V.S., 20 0 0. Affective disorders and suicide risk: a reexamination. Am. J. Psychiatry 157 (12), 1925–1932.
Canada, S. (2011). Physical Activity of Canadian Adults: Accelerometer Results From the 2007 to 2009 Canadian Health Measures Survey. http://www5.
statcan.gc.ca/access_acces/alternative_alternatif.action?l=eng&loc=11396-eng.pdf.
Chatterji, P., Alegria, M., Takeuchi, D., 2011. Psychiatric disorders and labor market outcomes: evidence from the national comorbidity survey-replication. J.
Health. Econ. 30 (5), 858–868.
Chetty, R., 2006. A new method of estimating risk aversion. Am. Econ. Rev. 96, 1821–1834.
Clark, A.E., Fleche, S., Layard, R., Powdthavee, N., Ward, G., 2017. The Key Determinants of Happiness and Misery. World Happiness Report 2017, New York:
Sustainable Development Solutions Network, United Nations.
Clark, A.E., Frijters, P., Shields, M.A., 2008. Relative income, happiness, and utility: an explanation for the easterlin paradox and other puzzles. J. Econ. Lit.
46, 95–144.
Cutler, D.M., Glaeser, E.L., Norberg, K.E., 2001. Explaining the rise in youth suicide. In: Risky Behavior Among Youths: An Economic Analysis. University of
Chicago Press, Chicago, pp. 219–270.
Dalgaard, C.-J., Hansen, C.W., Strulik, H., 2017. Accounting for Fetal Origins: Health Capital vs. Health Deficits. Discussion Paper, University of Copenhagen.
Dalgaard, C.-J., Strulik, H., 2014. Optimal aging and death: understanding the preston curve. J. Eur. Econ. Assoc. 12, 672–701.
Darden, M., Gilleskie, D.B., Strumpf, K., 2015. Smoking and Mortality: New Evidence From a Long Panel. Discussion Paper.
De Neve, J.-E., Diener, E., Tay, L., Xuereb, C., 2013. The objective benefits of subjective well-being. In: Helliwell, J.F., Layard, R., Sachs, J. (Eds.), World
Happiness Report 2013, vol. 2. UN Sustainable Network Development Solutions Network, New York, pp. 54–79.
De Quidt, J., Haushofer, J., 2016. Depression for Economists. National Bureau of Economic Research Working Paper, W22973.
Deaton, A., 2008. Income, health, and well-being around the world: evidence from the gallup world poll. J. Econ. Persp. 22 (2), 53–72.
DePaulo Jr., J.R., Horvitz, L.A., 2002. Understanding depression: what we know and what you can do about it. John Wiley & Sons, Hoboken, New Jersey.
Diener, E., Chan, M.Y., 2011. Happy people live longer: subjective well-being contributes to health and longevity. Appl. Psychol. 3 (1), 1–43.
Drevets, W.C., Price, J.L., Furey, M.L., 2008. Brain structural and functional abnormalities in mood disorders: implications for neurocircuitry models of
depression. Brain Struct. Funct. 213 (1–2), 93–118.
Finkelstein, A., Luttmer, E.F., Notowidigdo, M.J., 2013. What good is wealth without health? The effect of health on the marginal utility of consumption. J.
Eur. Econ. Assoc. 11 (1), 221–258.
Fournier, J.C., DeRubeis, R.J., Hollon, S.D., Dimidjian, S., Amsterdam, J.D., Shelton, R.C., Fawcett, J., 2010. Antidepressant drug effects and depression severity:
a patient-level meta-analysis. JAMA 303 (1), 47–53.
Frey, B.S., Stutzer, A., 2002. What can economists learn from happiness research? J. Econ. Lit. 40, 402–435.
Grossman, M., 1972. On the concept of health capital and the demand for health. J. Polit. Econ. 80, 223–255.
Guven, C., 2012. Reversing the question: does happiness affect consumption and savings behavior? J. Econ. Psychol. 33 (4), 701–717.
Hamermesh, D.S., Soss, N.M., 1974. An economic theory of suicide. J. Polit. Econ. 82 (1), 83–98.
Harttgen, K., Chatterji, S., Kowal, P., Strulik, H., Vollmer, S., 2013. Patterns in disability and frailty in older adults: Results from the Survey of Health. Ageing
and Retirement in Europe (SHARE), PlosOne 8 (10).
Horwitz, A.V., Wakefield, J.C., 2007. The Loss of Sadness: How Psychiatry Transformed Normal Sorrow into Depressive Disorder. Oxford University Press.
Isen, A.M., 2008. Some ways in which positive affect influences decision making and problem solving. Handbook of Emotions. The Guilford Press, New York.
548–573
Jha, P., Ramasundarahettige, C., Landsman, V., Rostron, B., Thun, M., Anderson, R.N., Peto, R., 2013. 21st-century hazards of smoking and benefits of cessation
in the United States. N. Engl. J. Med. 368 (4), 341–350.
Jorda, O., Knoll, K., Kuvshinov, D., Schularick, M., Taylor, A.M., 2017. The Rate of Return on Everything, 18702015. National Bureau of Economic Research
Working Paper w24112.
Katolik, A., Oswald, A.J., 2017. Antidepressants for Economists and Business-School Researchers: An Introduction and Review. Discussion Paper, University
of Warwick.
Kessler, R.C., Zhao, S., Blazer, D.G., Swartz, M., 1997. Prevalence, correlates, and course of minor depression and major depression in the national comorbidity
survey. J. Affect. Disord. 45 (1–2), 19–30.
Layard, R., Chisholm, D., Patel, V., Saxena, S., 2013. Mental illness and Unhappiness. World Happiness Report 2013, New York: Sustainable Development
Solutions Network, United Nations.
Lecrubier, Y., 2007. Widespread underrecognition and undertreatment of anxiety and mood disorders: results from 3 European studies. J. Clin. Psychiatry
68, 36–41.
Lerner, J.S., Li, Y., Weber, E.U., 2013. The financial costs of sadness. Psychol. Sci. 24 (1), 72–79.
Levinson, D.F., 2006. The genetics of depression: a review. Biol. Psychiatry 60 (2), 84–92.
Lohman, M., Dumenci, L., Mezuk, B., 2015. Depression and frailty in late life: evidence for a common vulnerability. J. Gerontol. Ser. B 71 (4), 630–640.
Lorant, V., Deliege, D., Eaton, W., Robert, A., Philippot, P., Ansseau, M., 2003. Socioeconomic inequalities in depression: a meta-analysis. Am. J. Epidemiol.
157 (2), 98–112.
Lyubomirsky, S., King, L., Diener, E., 2005. The benefits of frequent positive affect: does happiness lead to success? Psychol. Bull. 131 (6), 803–855.
Machlin, S., Carper, K., 2014. Out-of-pocket health care expenses by age and insurance coverage. Statistical Brief 441, Agency for Healthcare Research and
Quality. Rockville, MD.
Mitnitski, A., Song, X., Skoog, I., Broe, G.A., Cox, J.L., Grunfeld, E., Rockwood, K., 2005. Relative fitness and frailty of elderly men and women in developed
countries and their relationship with mortality. J. Am. Geriatr. Soc. 53, 2184–2189.
Mitnitski, A.B., Mogilner, A.J., MacKnight, C., Rockwood, K., 2002a. The accumulation of deficits with age and possible invariants of aging. Sci. World 2,
1816–1822.
Mitnitski, A.B., Mogilner, A.J., MacKnight, C., Rockwood, K., 2002b. The mortality rate as a function of accumulated deficits in a frailty index. Mech. Ageing
Dev. 123, 1457–1460.
Moore, S.C., Patel, A.V., Matthews, C.E., de Gonzalez, A.B., Park, Y., Katki, H.A., Thun, M., 2012. Leisure time physical activity of moderate to vigorous intensity
and mortality: a large pooled cohort analysis. PLoS Medicine 9 (11), e1001335.
Murphy, K.M., Topel, R.H., 2006. The value of health and longevity. J. Polit. Econ. 114, 871–904.
NSDUH, 2006. Results from the 2006 national survey on drug use and health: national findings. Substance Abuse and Mental Health Services Administra-
tion. NSDUH Series H-32, DHHS Publication No. SMA 07–4293. Rockville, MD.
NSDUH, 2016. Key substance use and mental health indicators in the United States: results from the 2016 national survey on drug use and health. HHS
Publication No. SMA 17–5044, NSDUH Series H-52, Rockville, MD.
NVSS, 2014. National Vital Statistics Report 63(7). US Department of Health and Human Services, National Vital Statistics System.
OECD, 2016. Gross Pension Replacement Rates. OECD Data. https://data.oecd.org/pension/gross- pension- replacement- rates.htm.
Peng, L., Meyerhoefer, C.D., Zuvekas, S.H., 2016. The short-term effect of depression on labor market outcomes. Health Econ. 25, 1223–1238.
Preston, S.H., Glei, D.A., Wilmoth, J.R., 2010. Contribution of smoking to international differences in life expectancy. In: Cohen, B., Crimmins, E.M., Pre-
ston, S.H. (Eds.), International Differences in Mortality at Older Ages: Dimensions and Sources. National Academies Press, Washington, D.C., pp. 105–131.
JID: JEBO
Rao, S.K., Wallace, L.M.K., Theou, O., Rockwood, K., 2017. Is it better to be happy or not depressed? Depression mediates the effect of psychological well-be-
ing on adverse health outcomes in older adults. Int. J. Geriatr. Psychiatry. 32 (9), 10 0 0–10 08.
Rockwood, K., Mitnitski, A.B., 2007. Frailty in relation to the accumulation of deficits. J. Gerontol. Ser. A 62, 722.
SAMHSA, 2016a. Key substance use and mental health indicators in the United States: results from the 2016 national survey on drug use and health. Center
for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration. Publication No. SMA 17–5044, NSDUH Series
H-52, Rockville, MD.
SAMHSA, 2016b. Substance abuse and mental health services administration. Behavioral health spending and use accounts, 1986–2014. Substance Abuse
and Mental Health Services Administration. HHS Publication No. SMA-16-4975, Rockville, MD.
SAMHSA (2016c). NSDUH DATA REVIEW: Suicidal Thoughts and Behavior Among Adults: Results From the 2015 National Survey on Drug Use and Health,
September 2016. https://www.samhsa.gov/data/sites/default/files/NSDUH- DR- FFR3- 2015/NSDUH- DR- FFR3- 2015.pdf.
Schulz, R., Drayer, R.A., Rollman, B.L., 2002. Depression as a risk factor for non-suicide mortality in the elderly. Biol. Psychiatry 52 (3), 205–225.
Schünemann, J., Strulik, H., Trimborn, T., 2017a. Going from bad to worse: adaptation to poor health, health spending, longevity, and the value of life. J.
Econ. Behav. Organ. 140, 130–146.
Schünemann, J., Strulik, H., Trimborn, T., 2017b. The gender gap in mortality: how much is explained by behavior? J. Health Econ. 54, 79–90.
Searle, S.D., Mitnitski, A., Gahbauer, E.A., Gill, T.M., Rockwood, K., 2008. A standard procedure for creating a frailty index. BMC Geriatr. 8 (1), 24.
Silveira, H., Moraes, H., Oliveira, N., Coutinho, E.S.F., Laks, J., Deslandes, A., 2013. Physical exercise and clinically depressed patients: a systematic review and
meta-analysis. Neuropsychobiology 67 (2), 61–68.
Solomon, A., 2015. The Noonday Demon: An Atlas of Depression. Simon and Schuster, New York.
Stevenson, B., Wolfers, J., 2008. Economic growth and subjective well-being: reassessing the easterlin paradox. In: Brookings Papers on Economic Activity.
Spring, pp. 1–87. 2008.
Strulik, H., 2015a. Frailty, mortality, and the demand for medical care. J. Econ. Age. 6, 5–12.
Strulik, H., 2015b. A closed-form solution for the health capital model. J. Demogr. Econ. 81 (3), 301–316.
Strulik, H., 2018a. The return to education in terms of wealth and health. J. Econ. Age. 12, 1–14.
Strulik, H., 2018b. Myopic misery: maternal depression, child investments, and the neurobiological poverty trap. Macroeconomic Dynamics. forthcoming.
Strulik, H., 2018c. Smoking kills: an economic theory of addiction, health deficit accumulation, and longevity. J. Health Econ. 62, 1–12.
Tomitaka, S., Kawasaki, Y., Ide, K., Akutagawa, M., Yamada, H., Ono, Y., Furukawa, T.A., 2018. Distributional patterns of item responses and total scores on
the PHQ-9 in the general population: data from the National Health and Nutrition Examination Survey. BMC Psychiatry 18 (1), 108.
Townsend, N., Wickramasinghe, K., Williams, J., Bhatnagar, P., Rayner, M., 2015. Physical Activity Statistics 2015. British Heart Foundation, London.
Trimborn, T., Koch, K.J., Steger, T.M., 2008. Multidimensional transitional dynamics: a simple numerical procedure. Macroecona. Dyn. 12, 301–319.
Vos, T., Haby, M.M., Barendregt, J.J., Kruijshaar, M., Corry, J., Andrews, G., 2004. The burden of major depression avoidable by longer-term treatment strate-
gies. Arch. Gen. Psychiatry 61 (11), 1097–1103.
WHO, 2017. Depression and Other Common Mental Disorders: Global health Estimates. World health organization, Geneva.
Zivin, K., Ilgen, M.A., Pfeiffer, P.N., Welsh, D.E., McCarthy, J., Valenstein, M., Kales, H.C., 2012. Early mortality and years of potential life lost among veterans
affairs patients with depression. Psychiatric Serv. 63 (8), 823–826.

Jurnal Health Behavior

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jurnal Health Behavior

Uploaded by

Copyright:

Available Formats

JID: JEBO

ARTICLE IN PRESS [m3Gsc;December 10, 2018;13:45]

Contents lists available at ScienceDirect

Journal of Economic Behavior and Organization

An economic theory of depression and its impact on health

2 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 3

4 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

2. A theory of depression, health behavior, and health outcomes

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 5

−λD μγ Ahγ −1 = λk φ p (7)

6 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

For the proof, I apply the implicit

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 7

8 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 9

4. Results: depression, life-cycle health behavior, and longevity

10 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

Fig. 1. Depression, Health Behavior, and Health Outcomes.

Case Parameter change Remark T u/u h/h x/x k/k

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 11

5. Applications and extensions

5.1. Depression and labor supply

5.2. Delayed treatment

12 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

Fig. 2. Delayed treatment.

5.3. Income gradient

5.4. Feedback effects from exercise

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 13

Fig. 3. Depression shocks.

6. Depression shocks and depression periods

6.1. Depression shocks

14 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

Case Timing T u/u h/h x/x k/k

6.2. Depression periods

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 15

Fig. 4. Depression Periods.

Fig. 5. Two depression periods.

16 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

A1. Derivation of (14) to (18)

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 17

− δηe−ηm ≤ λk φm pm with = for m > 0. (A.5)

18 H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx

H. Strulik / Journal of Economic Behavior and Organization xxx (xxxx) xxx 19

You might also like

Case Parameter change Remark T u/u h/h x/x k/k

Case Timing T u/u h/h x/x k/k