This paper was submitted on May 6th, 2022 as my course paper for SS368: Econometrics.
Introduction
It is important to understand the relationship between economics and mental health because virtually all individuals are economic actors, and the human side of economic decision-making is essential to an ethical economic policy system. Applying economic analysis to medical and psychological problems can help us gain a deeper understanding of important issues. Studies, discussed below, have been done within countries to examine different factors that affect mental health, but I want to explore mental health across multiple countries and see if the relative wealth of those countries has any impact on mental health. If an effect is drawn between GDP and mental health, perhaps global financial resources could be directed to countries with lower GDP per capita to improve their aggregate mental health. My research question for this project is: what is the causal effect of GDP per capita on mental health, measured by suicide rate? My hypothesis is that suicides will decrease as wealth increases. Ultimately, as the regression results show, there is a zero causal relationship between GDP per capita and suicide rate. The most likely reasoning for this is omitted variable bias.
I believe GDP per capita to be a good measure of wealth because it provides the average living standard in a country. It does not quite represent the income of each person in a country’s economy, as many households have fewer earners than inhabitants, and there may also be significant wealth outliers in some countries that distort the average. However, it works well as a tool on a relative basis to other countries to compare standards of living and wealth. Suicide rate is a good empirical measure of mental health because it does not rely on reporting and diagnosis of mental illnesses. Suicides are included in cause-of-death data and are not more likely to be reported in countries with better health infrastructure like general mental health are. This may not capture all mental illness in the world, but I do not believe there is any data set that can, given the current state of world development.
Literature Review
It has already been established that there is a strong association between wealth inequality and psychological distress. However, this study was conducted in one country (New Zealand) and measures the relative attitudes and health outcomes of citizens of the same country. We can expect that the trend with inequality will be similar when drawn to an international scale and will explore if the same relative concept applies in an absolute scenario of wealth. It would be interesting to see if there was a “living standard” level of income after which mental health problems dropped off significantly (Carter et al. 2009, 221).
There is also evidence that poverty in early childhood results in worse mental health outcomes later in life. This points to cultural factors of poverty having a relationship with mental health. Because of wealth disparity being greater in certain countries than others, it would be interesting to see how a study controlled for both absolute and relative wealth would evaluate the relationship between wealth and mental health. If it is true that most psychiatric disorders are associated with childhood onset affected by wealth, this could greatly inform policy with regards to youth education and economic policy. It will also be interesting to see if the trends that appear on a national scale persist on a global scale, or if poverty has different effects in different parts of the world (Lê-Scherban, Brenner, and Shoeni 2016, 798).
A study from Child Development presents an alternative perspective on the relationship between wealth and mental health. The study identifies that children from affluent backgrounds, compared to middle-class children, are more likely to experience substance abuse, anxiety, and depression. This study was conducted within the United States and is likely not representative of the entire world. It points to an “affluenza” that may exist in wealthier nations, and I am interested in exploring if this trend has a certain threshold level up to which wealth improves mental health and after which mental health declines (Luthar 2013, 1590).
A study from the World Health Organization found a statistically significant relationship between poverty and diagnosis for mental disorders. This study differs from most others in that it examines the relationship between wealth and mental health on a global scale. However, it comes short of the scope of my research question because it does not look at mental health as a linear function of wealth. Rather, it uses a probabilistic model with “poor” and “wealthy” as categories. I want to examine the relationship of wealth and mental health at all wealth levels (Patel and Kleinman 2003, 609-610).
Data and Sample
The six variables used in this analysis are suicide rate measured in suicides per 100,000 people, GDP per capita in thousands of dollars, percent of total GDP comprised by industrial output, Gini coefficient, population density measured in population per square kilometer of land mass, and percent of total GDP comprised by military spending. All these data represent one country in the year 2013, from a data set of 164 countries selected on availability of data. All countries on earth for which there were data were included in the sample. I chose 2013 because it was a “normal” year in my assessment, with no significant financial crises that might affect data points like data from the 2008 crisis or COVID-19 years might.
Column 1 lists the variables used in analysis. Suicides per 100,000 people is the dependent variable and measures the suicide rate in 164 different countries. The data comes from the Global Burden of Disease Study, most recently conducted in 2017. I chose to examine the year 2013, due to the completeness of the data. It is fair to assume that suicides are well-reported in all countries, unlike a mental health diagnosis or something similar. This reduces the likelihood of selection bias.
The primary independent variable is GDP per capita. I want to examine the effect of this variable, representing wealth, on the suicide rate. This data comes from a World Bank 2013 survey and reports values in thousands of 2020 US Dollars. This variable appears to have a wide distribution, given that the mean is hardly larger than the standard deviation. This is one of the common problems with averages in that it is vulnerable to outliers. However, this is the best indicator of wealth that is available for empirical study. Individual income data is likely not as well reported as national GDP is, introducing room for selection bias.
The industrial output control takes the GDP of a country’s industrial sector and divides it by the total GDP. This data comes from The World Bank. I selected this variable because it might reflect the lifestyle or economic constitution of a country. A country with a lower percentage might be more agrarian and less urbanized, while high-percentage countries are likely more urbanized and have more specialized jobs.
Gini coefficient is included to ensure that GDP per capita measurements are controlled for wealth inequality. The distribution of wealth may have some effect on mental health. This data comes from GapMinder, which compiled multiple 2013 surveys into one more complete data set.
Population density, measured in people per square kilometer, is used as another control for GDP per capita, to ensure that urban and rural lifestyle differences are accounted for. This data comes from the 2013 United Nations World Population Prospects report.
Military spending was sourced from the Stockholm International Peace Research Institute in 2013. I hope to use this as a measure of the general peacefulness of a country, which may influence the suicide rate in a way not captured by GDP.
A potential weakness of the data is the inherent selection bias with any large-scale study. The data are not perfect and there is not a data point that can represent every factor about a country.
However, this data is still useful and appropriate to address my research question because it contains a wide enough range of countries to examine an effect. The countries span multiple continents and have cultural and geographic factors which distinguish them, providing multiple variables which could affect the mental health of a country. The variables themselves are strong, as suicide rate is a good indicator for mental health in a country, and GDP per capita provides the best estimate we can get for how wealthy an average citizen is, barring countries with extreme wealth inequality, which is controlled for with the Gini coefficient.
Economic Theory and Econometric Model
My econometric model is as follows:
I will use multiple linear regression as my identification strategy to analyze the significance of the correlation between GDP and suicide rate as I add more control variables and see if more of the effect on mental health is due to GDP or is affected by other variables. I will look at how well my variables account for the variance in the model and make a reasoned judgement as to whether there are potential omitted variables, but I want to avoid adding bad controls which would introduce selection bias to me model. My estimation method will be ordinary least squares, using one left hand side variable and 5 right hand side variables.
The analysis will generate a coefficient for each right-hand-side variable that can be examined for significance and size. The size and sign of the 
The primary theory behind this model is that as countries become wealthier, citizens are afforded a greater standard of living. This would, in my hypothesis, improve the mental health of those citizens and reduce the number of people who take their own lives. I believe that there is a linear relationship between GDP per capita and suicide rate, although at some point of wealth there may be a carrying capacity where more money does not necessarily reduce suicides in a meaningful way.
Results
In column (5), we see that on average, for every one-thousand dollar increase in GDP per capita, a country experiences about .066 more suicides per 1,000 people when controlled for industrial output, wealth distribution, population density, and military expenditure. The direction of correlation contradicts what my states hypothesis was. Based on the body of research, I predicted that an increase in GDP would result in a reduction in suicide rates. With the coefficient being .0660067 and the standard error being .047689, the coefficient is not statistically significant at the 95% level. This results in a t-statistic of 1.38, which has a p value of .168. This estimate is also not economically significant. To raise the suicide rate by even one suicide, it would take an increase of over $15,000 in GDP per capita. To raise the suicide rate by one standard deviation, it would take a $144,000 increase in GDP per capita. This is not a realistic scenario. This estimate is moderately stable, however. As we add more control variables, the t-statistic of our estimate does not change drastically. We do see a slight jump when we add the control for population density, but this reverts once we add the military spending control. The confidence interval of the estimate is [-.0282172, .1602307], which clearly includes zero. Thus, it is possible that GDP per capita has a zero-effect on suicide rate and we can rule out a large effect.
The results in column (4) might provide an interesting insight into what could affect suicide rates. We see that, when controlled for population density, the .0645518 coefficient for GDP per capita from column (3) increases to .0718305. For context, this causes the t-statistic to increase by .12 and the p-value to decrease from .159 to .126. While not massively significant of a change, this is the variable which produces the most change in results. This suggests that population density, which is an indicator for urbanization, might have an effect on suicide rates that can be explored with further research.
Omitted variable bias will be explored with more depth in the discussion section, but it is highly likely that omitted variables influence the results of this analysis. There is a discrepancy between my analysis and the body of literature on the subject with regards to the direction or even existence of a relationship between wealth and mental health. Omitted variables are the most likely explanation for this discrepancy.
I do not believe there is as much measurement error in my analysis as there is omitted variable bias. This was a deliberate choice on my part. I chose concrete measures such as GDP per capita and suicide rate to analyze, rather than self-report or voluntary individual studies that might be measures inaccurately. All variables in my model are macroeconomic indicators that rely mostly on measurement of a country’s economy, which is hard to misrepresent.
Based on the use of controls and the size of the sample, I believe that the identified relationship between GDP per capita and suicide rate is causal. Despite zero being captured in the confidence interval, it is possible for a causal effect to be zero. This would simply mean that the true effect of GDP per capita on suicide rate is zero. Further research into omitted variables would be required to confidently differentiate between a causal relationship or a mere association between the two variables.
Discussion
There is lots of room for omitted variable bias in this analysis. There are no reliable variables, for instance, that can measure a country’s culture, history, beliefs, or attitudes. These factors can be attempted at with self-report surveys, but I strongly doubt the reliability of those in giving a valid account of the impacts of these on things like mental health. It is simply impossible to quantify certain things. There are also other variables, which we can quantify, that are not included in this analysis. I attempted to pick the most likely variables to have an effect on suicides a priori from the body of research I had reviewed. It is likely that I included variables which I ought to have omitted, and omitted variables I ought to have included. The direction of this omitted variable bias is more likely than not to be negative. The research I reviewed established a clear positive relationship between wealth and mental health, so I would expect that inclusion of the right variables would increase the coefficient on GDP per capita, if that is truly the right measure of wealth to use in the study.
Conclusion
In summary, the linear regression model of suicide rate on GDP per capita produced an estimate of .066 additional suicides per 1,000 people for every $1,000 increase in GDP per capita. This went against my hypothesized negative relationship, but does not appear to be statistically or economically significant of an estimate when put into context.
With regards to policy, these results do more to eliminate nonworking solutions than to encourage investment in working ones. Nonetheless, this improves our understanding of the ways we can approach global problems. Seeing that personal wealth on average does not have a significant effect on suicide rate, we can focus research and funding on things like medicine, education, and social programs rather than trying to increase global wealth parity.
These results lend well to future research. As mentioned in the discussion of findings, there are many potential omitted variables which could affect suicides. The ones that would interest me the most are strength of community/social bonds, education, and types of employment. The concept of a relationship between wealth and mental health should not be discarded, either. Perhaps GDP per capita is not the best variable to measure the way wealth affects a person’s mental health. The studies discussed in my literature review studied the relationship between wealth and mental health on smaller scales. I was unable to physically go out and collect data, but given the time and the resources I might conduct a survey of earnings and net worth on representative samples from all countries and compare that to different mental health indices that attempt to quantify mental health. Suicide as the measure for mental health might not give the full picture, but it is the simplest to analyze without selection bias. It might also be helpful to conduct research within groups of countries that share common traits, such as by continent or by languages spoken. This could be done with fixed effects, but I think conducting this research on a smaller scale would produce more meaningful results and give researchers more flexibility in their methodology.
Table 1—Summary Statistics
| Variables | Mean | SD | 25th percentile | 75th percentile | Observations |
| Suicides per 100,000 people | 10.44 | 9.51 | 5.35 | 12.85 | 164 |
| GDP per capita ($1000s) | 12.96681 | 17.95685 | 1.71 | 15.6 | 161 |
| Industry as percent of GDP | 28.11 | 12.40 | 19.9 | 32.5 | 160 |
| Gini Coefficient | 38.47 | 7.75 | 32.9 | 43.4 | 164 |
| Population per square kilometer | 180.47 | 643.24 | 30.8 | 133.5 | 164 |
| Military spending as % of GDP | 2.20 | 2.74 | 0.99 | 2.29 | 161 |
Notes: The subjects of this data set are 164 countries chosen based on availability of data and samples are collected from the year 2013. All data was initially found from GapMinder’s data visualization tool and then followed to its original source. All source data is listed in the bibliography.
Table 2—Joint Regression Results
| Variables | OLS Results (1) (2) (3) (4) (5) | ||||
| GDP per capita ($1000s) | .0570284 (.0419513) | .0570446 (.0423243) | .0645518 (.0456074) | .0718305 (.0467512) | .0660067 (.047689) |
| Industry as percent of GDP | -.0561884 (.0614558) | -.0556998 (.0616233) | -.0578081 (.0617838) | -.024828 (.0691071) | |
| Gini Coefficient | .0470294 (.1047288) | .0511505 (.1050386) | .0493808 (.106028) | ||
| Population per square kilometer | -.000878 (.0012024) | -.0007623 (.0012146) | |||
| Military spending as percent of GDP | -.5525636 (.5049676) | ||||
| R2 | .0115 | .0163 | .0176 | 0.0210 | 0.0266 |
| t-statistic of GDP per capita | 1.36 | 1.35 | 1.42 | 1.54 | 1.38 |
| Observations | 161 | 159 | 159 | 159 | 157 |
Notes: Table reports the coefficient of the predicted effect on suicide rate (in number of additional suicides per 100,000 people) by each variable included in the regression. Standard errors are reported in parentheses below coefficients. Coefficients marked with one asterisk (*) are statistically significant at the 10% level, those marked with two asterisks (**) are significant at the 5% level, and those with three (***) at the 1% level.
Bibliography
Carter, K. N., T. Blakely, S. Collings, F. Imlach Gunasekara, and K. Richardson. “What is the association between wealth and mental health?” Journal of Epidemiology and Community Health 63, no. 3 (March 2009): 221-226. https://doi.org/10.1136/jech.2008.079483.
Gapminder. “Gini – Data Documentation.” https://www.gapminder.org/data/documentation/gini/.
Global Health Data Exchange. “Global Burden of Disease Study 2013.” https://ghdx.healthdata.org/record/ihme-data/gbd-2013-all-cause-and-cause-specific-mortality-1990-2013.
Greenberg, CDT Samuel ’24 CO E1.
CDT Greenberg conducted a referee report of my paper and offered suggestions on grammar as well as that I should include more substance in explaining the economic theory behind my model.
Lê-Scherban, Félice, Allison B. Brenner, and Robert F. Schoeni. “Childhood family wealth and mental health in a national cohort of young adults.” SSM-Population Health 2, no. 1 (December 2016): 798-806. https://doi.org/10.1016/j.ssmph.2016.10.008.
Luthar, Suniya S. “The Culture of Affluence: Psychological Costs of Material Wealth.” Child Development 74, no. 6 (November 2003): 1581-1593. https://doi.org/10.1046/j.1467-8624.2003.00625.x.
Patel, Vikram, and Arthur Kleinman. “Poverty and common mental disorders in developing countries.” Bulletin of the World Health Organization 81, no. 8 (2003): 609-615. https://www.who.int/bulletin/volumes/81/8/Patel0803.pdf.
Stockholm International Peace Research Institute. “SIPRI Yearbook 2013.” https://www.sipri.org/yearbook/2013.
United Nations. “World Population Prospects.” https://population.un.org/wpp/.
World Bank. “GDP per capita (current US$).” https://data.worldbank.org/indicator/NY.GDP.PCAP.CD. World Bank. “Industry (including construction), value added (% of GDP).” https://data.worldbank.org/indicator/NV.IND.TOTL.ZS
Cammack Shepler 
Leave a comment