The “Happiness Experiment” Report: How NOT to Measure Happiness
About a month ago, inspired by the conversation with one of my readers, I embarked on a personal research project, which I dubbed “Happiness Experiment”. The objective was to design a series of self-experiments to find out what makes me happy. I decided to start last month, with a couple of easy things, like quality of sleep, and stress levels. The experimental design was simple: to measure my sleep, stress and happiness levels daily for a month, and then look at the correlations among them. Only after a couple of weeks into actual experiment I realized that my design had some flaws: I chose the wrong scale to measure happiness.
Since the project essentially about life and its meaning, I opted for a Life Satisfaction as a proxy for happiness. In fact, Dr. Diener’s “Satisfaction with Life” questionnaire is often cited as one of the methods to measure happiness and well-being. The original scale has 5 items, but I went with four:
- In most ways my life is close to my ideal
- So far I have gotten the important things I want in life
- The conditions of my life are excellent
- If I could live my life over, I would change almost nothing
I answered each of these questions, using 10-point scale (1 = “Disagree Completely”, 10 = “Agree Completely”) at the end of each day; the life satisfaction score was computed as an arithmetic average of responses. After second week, I realized that these questions do not really reflect my happiness on that particular day; when answering them, I tend to think about my life in general. As a result, the life satisfaction score was pretty much the same every day (the chart is based on three weeks of data, averaged daily life satisfaction levels):
With the exception for a couple of spikes, my life satisfaction most of the time was around 5.8. With the standard deviation of 0.39 (!), there is no much variability: in 95% of cases, my life satisfaction level will be “locked” between 5.0 and 6.5. While from psychometric point of view, such stability is great, for my current research project this scale may not be “sensitive” enough to capture the current emotional and mental state. Just for comparison, here is distribution of average daily negative stress levels for the same time period (average stress level was 3.2, with standard deviation 1.2):
Unhappy with the happiness scale, I stopped collecting data last week. Still, by that time I already had accumulated enough data points, so I ran some analysis. The findings were somewhat encouraging. The patterns are hard to spot in the chart (I added current mood to the mix, out of curiosity):
As you would expect, there was moderate negative association between the stress levels and life satisfaction. The relationship between life satisfaction and current mood was surprisingly weak, although positive, as expected. However, the correlation between life satisfaction and sleep quality is puzzling. I have no explanation whatsoever why would sleeping better make me feel less happy. Hopefully, a more proper measure of happiness will fix this. I actually have looked into other measures of happiness, and already settled on the short version of the Oxford Happiness Questionnaire. The “Happiness Experiment” will resume in September. Stay tuned!