Effect of Generation on Suicide

Photo by Dan Meyers on Unsplash

Background

As many countries face new stay-at-home restrictions to curb the spread of covid-19, there are concerns that rates of suicide may increase — or have already increased. Several factors underpin these concerns, including a deterioration in population mental health, a higher prevalence of reported thoughts and behaviors of self-harm among the masses during this coronavirus pandemic.

We decided to go deeper into the subject and we found that the global suicide rates were already on the rise, even before the pandemic struck.

The World Health Organization (WHO) estimates that each year approximately one million people die from suicide, which represents a global mortality rate of 16 people per 100,000 or one death every 40 seconds. It is predicted that by 2020 the rate of death will increase to one every 20 seconds.

Further findings by the WHO stated:

• Mental health disorders (particularly depression and substance abuse) are associated with more than 90% of all cases of suicide.
• Suicide rates today are most prevalent among the elderly population.
• However, suicide results from many complex sociocultural factors and is more likely to occur during periods of socioeconomic, family, and individual crisis. These factors, more often than not, prove to be causal for the Mental Health disorders mentioned in the earlier point.

After going through these findings and analyzing different sets of data, we realized that the rates of suicides seemed to change as we moved from people of one generation to the other generation and as the above findings dictated suicide today is most prevalent in elderly men today. Was the chart-topping suicide rates of such people due to the generation they belonged to or was there some other factor causing this trend?

This we did not know and so decided to find out the truth about whether the generation people belong to changes the likelihood of them committing suicide. This could greatly help in suicide prevention as, if we know for sure about the main cause of suicide in the masses then there can be collective and concentrated effort to help the affected people.

Datasets Used

We have used two datasets in this in-depth analysis of suicide rates:

1. Suicide Rates Overview 1985–2016 from Kaggle. We used this dataset because it has high integrity, as the datasets which were used to make this dataset belonged to the World Health Organization. Further, it covers a broad spectrum and has a host of relevant attributes. This was our primary suicide overview dataset.
2. A survey that we conducted to gather further information that was missing in the Kaggle dataset. This really helped us gather further insights into the dynamics of suicide in our society. We managed to gather a total of 129 responses with a 49:50 female: male response ratio.
3. World Happiness Report from Kaggle. We used this dataset because it has high integrity, as it was compiled by the United Nations (UN). Further, it has data regarding a good majority of countries in our other dataset.

Data Cleaning

The following steps were taken to clean dataset (1):

1. Removed the Human Development Index column, due to it having more than 70% missing values.
2. Removed the data collected in 2016, because it seemed a lot less than the data from the rest of the years i.e. 1985–2015 inclusive. This seemed like an anomaly and might have affected our results.
3. Removed countries with less than 4 years of data.
4. Attribute names and data types were changed to suit our requirements.

For the other two datasets, the columns which were not needed were dropped.

Generations Into Consideration

The generations we are taking into consideration for this project and their respective time periods are given below:

1. G.I. Generation: 1901–1927
2. Silent Generation: 1928–1945
3. Boomer Generation: 1946–1964
4. Generation X: 1965–1980
5. Millennial Generation: 1981–1996
6. Generation Z: 1996-current

EDA Divided Into 5 Parts

We have separately and independently analyzed our main causal factors for suicide:

1. Generation
2. GDP
3. Happiness
4. Gender
5. Age

Generation

It is commonly perceived that suicidal behaviors are most common in people of current generations. They are most prone to depression, loneliness, and other mental health issues. According to a report Generation Z (which is the current generation) is significantly more likely (about 27 %) than other generations to report their mental health as poor. However, at the same time, they are 37% more likely to receive mental health treatment compared to other generations. To see the actual relationship between the suicide rate and generation we analyze the data which pictures us some important insights. According to it, amongst all the generations, the suicide number is highest among the Silent and second-highest among the Boomers generation.

These results seem plausible when we consider the fact that these generations were from the time period of World War II. Silent defines people born in 1928–45; while the Boomers generation (who precedes silent) consists of individuals who were born in 1946–1964. This provokes us to further analyze the generation with the suicide rate.

Photo by The New York Public Library on Unsplash

This was obviously an unfortunate time period for all of humanity, people going to war experienced unimaginable events. Cities were destroyed, people were put into concentration camps, often separated from their families, the atom bomb was dropped in Japan which was one of the most devastating incidents anyone could ever think of. Over 50 million people lost their lives due to this war, and thus, many people lost their fathers, mothers, sisters, brothers, and friends. People were left alone and were saddened to such an extreme extent that they no longer wanted to live anymore, and unfortunately might have resorted to suicide. Not to mention the trauma of the events would remain to haunt them throughout the rest of their lives. Of course, this is not to disregard any other reasons for why anyone within that time period committed suicide.

GDP

It is a common assumption that wealth is a source of happiness. Therefore, having more money means fewer worries. Hence, in theory, an increase in GDP per capita should decrease the rate of suicide. However, our data had some interesting insights to show.

While observing the graph of year-wise suicide rate with GDP, we see that from 1995 to 2013, the total suicide rate has been decreased while the GDP per capita has been increased. Here, we can infer that increase in wealth can decrease the risk factor of suicide.

But when we look at the suicide rate of different countries with the highest suicide rates, we see different trends. For example, in the United States, as well as in the Republic of Korea, the rate of suicide increase with the increase in GDP per capita.

In countries like the Russian Federation and Germany, the rate of suicide is decreasing as the GDP per capita is increasing.

And in countries like the United Kingdom, the rate of suicide is neither increasing nor decreasing with the increase in GDP per capita.

By observing these trends, we can infer that the GDP is not the sole factor that affects the rate of suicide. For different countries, the definition, as well as the reason for suicide, is different. In some countries, there are specific factors that adversely affect the mental health of the people and hence result in an increased suicide rate for that country. But when we observe the overall trend, the rate of suicide is inversely proportional to the GDP per capita.

Happiness

One common assumption about suicide is that people living in “happier” environments are less prone to committing suicide. To look into any possible correlation, we looked at the world happiness reports, released by the United Nations from 2015 to 2019, to see if generally happier countries have a lesser suicide rate. The world happiness report weighs metrics like healthcare, GDP per capita, and freedom to give a score to each country.

Firstly, we saw the countries with the highest suicide rate and extracted their happiness values from the world happiness reports. The following figure shows the happiness rank of the top 20 countries with the highest number of suicides.

Photo by Caju Gomes on Unsplash

However, once we visualized these numbers, we realized that there is considerable variation in the rank and score of the top 20 countries. To see if this variation is present in the 20 countries with the lowest suicide number as well, we visualized the happiness ranks of the countries with the lowest number of suicides.

As can be seen in the figure above, even the 20 countries with the lowest number of suicides show a lot of variation, with Iceland being ranked number 2, while Armenia is at number 127. However, as we further analyzed the data, we found out that while the average happiness score of the top 20 countries across 5 years was 41.73, the average happiness for the countries with the lowest suicide rate was 56.66. These findings hinted that happiness does have as significant a correlation with the number of people committing suicide, so we decided to probe further, and visualized the number of suicides in the most, and the least happy countries which overlapped with the countries in the suicide information dataset.

These figures show that apart from a couple of exceptions, the number of suicides tend to be lower in the least happy countries. However, it is not because as many people in these countries do not commit suicide. Rather, there may be a host of factors explaining why the number of suicides is low in these countries.

According to an article published by BBC outlining the difficulties of data collection, it is seldom convenient to collect and compile data in underdeveloped countries, like the ones in our bottom twenty list. This is because while there is a lack of infrastructure in some countries like Kyrgyzstan, there may be cultural or contextual biases to answering survey questions in other countries like Saudi Arabia, where suicide is considered a religious taboo, and not talked about or discussed very often. Further, most of the countries with a low suicide number have a very low population (for example, Bulgaria has a population of 7 million), and thus, the number of suicides cannot be fairly compared to countries having a higher number of suicides.

One notable spike in the least happy countries is Russia. While this spike may be attributed to a relatively higher population of 144 million, the statistics regarding alcohol consumption in Russia may provide an alternate explanation. According to a report by the World Health Organization (WHO), Russia has a greater than average per-capita consumption of alcohol. A study by NCBI uncovered the correlation between high alcohol consumption and a considerable number of suicides in Russia.

On the other hand, the countries in our top 20 list are generally highly developed countries, which means they have a higher level of freedom, and its people can more freely talk about topics like suicide and mental health. This means that more data is available, and researchers have access to better infrastructure to survey and collect data. In countries with the highest number of suicides, the USA really causes a sharp spike in the graph. In fact, according to a report by the Centers for Disease Control and Prevention, the suicide rate in the USA has increased by 24% from 1999 to 2014 and is amongst the highest in wealthy countries. According to Prof. Julie Cerel, president of the American Association of Suicidology, one reason for this increase may be that the improvement in reporting standards over the years have led to this increase, as the government can not better compile data of the people who have committed suicide in a specific year. However, Prof. Julie Cerel also argues that mental health issues and preventative measures also require attention, and believes that a lack of access to therapy and the fact that mental health issues are not always covered by health insurance are further contributing to an increase in suicide rates.

Additionally, another reason for so many “developed” countries in our top 20 list may be due to the greater use of social media in these countries. As these countries have better education, GDP per capita, and connectivity services, people are more likely to use social media over the internet. According to a study published by NCBI, people who use social media are more likely to be depressed, as they have a false perception of other users. For instance, seeing an Instagram influencer living a seemingly perfect life in a house in Beverly Hills may make a minimum wage worker seem like he/she hasn’t accomplished anything in life.

Further, in countries like Japan, which rank highly in the world happiness reports, people are known to overwork at the expense of their health. In fact, the Japanese word Karoshi (過労死, Karōshi) literally translated to “overwork death.” In 2017, a Japanese media worker died after logging in more than 159 hours of overtime work in a single month. Factors like these cause people to lose out on life outside work, and a small hiccup in their professional life pushes them towards suicide.

Thus, we can conclude that while the overall happiness of a country may impact the number of suicides, a host of other factors like the work environment and alcohol consumption also contribute to the number of people who commit suicide in a given country. Further, while the suicide data is from 1985 onwards, the happiness data is from 2015 to 2019. While the happiness rank and rating didn’t vary considerably for a majority of these countries over these years, we cannot rule out the fact that the happiness ranks and scores may have been considerably different several decades ago. Therefore, going by the available data, happiness alone cannot justify the trend of global suicide rates.

Gender

While visualizing and analyzing our dataset and our survey, we found various instances where we felt the need to analyze the two genders separately, and dig deeper into finding out about how the people belonging to different genders have different approaches towards tackling their mental health issues and consequently — suicide. A prime example of this would be the following figure:

It is quite astounding how the number of suicides committed by males is so much higher than the suicides committed by females, and so much so, this trend is consistent along with each generation that we had data for.
A few reasons for this might be:

• From their early ages, men are often told things like “boys don’t cry”, and the usual social construct around the behavior of men is that they are independent, risk-taking, and strong, this wires into their brains that they should be tough. This results in most men not being open about their feelings for all their lives, thus, their feelings usually bottle up to the point that they are not able to control their feelings anymore, and unfortunately resort to suicide. A UK British Medical Journal study found general primary care consultation rates were 32% lower in men than women.
• Men tend to consume alcohol more than women when faced with a mental health crisis, and excessive alcohol consumption is a major risk factor for many people to commit suicide.

Even though the suicide rate in men is much higher, the suicide rate for females cannot be overlooked. Females around the world have also, sadly, been losing hope in their life and resorting to suicide. A few reasons for females to commit suicide would be:

• In this patriarchal society, the efforts and contributions of many women are overlooked, they are not appreciated as much as they rightfully should be. Be it socially, economically, it is unfortunate how this causes mental illness for a lot of women across the world.
• In most cases, women are the victims of harassment cases, many women are often unable to deal with the mental trauma that results due to these unfortunate events.

Our own little study

Seeing that the trend in suicide rates between men and women has shown, for every generation, that males commit more suicide, we sought out to research on why this is the case. According to a study, depression is more prevalent in women than it is in men, then why are men more prone to committing suicide than women are? Is depression not a factor that leads to it? This was quite surprising, and we could not help but to dig deeper into it and try to find what was going on here. We decided to conduct a local survey to further analyze the correlation between depression and the actual act of suicide. In this survey, we included a question on “How often do you get suicidal thoughts?” on a scale of 1–5, with 5 being “always” and 1 being “never”. The figure below shows the results of that question, and it is clearly visible how females think more about committing suicide than men do.

It is important to keep in mind here that the 1985–2016 Kaggle dataset and our survey both had almost equal representation from both men and women, this removes any possibility of having a case of Simpson’s Paradox in our statistical analysis. In the Kaggle dataset, the values for males and females were exactly on a 1:1 ratio. However, in our survey the participation from each gender was as follows:

We wanted to know if gender actually had a causal relationship with the act of suicide, for this we did an A/B test on the 1985–2016 suicide dataset. Our hypothesis was the following:

• Null hypothesis: The ratio of suicide in Females and males should be more or less the same and thus gender does not correlate with the suicide rate. The difference in the sample is due to chance.
• Alternative Hypothesis: The risk of committing suicide in males is higher than in females.

The observed statistic which is the difference in the mean number of suicides committed by males and females is 14.89. Now we will predict the statistic under the null hypothesis. For this, we create a table in which the gender column is randomly shuffled. Next, we found the simulated statistic from this shuffled table to replicate the null hypothesis and repeated this over 1000 times. A histogram of simulated statistics is then plotted along with the observed statistic in the red line. We next computed the empirical p-value and found that none of the 1000 samples resulted in a difference of 14.89 and higher.

Conclusion: We reject the null hypothesis and accept the alternative hypothesis which says that gender affects the risk and thus the rate of committing suicide.

We will now see if the thought of suicide (assumption of suicide) correlates/has a causal effect with/on suicide rate or not.

• Null Hypothesis: From all those who have suicidal thoughts, the probability that a female within our survey has suicidal thoughts is equivalent to the female committing suicide from the Kaggle data set i.e any difference between the survey and data is due to the random chance in the selection of the participants.
• Alternative hypothesis: From all those who think about suicide, the probability that a female will think of suicide is different from the probability of her actually committing suicide. Please note that the probability is calculated concerning gender committing suicide and not the population as a whole.

We used the absolute distance between the survey and true data as our test statistic. From the data set, we see the probability that it is a female who committed suicide is 0.21 while that of a male is 0.79. The probability of female thinking of suicide is 0.54 whereas that of a male is 0.45.

Now through the p-value, we will see if the difference is actually significant enough to reject the null hypothesis. For this, we use a multinomial function to make a large data i.e. 5000 values set with a binomial distribution of the Kaggle data set. We append the absolute difference of each of the 4000 data in an array and plot the histogram along with the observed statistic. Through the p-value, it is observed that none of the 5000 samples resulted in a difference of observed survey distance or higher. Thus, we reject the null hypothesis and accept the alternative hypothesis which suggests that the probability of a woman thinking of suicide is much different from the probability of her actually committing suicide.

This analysis/conclusion is very weird made us look into this strange result that while females have more suicidal thoughts than males, it is actually the males who end up committing more suicide.

This did give us a conclusion, however, we did even further research and found out that this is actually known as “The Gender Paradox in Suicide”. There is also a major difference in how the two genders attempt suicide, and the methods that females mostly use are less lethal like drug overdose, while men mostly commit suicide through hanging and Asphyxia. The figure below shows how the number of suicides committed by men and women was different across the world from the year 1985–2015.

Age

Suicide at any age is a tragedy for the individual and has a ripple effect on the individual’s family and community. While the media highlights high-school kids and youth suicides, one must not ignore the age group which is most affected by this…the older adults. As you can see in the graph, surprisingly it is the old people who commit the most suicide. Although since 1995 the rate is decreasing, the number is still much greater than any of the age groups.

The young generations’ suicide rates are already on the rise again and in a few years they will become part of the 75+ age group(higher risk of suicide), and to prevent the rise in suicide rates of this age group, we must understand the underlying reasons behind this.

We found out that there are multiple reasons which cause such high rates of suicides among the elderly.

Poor Mental Health: As we can see from the table, major depressions and therefore poor mental health is closely related to suicide. As people grow old their mental health can deteriorate due to several factors such as health, social factors, and depression because of the loss of loved ones, for example, their life partner. Moreover, with retirement nearing, many people get anxious about how they will manage their expenses after retirement. We will now discuss these factors and how they increase the risk of suicide in older adults.

Poor Physical Health: Health problems arise as a person grows old, this becomes very depressing and frustrating for some adults who just want to get out of this agony, this build-up of frustration and anxiety eventually leads to them committing suicide. Research shows that people with malignancies, HIV/AIDS, heart problems, back problems, Huntington’s disease among many others are more prone(1.5–4 times) to committing suicide than those without these impairments(Suicide as an outcome for medical disorders.

Social Factors: Now that we have seen some of the reasons as to why suicide rates are high in elder adults, we will shed some light on some other issues old adults face which might increase the likelihood of them committing suicide.

The 75+ age group goes through a rough period in this part of their lives. Many people lose their loved ones and the loss of a loved one because of suicide is even more devastating. Elders are sent to old-age homes which many consider being a prison for them. A lot of old people face financial issues at this age. All these factors together increase the chances of committing suicide.

As we can see in the first graph, the age-group trend is always the same in every generation (i.e rate of each sub-group does not overlap with any other age group, and the suicide rate increases with age). We can thus conclude that the age groups in each generation have similar underlying reasons (like those of the 75+ group) which contribute to the suicide rate. Had the reasons been different the trends would not strictly follow as shown in the graph. Therefore there must be some other factor that produces the general trend of suicide rates across the years.

Generation and The Confounding Candidates

As we’ve seen so far none of the factors that we analyzed produced significant enough evidence (on their own) to establish confounding attributes in the causal relationship between generation and suicide rate.

To discover how other factors have varied the rates of suicide across different generations, we further analyzed several other variables to know how the suicide rate has fluctuated across the different generations.

We examine how GDP per capita varies with generation. Generation is negatively correlated with GDP. As the generation grows old its contribution to GDP per capita of the country decreases and the newer generation takes over. From the graph, we can see that the G.I. generation (the oldest generation) has the lowest contribution to GDP while Generation X (current generation) has the highest contribution to GDP.

Furthermore, on analyzing the suicide rate alongside GDP in each generation we clearly observe that as the GDP of a generation decreases its suicide rate increases and this trend is significant in all the generations which indicates some correlation of generation with GDP alongside suicide.

Moving on, for further analysis, we compare Generation and gender relations alongside suicide rates. The results show that in both genders suicide rate is highest in the Boomers generation which, as discussed earlier, seems reasonable because these were individuals who witnessed the impacts of the period of the great depression and WW2.

Furthermore, analyzing the suicide rate in terms of age group within each generation shows that within the most suicide committing age group, the G.I. generation has the highest suicide rate in terms of age which seems inconsistent with what we observed from the generation versus suicide analysis. Though this was the only exception to the trend and we can possibly treat it as an anomaly.

Moving on, we can also compare happiness scores with suicide rates and how they vary across various generations. A high happiness score is commonly interpreted as a signal of less social failure and suicide rate. However, our data provide evidence against this misconception. And we can see that there is no such relation between suicide rate and happiness score. In terms of generations apparently, the suicide rate is again highest in the silent generation for all Happiness ranks.

But when we draw the correlation matrix we see that for all generation’s happiness score and happiness rank comes out to be the same which indicates that generation and happiness scores are independent and there are some other confounding factors which require further exploration

Our Take on Machine Learning

Photo by Andy Kelly on Unsplash

We now sought to develop a machine learning model that would help us find out the chances of a particular individual resorting to suicide in the future. However, we faced a few drawbacks. The results of this machine learning model would not add value to our thesis for the following reason:
From the extensive data analysis that we have done, we have found out that the people belonging to the older generations i.e. silent generation, boomer generation, G.I Generation, and Generation X are more prone to resort to suicide than the newer generations, i.e. millennial generation and Generation Z. If we predict the suicide rates for the older generations, say, 25 or 30 years down the road, the results of our model would be insignificant. This is because there will not be many people belonging to the older generations alive by then. This is due to the fact that a vast majority of people from those generations will most probably have died, keeping into consideration that the world-wide average life-span of a human being is around 70 years. An extensive visual analysis can be done using the life expectancy map shown below and it can be seen that the average life-span is actually around 70 years.

Our machine learning model will only have our dataset available for training, and according to this dataset, the suicide rates for the older generations have always been higher than the newer generations. The machine learning model’s prediction will also show the older generations to have a higher rate of suicide, whereas we just proved that there is a very high probability of these older generations having a lower suicide rate because of a lower total population.

As for the alternate machine learning model for predicting whether or not a person will resort to suicide based on their respective generation, age, gender, country, and all other attributes we consider would be unrequired because as we proved above in the individual analysis of each possible confounding factor, we can predict the likelihood of a person resorting to suicide without the need of a machine learning model.

Conclusion

All this converges to prove our primary hypothesis, that the generation is a leading factor that dictates the global trend in suicide rates across the years and we have strong evidence to conclude that the generation a person belongs has a significant effect on the possibility of that person resorting to suicide.

This really shows us how the issue of suicide is so relevant in every generation that we look into. No matter how less or more, in every generation there have been a significant number of suicides. This is a harsh reality that we all must realize as a society and work together in harmony to find solutions to bring an end to such tragical events.

“Place your hand over your heart, can you feel it? That is called purpose. You’re alive for a reason so don’t ever give up.” — Unknown

Contributors

Left to Right: Faraz Karim, Hamza Khalid, Muhammad Muzammil, Zafir Ansari + Amna Khalid & Fareeha Mohsin