Living with a mental illness - the economic consequences of psychiatric disorders on income and portfolio selection

(1)

1 University of Amsterdam

Faculty of Economics and Business

Living with a mental illness - the economic consequences of

psychiatric disorders on income and portfolio selection.

Name:Lora Ivanova Student ID: 10004170 Supervisor: Jindy Zheng

BSc in Economics and Business Specialization: Economics and Finance

(2)

2

When offered a fair bet 50-50 chance of winning 200 euros or losing 150 most people would refuse to participate despite the high expected return (B. Shiv, 2005). Individuals are often reluctant to accept bets associated with extreme losses even if the expected gains are even bigger. The condition, called myopic loss aversion is only a small example of how feelings get in the way of rationality.

In contrast with most other social sciences, Economics assumes that ratio nal agents with well-defined and constant preferences interact in markets that eventually clear. Events that seem to be difficult to rationalize are often claimed as “anomalies” (J. Siegel, 1997). In the meantime, scholars in the fields of psychology, sociology and even game theory have witnessed the fact that such a generalization about human behaviour lacks descriptive realism. Empirical studies and experimental observations even among economists reveal that human behaviour is a lot more complex and diverse than depicted in the fictional world of rational models (Huang, 2002).

The reason why individuals take decisions that are not in their greatest interest or choose to behave irrationally when it comes to economic decisions is still unclear. Fear, stress,

impatience and emotional expectations all have a negative impact in the decision making process. Mental health problems can have an influence on individual’s emotions, compromising their ability to valuate decisions objectively (Bogan & Fertig, 2013). Furthermore, mental disabilities are linked to lower income and diminished working potential which further has an effect on confidence and investment decisions (Bartel & Taubman, 1986; M. Topper, 2010). Considering the crucial connection between mental health and the personal decision making process of individuals, the question whether mental health can have a significant effect on investment choices remains open in economic literature (Dahal & Fertig, 2013).

(4)

4 In their study, Eva Selenic and Bernard Batnic (2011) argue that there is a high

correlation between various mental health problems and financial hardship. Bartel and Taubman (1986) have found that individuals who suffer from psychosis or neurosis have significantly reduced earnings and a higher probability to remain single. The study demonstrates that mental health issues can have a substantial effect on economic and demographic characteristics. The authors claim that the pervasive effects that a mental health issue can cause, clearly demonstrate the benefits of increased government intervention, financial and moral support and better

treatment opportunities. Vicky Bogan and Angela Fertig (2012) speculate that mental health issues might lead to decreased amount of investments for households and have a long term effect on economic status.

In the USA when an asthma patient does not meet the requirements of a chronic physical disability, he/she can still be considered as a chronic disability patient if his job has restrictions that cannot be met due to the specific physical health condition (Laurence, 2014). On the

contrary, mental health issues, while officially recognized, are still burdened with a certain social stigma and diagnosed mental patients are still avoided or discriminated against (Hinshaw, 2007). Quite often, the criterion for mental disability claims is subjective and difficult to assess

(Laurence, 2014). According to Beth Laurence (2014), a disability claims examiner in the US , when it comes to mental health, disability examiners might be biased, perceiving mental health sufferers as lazy or malingering( faking their condition for money) . As compared to an asthma patient, a person with bipolar disorder is a lot less likely to announce his condition at work or search for any accommodations in regards to his condition, due to the social stigma connected to the diagnosis (Hinshaw, 2007) . In his book Stephen Hinshaw writes about the long held tradition to stigmatize mental health patients and provides multilevel strategies to fight against that

tendency, including innovative social policies, media coverage and particular treatment sessions. The above mentioned studies, while few and limited, demonstrate the daily suffering and economic effects of mental health issues on the affected individuals. It is essential to understand the differences in the decision-making process of those individuals in order to allow future policy makers to account for the changes and act accordingly. The limited collection of scientific studies that sheds light on the issue, unavoidably leads to the question – could mental health be an

omitted variable that should be included into consideration in future portfolio decision-making models ?

(5)

5 The purpose of this paper is to test empirically the effect of mental illnesses on economic decisions and income. The particular approach, applied, aims to compare investment decisions and income levels of individuals with different medical status. Similar to previous studies of mental health and economic status, mental illness is defined as psychological and emotional problems including but not limited to depression, bipolar disorder, schizophrenia, obsessive-compulsive disorder, anxiety and general emotional well-being.

According to the National institute of mental health around 26.6% of the US population is affected by a diagnosable mental health disorder each year. When applied to the population of US residents of 2004 the number translates to 57.7 million adults age 18 and older. Mental health disorders are one of the main causes of disability in the US and Canada and an increasing international concern globally. Nearly half of the individuals diagnosed with a mental disorder meet the diagnostic criteria for second or even third mental condition, the severity of which is strongly correlated to comorbidity (NIMH, 2009). Living with a mental illness might lead to substantial differences in income and investment behaviour in households. Investigating the relationship between the three might prove to be beneficial not only for predicting individual investment decisions, but also as guidance for future policy makers to track health disparities and design appropriate interventions.

There are several reasons why having a mental illness can impact household investment decisions and income. First, the presence of a mental disorder has a significant influence on individuals’ ability to regulate their mood and emotions (Bogan & Fertig, 2013). Rapidly changing emotional states as in the case of bipolar disorder, general helplessness and negativity in the case of depression or even irrational routines and superstitions in the case of obsessive -compulsive disorder might have a severe effect on the objective evaluation of investment opportunities.

Furthermore, the inability to control one’s emotions might cause significant difficulties at work, decrease productivity and at the same time increase spending for medical bills and therapy (Bartel & Taubman, 1986). As a consequence the disposable family income might be lower, which also leaves less funds to invest and less financial freedom when making allocation

decisions. A financial shock in the family might also promote the preference towards more liquid assets, more conservative investment strategies and altered individual discount rate for future

(6)

6 returns. In other words, mental health might also affect the degree of risk aversion of the

individuals, which is a crucial factor for making investment decisions.

2. Background and theory

There is a heated discussion on the correlation between health problems and income. Numerous papers outside the field of economics are tracing the relationship between physical health and economic status (Dahal & Fertig, 2013).The link between mental health and income has been a less disputed topic, yet there are papers that review the problem in question. Some argue that low income and a low job position create a stressful environment and are more likely to predispose to deterioration in healthand mental disorders (Lund, Myer, Stein, Williams, & Flisher, 2013). Others claim that most of the correlation actually comes from mental health affecting income (Bartel & Taubman, 1986). While there are economic papers supporting each side of the debate and even some that review both directions, Bartel and Taubman (1986) suggest that a way to reduce the reversed causality is to combine information on mental health with information of economic changes of a few years later. The purpose of this approach is to try to limit the cases of mental health issues caused by poor economic status and concentrate on the changes of income and portfolio selection a few years after a mental health condition is present.

While it is impossible to eliminate the reversed causality problem completely, the

specified approach ensures that a mental health diagnosis precedes changes in economic status. If the mental health condition was caught on time, using economic variables of a few years later would ensure that the economic changes were a probable cause of mental health deterioration and not vice versa. Even if the condition existed a lot before the diagnosis was made, the moment of receiving the official diagnosis often signifies, that the condition got more severe and as such it was easier to recognize from a health care specialist or it became alarming enough for the patient to finally seek medical attention. In this case, using economic variables of a few years later would allow tracing the changes from a less severe to a more serious condition and its effect on income and portfolio selection. In either case, using variables of income and portfolio selection of a few years later is a step further in reducing the reversed causality problem.

It is proven that the portfolio selection of individual investors is subject to a variety of behavioural biases (Bailey, Kumar, & Ng, 2011).Over the last century stocks have outperformed

(7)

7 bonds dramatically. One would expect that arbitrage opportunities would lower the difference, reflecting only the risk premium of investing in stocks. On the contrary, the intuitive expectation that stocks are riskier than bonds is not enough to explain the magnitude of the difference – the disparity is so high that it suggests implausibly high levels of risk aversion, a pattern referred to as the equity premium puzzle (B. Shiv, 2005). The emotional instability that goes hand in hand with mental health issues might be a piece in that mysterious puzzle.

The correlation between mental health and investment decisions is a relatively unusual topic in the fields of economics and finance. The lack of representative and complete data on the subject creates substantial difficulties in outlining the relation. Bogan and Fertig (2 013) write about the connection between portfolio choice and mental health. The authors argue that having a mental disorder substantially reduces holdings of risky assets. They claim that mental health has an impact on cognitive abilities, mood regulation, risk aversion and individual’s discount rate, and relate them to financial decision making. The authors use data from multiple waves from the Health and Retirement survey (HRS). Even though the study in question has detailed information on financial assets and mental health, it tracks only older individuals and is therefore not

representable for the entire population. Furthermore, the study only traces the impact of depression, memory and general well-being.

The current paper aims to trace connections between actually diagnosed mental conditions, self-reported well-being indicators and economic and investment indicators. Specifically, the health conditions of interest are depression, bipolar disorder, anxiety and obsessive-compulsive disorder.

Depression is one of the most common mental health issues reported worldwide.

According to Kessler (2006) a single major depressive episode leads to approximately 5 weeks of lost productivity and earnings. There is already a sound scientific evidence that the relation between depression and poor school performance, lower grades, health problems, smoking, poor social performance, drug use and engagement in highly risky behaviours (M. Topper, 2010). Depressed individuals might experience fatigue and low energy, physical and mental lethargy and concentration and memory problems. Such symptoms can decrease their working capacity and impair their rational judgement and as a result have an effect on family earnings and rational investment decisions.

(8)

8 The majority of individuals who suffer from bipolar disorder tend to circulate between states of mania and depression. They express rapid and sudden mood changes from general excitement to irritability and hostility. The maniac episodes are often reinforced with the

tendency to make grand and unattainable plans, unrealistic beliefs in one’s abilities, impulsivity, poor judgement of the situation at hand and increased reckless behaviours. The above mentioned characteristics are highly relevant for investment decision making and family income. Individuals in the state of mania might engage into lavish shopping sprees, invest into a certain asset

impulsively or decide to quit their job prematurely (Angus, 2013).

One of the main features of anxiety is reduced risk taking (Giorgetta C1, 2012). Anxious individuals engage in risk avoidance, their hypersensitivity to potential threats and pessimistic evaluation of future events reduce risk-taking behavior and have a significant impact on their decision making abilities (Giorgetta C1, 2012). The observed behaviour is highly relevant when making rational investment decision and might have an effect in portfolio choice, income and general life-style tendencies for the specified individuals.

Individuals suffering from obsessive-compulsive disorder might follow a certain strict routine or engage in repetitive and irrational practices. Experiencing such symptoms might impact their investment decisions and interfere with their rational judgement.

Based on all of the above mentioned studies, the predicted results of the paper are speculated to prove that depressed individuals would avoid riskier investments; bipolar

individuals in the state of mania would show a positive preference with regards to riskier choices and on average women would prefer safer investment options than man. Furthermore, for all diagnosed mental health conditions, income is expected to be negatively related with metal health status.

3. Data

The data set used for this paper comes from Panel Study of Income Dynamics (PSID). It provides data on waves, starting as early as the 1968 and together with the National Comorbidity survey is one of the very few national representative studies that provide data on economic dimensions and

(9)

9 mental health. Until 1997 the data has been collected annually and after that biannually.

Household asset allocation questions and mental state questions were asked in all of the six most recent waves. From 2005 the survey also includes information about diagnosed mental illnesses. For the purpose of this paper the waves from 2005 until 2011 are used and adjusted with weights provided by the PSID in order to approximate a representative sample of the US population.

3.1 Subgroups

Research shows that the decisions and earnings of single households vary substantially from those of families (Dahal & Fertig, 2013). In addition, in order to distinguish the direct effect of mental illness of the head of the household on asset allocation and inco me, only single

individuals are taken into consideration. Furthermore, gender differences in investment styles exist (Barber & Odean, 2001) and should be accounted for. Barber and Odean claim that man trade 45% more than women due to overconfidence. Also, on average male investors are less risk averse than female ones and there are also differences in their discount factor. Due to the

significant variances in investment decisions between the two genders, single households are further separated into male and female. After deleting all observations with a missing marital status and getting rid of individuals that do not participate in all of the studied years, the

remaining observations represent a balanced data panel. The data is then weighted with weights from the last wave used (2007) in order for the weights to be consistent within the panel.

(10)

10 3.2 Mental health measures

The PSID data allows assessment of mental health in two different distinctive ways. First, in the last 4 waves namely 2005, 2007, 2009, 2011 respondents were asked if they were ever diagnosed with a certain mental illness. A dummy variable is created for those answers.

Table 2: Weighted mean and standard error of females and males that have reported a mental problem.

A follow up question allows respondents to specify if their mental illness corresponds to one of 9 namely: depression, schizophrenia, anxiety, phobia, bipolar disorder (only state of mania),

alcohol abuse, drug addiction, obsessive compulsive disorder and other (unspecified). The questionnaire permits each individual to state up to three mental disorders. An indicator variable is created for each of depression, bipolar disorder, obsessive-compulsive disorder and anxiety.

Women .126612 .0123699 .1023549 .1508692 Men .0960644 .0243682 .0482788 .1438499 ifprob

Over Mean Std. Err. [95% Conf. Interval] Women .0039749 .0017978 .0004495 .0075002 Men .0221951 .0141422 -.0055374 .0499276 OCD Women .0356182 .0072454 .0214101 .0498263 Men .0192202 .0111366 -.0026183 .0410587 Anxiety Women .0238726 .0051758 .013723 .0340221 Men .0118848 .0055554 .0009908 .0227788 Bipolar Women .0779797 .0104802 .0574282 .0985311 Men .0227759 .0143737 -.0054106 .0509623 Depression Over Mean Std. Err. [95% Conf. Interval]

Table 3: Weighted means and standard errors for females and males who reported depression, bipolar disorder, anxiety and obsessive-compulsive disorder

(11)

11 One of the main drawbacks of using only diagnosed disorders is that it might not take into account individuals with a lower socio-economic background who cannot afford mental health services and a professional opinion. Furthermore, the diagnosis might not be an indicator of the present health condition of the individual.

As a solution, the second measure of mental health used is a K-6 non-specific

psychological distress indicator. The scale is a tool developed by R. Kessler, Harvard Medical school professor, in order to capture the different stages of psychological distress based on a number of questions about the current state of the individual (Kessler, 2003). The respondents are asked to indicate on a scale from 1 to 5 how often they have felt sad, nervous, restless, hopeless and that everything is an effort in the last month (Dahal & Fertig, 2013). An answer “all of the time” gives 4 points, “most of the time”- 3,”some of the time” – 2, “a little time” -1 and none of the time gives 0. The K-6 indicator represents the sum of all those scores where a score of 13 or higher indicates the threshold for the clinically significant distribution of nonspecific

psychological distress (Kessler, 2003). On the basis of this information an additional variable that signifies a score above 13 is created.

Table 4: Weighted means for K-6 psychological distress indicator

The K-6 score might be less objective since the responses might depend strongly on the current emotional state of the individual. On the other hand, since the measure ranges between 0-24 it is good for estimating mild and more serious mental issues (Dahal & Fertig, 2013). Some studies claim that the indicator might predict depression and anxiety quite accurately, but the exact condition cannot really be distinguished for research purposes.

In order to avoid reverse causality, mental health responses are lagged by one wave - 2 years. To prevent catching the effect of economic distress on mental health, instead of the effect of mental health on economic variables, the information about mental health is always 2 years

Men .0480662 .0181062 .0125538 .0835786 Women .0521065 .0092005 .0340612 .0701518 k6Above13 Men 3.704643 .3839595 2.951568 4.457719 Women 3.796024 .1912034 3.421008 4.171039 k6 Over Mean Std. Err. [95% Conf. Interval]

(12)

12 older. In the case of chronic mental illness, if the severity of the condition remains constant through a long period of time, there should be no change in the economic measures of the individuals and such results would not be taken into account from the fixed effects economic model, applied later in the paper. Even though reversed causality in the case of chronic mental patients could not be avoided, the fixed effects model automatically excludes observations that show no variance between the subsequent years and as such, it will exclude patients that have had the diagnose through the whole observation period. Bogan and Fertig (2013) argue that “lagging” the mental health variables allows significant time for mental health issues to change the

investment and economic behaviour of individuals.

3.3 Assets and income measures

In order to account for the types of assets, held by individuals with mental health problems and the effect of mental illness on income, both measures are represented by their appropriate variables and two different types of regression models are performed at the end.

Like most large longitudinal data studies, PSID divides the information of specific assets in categories. For the purpose of this study, two asset categories are evaluated, where specific assets are collapsed together based on the volatility of the investment. Holding money in

checking and savings accounts, money market funds, certificates of deposit, government bonds, valuable collection, cash in an insurance policy or treasury bills is considered relatively safe. This does not include valuables in employer-based pensions or IRA’s. An interview question in each of the four waves is designed to record if respondents hold any of the above mentioned assets. A dummy variable named “Safe” is created accordingly.

Assets such as shares of stocks in publically held corporations, mutual funds or

investments in trusts are collapsed together to represent riskier investments. The dummy variable “Risky” indicates if the respondent holds such assets.

(13)

13 Table 5: Weighted mean of safer and riskier asset holding indicators

In order to account for the effect on income, the variable income represents the annual gross income of every head of household.

Table 6: Weighted mean of gross income for single women and men

3.4 Control variables

The tables below provide summary of the control variables used in the study. The variables include demographic characteristics of the individuals, socioeconomic variables, physical health measurements, medical expenses and financial sophistication variables. In addition dummy variables for the year of each of the four waves are created.

The demographic characteristics include age, race and marital status of the individuals, where a dummy variable is created for each separate option.

Women .0679839 .0097266 .0489102 .0870576 Men .1384164 .0276371 .0842207 .1926122 Risky Women .628181 .0174848 .5938936 .6624685 Men .6250987 .0361513 .5542068 .6959907 Safe Over Mean Std. Err. [95% Conf. Interval] Women 23848.7 947.7734 21990.07 25707.33 Men 31128.49 4100.38 23087.45 39169.53 Income Over Mean Std. Err. [95% Conf. Interval]

(14)

14 Table 7: Weighted demographic characteristics of the sample.

The socio-economic characteristics include the total amount of years of education, employment status, a dummy variable that accounts for work disabilities and a variable that measures total household net worth including real estate value if the head of the household owns a house.

Women 1.129612 .0431731 1.04495 1.214273 Men .9253554 .0788603 .7707121 1.079999 Children Women .102269 .0102903 .08209 .1224481 Men .0946771 .0201485 .0551663 .1341879 Separated Women .1051772 .0129199 .0798415 .1305128 Men .0097348 .0047225 .000474 .0189956 Widowed Women .4451839 .0184754 .4089541 .4814138 Men .5666145 .0360629 .495896 .6373331 Divorced Women .3473699 .0172894 .3134657 .3812741 Men .3289735 .0325553 .2651332 .3928138 Single Women .0121661 .0036636 .0049817 .0193504 Men .0040413 .0020392 .0000425 .0080401 Asian Women .0051036 .0025486 .0001059 .0101014 Men 0 0 . . Indian Women .4604683 .0181969 .4247846 .496152 Men .1329847 .0198423 .0940743 .1718951 AfricanAmerican Women .4950957 .0185629 .4586943 .5314971 Men .7995216 .0251319 .7502385 .8488047 White Women 46.10118 .4350104 45.24813 46.95422 Men 43.99672 .9602895 42.11362 45.87983 Age Over Mean Std. Err. [95% Conf. Interval]

(15)

15 The financial sophistication variables include a dummy variable that records whether the head of the household received any inheritance in the last year. The second variable measures the total amount of financial support spent on other family members. To account for any permanent physical health problems, a dummy variable that records whether or not the individual has a chronic medical condition is created. Additionally, when accounting for portfolio choice

allocation, the variable that accounts for income (presented in the previous part) is included as a control variable. Women 48891.4 12223.95 24920.09 72862.71 Men 77488.65 15360.67 47366.18 107611.1 NetWorth Women .0560975 .0079219 .0405626 .0716324 Men .0895099 .0235015 .0434233 .1355966 Disabled Women .0757671 .0082934 .0595036 .0920306 Men .1232876 .0278664 .0686412 .177934 Retired Women .7081118 .0169999 .6747747 .7414488 Men .6482935 .0374791 .5747966 .7217904 Working Women 12.55302 .0916227 12.37334 12.73269 Men 11.89752 .2129074 11.48 12.31503 YearsOfEducation Over Mean Std. Err. [95% Conf. Interval] Women .1535641 .014506 .1251181 .1820101 Men .1618211 .0308491 .1013264 .2223158 ChronicCondition Women 390.6415 98.84252 196.8128 584.4703 Men 1535.001 409.6969 731.5918 2338.411 TotalSupportOfOthers Women .0274219 .0067823 .014122 .0407219 Men .0266654 .0131776 .0008244 .0525065 ReceivedInheritance Over Mean Std. Err. [95% Conf. Interval]

(16)

16 Table 9: Financial sophistication weighted variables

4. Empirical strategy

4.1 Preferred selection of assets

In order to measure the impact of mental health issues on the probability of holding a specific type of asset (safe or risky as already specified) the following model is developed:

The dependent variable represents holding a specific type of asset for household n at time t. is one of the above specified mental health measures for household n and lagged 2 periods: t-2. represents all the control variables given for household n at time t.

Realistically, even the most sophisticated economic models are likely to suffer from omitted variable bias. Using panel data for research purposes, allows an alternative approach- since using a panel implies that data is available for the same individuals in different points in time, the subjects can be used as their own controls (Allison, 2009). Stable characteristics that do not change through time such as sex, gender or even one’s intelligence can be accounted for, no matter if they are measured or not. When the dependent variable is binary, this can be achieved with conditional/fixed effect logit models. Using the fixed effect logit model provides satisfactory results since it keeps the time-invariant characteristics constant (Bogan & Fertig, 2013) . The emphasis of this model is to catch the significant differences in the change of mental health status or the severity of the condition causing drastic changes in asset holdings.

While the fixed effects model provides some useful insights, given that the available for analysis waves that contain mental health characteristics are limited- only 4, in a lot of cases the change in mental health variables might have been a fact long before that, therefore the fixed effects model would not capture those cases. Furthermore, in a lot of households, the mental health issue is present long before the person was diagnosed, thus it will have an impact on asset allocation long before that. Due to those concerns all logit regressions are presented both with and without the fixed effects option.

(17)

17 4.2 Effect of mental health on income

The second analysis of the paper traces the connection between mental health indicators and income.

The model is:

Where represents annual gross income for household n at time t and the rest of the

variables are as described in the previous section. The control variables for net worth, inheritance and total support of others are dropped from the regression, because they are not relevant in this case. The Tobit random effects model for panel data is applied.

5. Results

5.1 Portfolio selection

The following extracts from the regression tables represent the standard errors and coefficient for each of the different mental health measures, provided separately for men and for women. Due to the high amount of performed regressions, only the relevant coefficients concerning mental health and asset holdings are presented, while the control variables are omitted from display. For each table random effects are presented first, followed by fixed effects.

5.1.1 Depression

Table 10: Random and fixed effects of depression on safe assets among single males

L1. -1.646133 .4854545 -3.39 0.001 -2.597606 -.6946593 depression

Safe Coef. Std. Err. z P>|z| [95% Conf. Interval]

L1. 3.94596 2.895346 1.36 0.173 -1.728814 9.620734 depression

(18)

18 Table 11: Random and fixed effects of depression on safe assets among single females

The random effects measurements of holding safe assets for both single females and males are significant and suggesting a negative relationship. The results imply that depressed individuals are less likely to hold safe assets. On the other hand, the results from the fixed effects regressions, while insignificant, both show a small positive relationship between depression and safe asset holdings. While an insignificant result shows that there is a high possibility for the null hypothesis to be true, it still does not really reject the opposite.

The fixed effects model allows for time invariant characteristics to be held constant and to use individuals as their own controls, accounting only for changes through time. As such the fixed effects regression’s results are preferred for the purpose of this analysis. The results for safe assets attest that there is a probable relation between depression and holding safe assets, yet the exact connection (positive or negative) might still be a bit unclear.

Table 12: Random effect of depression on risky assets among single males L1. -1.291682 .1851369 -6.98 0.000 -1.654543 -.92882 depression

Safe Coef. Std. Err. z P>|z| [95% Conf. Interval] L1. .0645388 .120524 0.54 0.592 -.1716839 .3007616 depression L1. -12.31663 7128.808 -0.00 0.999 -13984.52 13959.89 depression Risky Coef. Std. Err. z P>|z| [95% Conf. Interval]

L1. -.7414154 .2589271 -2.86 0.004 -1.248903 -.2339276 depression

Risky Coef. Std. Err. z P>|z| [95% Conf. Interval]

(19)

19 Table 13: Random and fixed effects of depression on risky assets among single females

For riskier assets, all of the results for females are significant, suggesting a strong negative relationship between holding risky assets and depression for females. While

insignificant, the result for single man shows a strong negative trend for holding negative assets. The obtained results are in agreement with the theory that the discount factor for

individuals with mental health issues may be higher and so reducing their desirability to invest in assets with a delayed reward. Furthermore, they coincide with the expectations of the result, stated in the theoretical framework.

5.1.2 Bipolar disorder

Due to lack of change through the years in bipolar diagnosis, the fixed effect regression is omitted.

Table 14: Random effects of bipolar disorder on safe and risky assets among single males

L1. -7.214667 1.781378 -4.05 0.000 -10.7061 -3.723229 depression

L1. 4.31608 .4824086 8.95 0.000 3.370576 5.261583 bipolar

L1. -13.74748 7916.391 -0.00 0.999 -15529.59 15502.09 bipolar

L1. -2.33327 .753825 -3.10 0.002 -3.81074 -.8557999 bipolar

L1. -21.25543 25007.88 -0.00 0.999 -49035.8 48993.29 bipolar

(20)

20 Table 15: Random effect of bipolar disorder on safe and risky assets for single females

For bipolar disorder a strong, significant result shows that men prefer to hold safer assets and women diagnosed with the condition actually have a negative relation to safe asset holding. The results for riskier assets, while insignificant, demonstrate a strong negative trend.

5.1.3 Anxiety

Table 16: Random and fixed effects of anxiety on safe assets for single males

Table 17: Random and fixed effects of anxiety on safe assets for single females

From the individuals diagnosed with anxiety women have a significant negative correlation with safe asset holding, while men actually prefer to pick safer assets

L1. -1.189756 .4503519 -2.64 0.008 -2.07243 -.307083 anxiety

Safe Coef. Std. Err. z P>|z| [95% Conf. Interval] L1. -.0627081 2.837229 -0.02 0.982 -5.623574 5.498158 anxiety L1. 1.799523 .2044795 8.80 0.000 1.398751 2.200295 anxiety Safe Coef. Std. Err. z P>|z| [95% Conf. Interval] L1. 16.21893 771.3841 0.02 0.983 -1495.666 1528.104 anxiety L1. -23.64314 8979.104 -0.00 0.998 -17622.36 17575.08 anxiety Risky Coef. Std. Err. z P>|z| [95% Conf. Interval]

(21)

21 Table 18: Random and fixed effects of anxiety on risky assets for single males

Table 19: Random and fixed effects of anxiety on risky assets for single females

For both genders the results from the risky assets regression is insignificant, yet with a high negative coefficient, suggesting a negative trend.

5.1.4 Obsessive-compulsive disorder

Table 20: Random and fixed effects of OCD on safe assets for single males

Table 21: Random effects of OCD on safe assets for single female

According to the results, male individuals diagnosed with OCD have a strong negative relationship with holding safe assets. While insignificant, perhaps due to the small amount of observations available, the regression for single females signifies a strong positive trend for holding safe assets.

L1. -6.439025 178516.4 -0.00 1.000 -349892.1 349879.2 anxiety

L1. 2.604257 .2422618 10.75 0.000 2.129433 3.079081 anxiety

Risky Coef. Std. Err. z P>|z| [95% Conf. Interval] L1. 3.049354 .4422855 6.89 0.000 2.18249 3.916218 anxiety L1. -1.896066 .3440834 -5.51 0.000 -2.570457 -1.221675 ocd Safe Coef. Std. Err. z P>|z| [95% Conf. Interval] L1. -10.97139 1.791475 -6.12 0.000 -14.48262 -7.460164 ocd L1. 19.76396 2145.493 0.01 0.993 -4185.325 4224.853 ocd Safe Coef. Std. Err. z P>|z| [95% Conf. Interval]

(22)

22 Table 22: Random effects of OCD on risky assets for single male

Table 23: Random and fixed effects of OCD on risky assets for single female

The results on holding riskier assets are all insignificant, with larger positive coefficients for females and negative coefficient for males, showing that females diagnosed with OCD might have a positive preference for investing in risky assets, while on the other hand, the results for men show that they would rather invest less in risky assets if they are diagnosed with obsessive -compulsive disorder.

5.1.5 K-6

Table 24: Random and fixed effects of a K6 distress scale above 13 on safe assets for single males

L1. -1.490407 2804.673 -0.00 1.000 -5498.549 5495.568 ocd

L1. 26.55716 57500.32 0.00 1.000 -112672 112725.1 ocd

L1. .8655316 1.238018 0.70 0.484 -1.56094 3.292003 ocd

L1. 1.456713 .2534174 5.75 0.000 .9600244 1.953402 k6Above13

Safe Coef. Std. Err. z P>|z| [95% Conf. Interval] L1. 6.86191 1.25394 5.47 0.000 4.404233 9.319588 k6Above13 L1. -.288704 .1433901 -2.01 0.044 -.5697434 -.0076646 k6Above13 Safe Coef. Std. Err. z P>|z| [95% Conf. Interval]

L1. -.082305 .1885114 -0.44 0.662 -.4517805 .2871705 k6Above13

(23)

23 Table 25: Random and fixed effects of a K6 distress scale above 13 on safe assets for single females

When it comes to safer assets, men with a high K-6 score have a preference of holding the safer options, while the result for women shows a small negative relation.

The amount of individuals with k-6 score above 3 does not change among the risky asset holders according to the fixed effects regression, therefore it is omitted.

Table 26: Random effects of K6 distress scale above 13, on risky assets for single males

Table 27: Random and fixed effects of K6 distress scale above 13, on risky assets for single females

Individuals with a K-6 result above 13, while with insignificant results, both (women and men) show a strong negative trend towards holding risky assets. A lot of researchers believe that the K-6 indicator is a reliable measurement of depression and the results of the regression are in an agreement with the ones obtained for diagnosed depression.

5.2 Effect of mental health variables on income

The following tables present the effect of having a mental health diagnosis on income. For each mental health measurement the first regression is for single male and the second for single female heads of households.

L1. -20.4975 4322.799 -0.00 0.996 -8493.029 8452.034 k6Above13

L1. -19.89188 4319.174 -0.00 0.996 -8485.318 8445.534 k6Above13

L1. -2.192414 .6529722 -3.36 0.001 -3.472216 -.9126118 k6Above13

(24)

24 Table 28: Random effects tobit regression for male and female heads with depression

When it comes to income, a diagnosed depressive condition leads to lower income for males and while the result for single females suggests a small positive relationship. The obtained results are completely in line with the expectations stated in the theoretical framework of the paper.

Table 29: Random effects tobit regression for male and female heads with bipolar disorder

As expected, the effect on income for both genders when diagnosed with a bipolar disorder is negative and highly significant.

L1. -1.170855 .2116373 -5.53 0.000 -1.585656 -.7560531 depression

lnIncome Coef. Std. Err. z P>|z| [95% Conf. Interval]

L1. .0824858 .0275558 2.99 0.003 .0284775 .1364941 depression

L1. -.9283102 .2063266 -4.50 0.000 -1.332703 -.5239176 bipolar

L1. -.4496067 .1025811 -4.38 0.000 -.650662 -.2485513 bipolar

L1. .5699509 .1337286 4.26 0.000 .3078477 .832054 anxiety

(25)

25 Table 30: Random effects tobit regression for male and female heads with anxiety

Being diagnosed with anxiety is related to a slightly higher income for men. This relates to the fact that individuals with anxiety disorder value highly security (both financial and

general). The result for women is insignificant, but the coefficient shows that if the null

hypothesis (no effect between income and anxiety) was rejected there would have been a small negative relation between income and anxiety for women.

Table 31: Random effects tobit regression for male and female heads w ith OCD

Table 32: Random effects tobit regression for male and female heads with k-6 above 13

L1. -.0806659 .0527008 -1.53 0.126 -.1839575 .0226257 anxiety

L1. -1.015471 .6865471 -1.48 0.139 -2.361078 .3301369 ocd

L1. .4367776 .6408588 0.68 0.496 -.8192825 1.692838 ocd

L1. .1465958 .0835858 1.75 0.079 -.0172293 .3104209 k6Above13

L1. .1128313 .0793202 1.42 0.155 -.0426334 .268296 k6Above13

(26)

26 For income, both coefficients for men and women with a high K-6 score show a small positive relationship – a higher distress score is related to receiving a slightly higher income, yet only the coefficient for single males is significant.

6. Conclusions

Approximately 30% of the US population has been diagnosed with a mental condition. The consequences of living with a mental illness are indisputably present while very few researchers have tried to investigate the connection between mental health issues and economic status. Understanding the investment patterns and income effects of mental patients is beneficial not only as a bizarre statistic. Understanding those 30% of the population might be useful for future policy decisions and laws, understanding of the financial world and explaining some peculiar financial puzzles of irrationality.

The conducted research clearly shows that various connections between investment decisions, income and mental health exist. The results of the two models show that there are also gender differences when it comes to investments and income. Most of the single women

diagnosed with a mental health condition show a negative relationship with holding safe assets, while most men prefer them. The regressions performed for riskier assets even though

insignificant in most cases show a reoccurring negative trend.

Due to the fact that PSID started to ask about diagnosed mental illness only in its last 4 available waves (2005-2011) the amount of observations available for this study has been limited. After deleting respondents that do not participate in all 4 waves and leaving only the ones that are single heads of a household to get rid of additional effects from the spouse, the sample of

consisted of 2343 pooled observations. The small number of observations is a possible limitation of the study, which might have caused the lack of significance in some of the regressions.

Furthermore, the sample consists of around 3 times more women than man each year, which means that the results for women might have been given with a lot higher precision. One of the last possible drawbacks of the model is that when working with panel data, the logit and tobit models require that weights should be constant across years. Due to that fact, the weights

(27)

27 provided for the last wave have been used across the whole sample instead of using the sample weights provided separately for each year, which might cause certain deviations from reality.

Regardless of the limitations of the models, some of the findings were in agreement with the speculated results. Depressed individuals have a negative correlation with investment in riskier assets and most of the health conditions are related to lower income. Contrary to

expectations, bipolar individuals do not show a positive connection with risky asset holdings and females with mental health issues do not demonstrate a positive preference for safer assets.

Understanding the relationship between mental health and economic variables has been proven to be important and does show significant relationship results. Further understanding and research on those effects might be a base for a lot of behavioral finance topics of interest and a future pool of information for economic analysis and policy-making strategies.

(28)

28

Bibliography

Allison, P. (2009). Fixed Effects Regression Models (Quantitative Applications in the Social Sciences). Sage.

Angus, J. (2013). Bipolar disorders in DSM-5: strengths, problems and perspectives. Zurich.

B. Shiv, G. L. (2005). Investment behaviour and the negative side of emotion. American psychological society.

Bailey, W., Kumar, A., & Ng, D. (2011). Behavioral biases of mutual fund investors. Journal of financial economics 102, 1-27.

Barber, B. M., & Odean, T. (2001). Boys Will be Boys: Gender, Overconfidence, and Common Stock Investment. The Quarterly Journal of Economics, 116(1), 261-292.

Bartel, A., & Taubman, P. (1986, Apr.). Some economic and demographic consequences of mental illness. Journal of Labor Economics , 4(2), 243-256.

Bogan, V. L., & Fertig, A. R. (2013). Portfolio choice and mental health. Review of finance, 955-992. Dahal, A., & Fertig, A. (2013). An econometric assesment of the effect mental illness on household

spending behavior. Journal of Economic Psychology, 37, 18-33. David Pevalin, K. R. (2009). The Stata Survival Manual. McGraw-Hill.

Giorgetta C1, G. A. (2012). Reduced risk-taking behavior as a trait feature of anxiety. Emotion(12(6):1373-83).

Hinshaw, S. P. (2007). The mark of shame: Stigma of mental illness and an agenda for change. New York: Oxford University Press.

Huang, P. (2002). Beliefs, Fears, & Feelings of Guilt in Securities Investing. Berkeley Program in Law and Economics, Working Paper Series.

J. Siegel, R. T. (1997). Anomalies, The equity premium puzzle. Journal of economic perspectives, 11(1), 191-200.

Jitender, S., Afifi, T. O., McMilan, K. A., & Amudson, G. J. (2011). Relationship Between Household Income and Mental Disorders. Arch Gen Psychiatry 68(4), 419-427.

Kessler, R. e. (2003). Screaning for serious mental illness in the general population. Archives of general psychiatry(60), 184-189.

Laurence, B. (2014). Can You Get SSD or SSI Disability for Asthma? Retrieved from Disability Secrets. Laurence, B. (2014). Is Social Security's Assessment of Mental Disorders Fair? Retrieved from Disability

(29)

29 Lund, C., Myer, L., Stein, D. J., Williams, D. R., & Flisher, A. J. (2013). Mental illness and lost income

among adult South Africans. Social Psychiatry and Psychiatric Epidemiology, 845–851. M. Topper, P. M. (2010, June). Improving prevention of depression and anxiety disorders: Repetitive

negative thinking as a promising target. Applied and preventive psychology, 57-71. NIMH, N. I. (2009). The Numbers Count: Mental Disorders in America. Retrieved 2014, from

Transforming the understanding and treatent of mental illness: http://www.nimh.nih.gov/ P. M. McEvoya, A. E. (2013, March). Intolerance of uncertainty and negative metacognitive beliefs as

transdiagnostic mediators of repetitive negative thinking in a clinical sample with anxiety disorders. Journal of anxiety disorders, 27(2), 216-224.

S. Lee, A. Y. (2010). Individual and societal impact on earnings associated with serious mental illness in metropolitan China. Psychiatry research, 180, 132–136.

Selenko, E., & Batinic, B. (2011). Beyond debt. A moderator analysis of the relationship between perceived. Social science and medicine, 73, 1725e1732.

Living with a mental illness - the economic consequences of psychiatric disorders on income and portfolio selection