Happiness, language, and personality

The “pursuit of happiness” is probably one of the most iconic catchwords in modern Western societies. Originating in the American Declaration of Independence, its claim as one of man’s unalienable rights comes next to the French revolutionaries claim of “Liberté, Egalité, Fraternité”. It refers to an apparently universal longing, that, once basic material and security needs have been met, happiness will be found. However, the writers of the Declaration of Independence wisely refrained from defining what happiness was, leaving it up to the individuals to seek their own definition of happiness. Not surprisingly, though, the discussion of what happiness is, how it can be achieved, what societies (or the state’s, if you like) role should be has been ongoing since the ancient Greek thinkers, and is most likely to continue. What has changed is that next to an analytic deliberation of happiness, we now have both the means and the data to explore this concept empirically. So now we can try to measure happiness as a concept, describe the individual and/or societal levels of happiness, explore its antecedents and conditions, as well as gauging the outcomes of different states of happiness. 

Since happiness is such a fundamental concept, it touches not only on philosophy, sociology, and politics, but of course also on psychology, health, religion, work, and – not the least – market research.

This is the starting point for our project. The overall question was: what role does personality play in explaining levels of happiness? How much more can it explain compared to explanations by income, wealth, etc.? From this, the question arose: How can we use the vast amount of open-ended data to derive at some measure of personality? Analysing the data, we also observed that the relation between happiness and other variables often was far from linear – which is usually the default-assumption analysing such data. 

So why might personality matter in explaining happiness?

Personality can be construed as an individual’s enduring pattern of thought, emotion, and behaviour. A classical approach is using 5 traits as in the OCEAN-model, such as openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism. Firstly, simply judging from our own common knowledge, we can all safely assume that people who are extroverted and/or agreeable probably are happier than those who are, say, neurotic. Secondly, the relation between happiness and personality has been thoroughly researched. A simple query with “personality” in the World Database of Happiness yields 3.650 published studies and articles. Thus, among different personality traits, extraversion and neuroticism have been shown to be the most consistently and strongly related to well-being. Extraversion has been shown to predict positive affect, while negative affect is strongly predicted by neuroticism. Openness to experience, which one would expect to relate positively with happiness, has been shown in some studies to relate negatively with happiness. Moreover, studies across nations show that extroverts everywhere tend to experience a higher number of positive feelings and experience them more intensely than introverts. After all, between 10 – 18% of the observed variance of happiness can be explained by personality, as measured by the Big 5. Thirdly, personality and happiness have a common denominator in terms of marketing research. On the one hand, most of today’s advertising carries at least implicitly the message that use of this or that product contributes to your personal happiness. Happiness is the default-mode in advertising – no TV-spot, at least that the authors are aware of, ends with unhappy faces. On the other hand, personality traits differentiate between people who e.g. might be early adopters compared to laggards. Therefore, including personality into psychographic segmentations would add further explanatory power and a sharper differentiation, which in turn might be used for addressing certain consumer segments. 

Language and personality

How, then, could we extract measures of personality from open-ended questions? Using text (letters, novels, articles etc.) has been used as an indicator for a writer’s specific personality for quite some time. The reasoning behind this is that word choices reflect stable psychological processes – since personality is an individual’s enduring patterns of thought, emotion, and behaviour, and language is a systematic means of conveying thoughts and emotions. With the advent of computerised text-analysis it has therefore been possible to gauge personality traits even in large datasets. For the present analysis we applied a linguistic approach based on Linguistic Inquiry and Word Count (LIWC, pronounced “Luke”), developed by James W. Pennebaker. We are grateful to Receptiviti, which is the commercial side of LIWC, for giving us access to their database. The basic idea behind this is using both a dictionary with words rated to psychological categories, and knowledge of the linguistics of word usage. Thus, words and expressions can be used as indicators for latent sociological and psychological constructs that lie behind the mere expressions, and that could otherwise either not be investigated or only by using large and time-consuming scales (e.g. Big Five Personality Inventory). 

From material to psychological happiness

Using this, we ran all the survey’s open questions through the API supplied by Receptiviti to calculate individual scores for the Big 5 personality traits plus some other traits. Out of these, we calculated three statistical models that built upon each other. Our reasoning behind this is based on one of the main findings throughout research on happiness between nations, that the level of its population’s happiness can be explained by three types of capital: 1) Monetary capital, i.e. how much money the country earns on average and how well this income satisfies the basic needs of citizens, 2) Social capital: whether citizens can count on others, how frequently they have experienced violence, and the level of government corruption and societal trust, and 3) Psychological capital: whether people feel free, learn new things, and are able to do what they enjoy on a daily basis. Importantly, the level of the latter two types of capital, social and psychological, best predict the emotional well-being of nations. We hypothesized that the findings on the national level should also be found on the individual level. Therefore, our first model simply comprised two kinds of variables: socio-demographic (gender, education, age) plus variables indicating material wealth (income and two sets of ownership of items - more general items plus rather special items, such as air fryers or smart speakers). Except for gender (female), all other variables show a positive direction in explaining happiness. The explained variance, however, yields a modest 3.6%, indicating that socio-demographics and material wealth alone are hardly a satisfying predictor for happiness. The second model, comprising model 1 plus two variables supposed to indicate an inward, resp. outward orientation, fares hardly better, with an explained variance of only 3.7%. It is only when personality comes into play that we see a significant contribution in explaining happiness. The model is now able to explain 8.6% of the observed variance, which is close to the values reported in scientific literature. As expected, the largest factor is extraversion with a positive effect on happiness, plus to a lesser extent agreeableness. On the negative side, again as expected, neuroticism yields the largest effect, followed by openness to new experiences. 

To sum up our findings this far, we were able to derive individual scores for personality traits based on the open-ended questions. In order to stress this point, none of these open-ended questions were designed to tap into personality traits, however, they could be used to explore what lies behind the mere literal expressions. More so, these scores showed relations that are in line with relations reported in many other scientific articles. Additionally, we could show that in order to understand happiness it is imperative to also consider psychological factors. 

During our analysis we encountered one additional issue that we wanted to share. Usually, in social as well as in market research, the underlying assumption is that relations between variables are linear (to be honest, we assumed this in our models, too). This would imply that e.g. the more income one has the happier one gets. But are the rich really happier? Or does the slope of happiness decrease at a certain point, indicating a marginal utility of happiness? Robert Putnam has pointed in his seminal book “Bowling alone” where he shows that e.g. volunteering, attending club meetings or entertaining at home strongly increase the level of happiness up to a certain point, and then ease down. We have examined some of the variables used in our analysis and found a similar relationship. This holds true for both psychological and material variables. Using polynomial regression, we observed that for agreeableness and extraversion types, the level of happiness first slightly decreases the higher the score for those traits, but then picks up and rises. Neuroticism, on the other hand, shows the opposite relation. According to our modelling, a slight increase in signs of neuroticism first supports happiness, but then strongly decreases. This is in line with general findings on personality that with a healthy state of mind all traits play a certain role, and only the extreme dominance of one or the other trait becomes problematic. The same goes for income. Of course, having a low income where one can hardly make ends meet, does not contribute to happiness, on the contrary. The higher the income gets the happier people are….but only up to a certain point, after which happiness decreases significantly with higher income. 

So regarding the keys to happiness, our conclusions for marketeers and market-researchers are:

• As most relations in human life, happiness is non-linear: Find the balance

• Don’t constrain yourself with standardized closed questions - let your respondents do the talking and listen! There are powerful tools around that help you to analyse the data

• After all, it’s not the more the merrier; tangibles only get you so far. Think more about agreeableness, starting today with the person in the mirror.

Interested in hearing more?