De Econometrist neemt een statistische kijk op de wereld.
Life expectancy, the average expected time for a person to live based on their year of birth, varies all over the world. People from third-world countries are inclined to have a shorter life expectancy than people from countries from the western world. This can be explained by several things, such as that people in the western world live under better conditions and have access to advanced health care. The World Happiness Report yearly publishes numbers on how happy people from certain countries are, on a scale from 0 to 10, and as it turns out, people from western countries are also happier than people from third wrold countries. Happiness makes life joyful and might grant people extra stimuli to live longer, consciously or subconsciously. Could it thus be that people who are happier and give their life a higher score on ‘the scale of happiness’, also live longer?
This article is based on a report that tries to answer the question above, made by the first year econometrics students Iris de Jong, Lisanne Herbers, Thom van Kemenade, Maarten Kant, Ettina Beiboer and Sabina Kamerling from the University of Groningen. The report used data from The World Happiness Report and data from The World Databank, described for 127 countries from all around the world.
To come up with an answer to the previously defined problem, an obvious first step was to check whether happiness and life expectancy are correlated. This is determined by applying a simple linear regression of life expectancy on the happiness score
The above results show that there is indeed a positive connection. The coefficient 5.359 means that if the happiness score of a country increases by 1 point, the average life expectancy of that country will increase by 5.359 years. This seems like a lot, but the happiness score is based on a scale that goes from 0 to 10, so an increase of 1 whole point already makes a huge difference. The constant means that if the happiness score is zero, the estimated life expectancy is about 43.21. Now, from the p-value that is below 0.01 (indicated by the three ***) shows that this is a statistically significant correlation. The R-squared of 0.576 means that the 57.6% of the variance in the variable lifeexp can be explained by the variance in happiness.
The next question that needed answering in order to draw conclusions, was which factors influence the happiness score the most. We did this because happiness is something that is quite hard to measure. With use of data from the World Happiness Report and the World Databank, we attempted to deconstruct this happiness variable and describe it in terms of easier observable variables. In order to do this, we started off with a simple regression of happiness on the variable gnipercapita (which is the Gross National Income per capita of a country in USD). Each time, we added another variable from our database to the regression and after evaluation, we left it in or took it out again. If the variable had a p-value greater than 0.1, we deemed it statistically insignificant and took it out again. The same was done if it affected the R-squared negatively, meaning it made the regression model less fitting. We also made sure the number of observations remained representable. After doing this, we had the following result:
In this way, the variable representing the happiness score could be described by the above set of other variables. These variables are of an objective nature, such as gnipercapita and ginihousehold (the gini factor is a way to measure the inequality of income within a country), or a more subjective nature, such as positive and negative (which describe the amount of positive or negative feelings people experience) and socialsupport (which measures whether people feel supported by friends and family). All these gave an R-squared of 0.808, meaning that this set of significant variables explains the happiness score for about 80.8%
What seemed like an interesting idea to us, was to pay some more attention to the influence that money has on happiness. Since GNI per capita can be taken as a measure of the average wage rate per country, we started investigating this variable in some more detail. To get a clear image of the effect of money on happiness, we created a scatterplot of happiness and gnipercapita:
As became clear from the graph, more money makes someone happier, but only to a certain extent. To show this in concrete numbers, we introduced the variable gnipercapita2, which are the values of gnipercapita, squared. Then we regressed happiness on both gnipercapita and gnipercapita2. We did not only do this for all countries, but also differentiated between ‘rich’ and ‘poor’ countries. The official partition line below which GNI per capita should be in order for a country to belong to the least developed countries is 1035 USD. However, if we used this as our measure for which countries we considered rich or poor, the number of poor countries became too low to get any significant results. That is why we chose to determine the median of gnipercapita as our measure, which turned out to be 6150 USD. So all countries with a GNI per capita above 6150 USD were the wealthier half of all our countries in the database and all countries with a lower GNI per capita belonged to the poorer half. We ran this regression for all countries in our database, for only the ‘rich’ countries and for only the ‘poor’ countries and got the following results:
Since for all categories, gnipercapita2 has a negative coefficient, the conclusion we drew from the graph is supported. These coefficients thus mean that the increase in happiness score becomes less as the GNI per capita becomes more. Furthermore, we can see that people from poor countries react a lot stronger to an increase in GNI per capita than people in richer countries.
Now that we have a grasp on which factors the happiness score is composed of, it remains to be seen whether these same ‘happiness influencers’ also have an effect on the average life expectancy. The way we have done this was by regressing the life expectancy on these happiness influencers we found in the first section. This gave the following result:
If happiness was included in this regression, the happiness influencers would already be taken into account, albeit partly. That’s why happiness was not one of the variables in the regression. The values of the coefficients in this final regression reveal that the variables that affect the score of happiness in a positive fashion also affect the life expectancy positively.
The goal of our research was to investigate the relationship between happiness and life expectancy. After doing a thorough analysis, we concluded that one can describe the life expectancy quite accurately in the following way:
Because the happiness score can mainly be explained by all variables above, we can indeed say that happy people live longer. This is also, of course, because people living under better conditions tend to be happier and give their life a higher score. However, one can not deny that trying to be as positive as possible will make life more worthwhile, however long that life may be.
This article is written by Sabina Kamerling.