Falling into the happiness rabbit hole: Part I

Philosophers, politicians and economists have long theorized about human happiness. Only in the latter half of the 20th Century, however, did researchers have the means to try and test those theories using large scale surveys that directly ask people about their happiness. The field of happiness research has exploded with voluminous research, its very own "Journal of Happiness and Well Being" and a sub-field of "happiness economics."

A future post will discuss the work that has been done on happiness across different countries. However, before plumbing the difficulties of measuring happiness across cultures, this post looks at some of the subtleties and difficulties of using happiness data within one country: the United States.  The more one examines the data (and the massive literature related to happiness), the more feels one has fallen down the rabbit hole with Alice.

The data used here was gathered by the World Value Survey from 2010 - 2014 (Wave 6) and is freely downloadable from their web site. The history of the survey and its methodologies are well described.

Happiness or Satisfaction?

Wave 6 had 2,232 respondents in the US who classified themselves as being "Very happy", "Rather happy", "Not very happy" or "Not at all happy." They also answered roughly 400 other questions about their age, health, income, political outlook, values etc.

One of the first difficulties we see with the happiness question is that only 10% of US respondents classified themselves as "Not very happy" or "Not at all happy." So inferring what correlates with happiness is hampered by the small number of people willing to describe themselves as unhappy!

In addition, the phrase choices of "rather happy" and "quite happy" are inherently imprecise. It's quite probable that some equivalently happy respondents interpreted these two phrases differently. Discrimination is further hampered by the fact that there are only 4 happiness categories.

The ambiguity of phrasing is seen in the relationship between the WSV happiness question and another question they asked: "All things considered, how satisfied are you with your life?" As shown in Table 1, the satisfaction question was the most highly correlated to the happiness question.  But the two aren’t perfectly correlated.  

As a quick reminder, correlation numbers range from -1 to +1. Two variables that are perfectly positively correlated (when one goes up, the other goes up by the same relative amount) have a correlation coefficient of 1; if they are perfectly inversely correlated they have a correlation coefficient of -1; if they are completely unrelated to each other, they have a correlation coefficient of 0. In looking at the correlations, it’s important to know questions are often coded going in different directions.  So for example “very happy” is coded as 1, while “I have no choice” as an answer to the “How much freedom of choice….” question is also coded as 1 (“I have complete control” is coded as 10). Thus the apparent “negative” correlation shown in Table 1 really means happier people also feel they have more control over their lives.  In general the relationships work as you might expect: happier people exhibit better health, more sense of freedom, more trust, religion and community is important to them, etc. Question codings can be found in the Codebook section at: http://www.worldvaluessurvey.org/WVSDocumentationWV6.jsp

Table 1: World Values Survey: US Correlations With Self-Reported Happiness

WVS Variable	Correlation	Question
V23	-0.61	Satisfaction with your life
V59	-0.40	Satisfaction with financial situation of household
V11	0.40	State of health (subjective)
V55	-0.39	How much freedom of choice and control over own life
V56	-0.24	Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?
V102	0.22	How much you trust: Your family
V170	0.22	Secure in neighborhood
V213	0.21	I see myself as part of my local community
V104	0.19	How much you trust: People you know personally
V4	0.19	Important in life: Family
V237	0.18	Family savings during past year
V239	-0.18	Scale of incomes
V94	0.18	Political action recently done: Any other act of protest
V145	0.18	How often do you attend religious services
V238	0.17	Social class (subjective)
V190	-0.17	In the last 12 month, how often have you or your family: Gone without needed medicine or treatment that you needed
V152	-0.17	How important is God in your life
V103	0.16	How much you trust: Your neighborhood
V9	0.16	Important in life: Religion
V95	-0.16	Self positioning in political scale

What is the difference between being "completely satisfied" and "very happy"?  Culturally some respondents might feel it’s boastful to say they are “very happy”, or feel that chasing happiness is silly and happiness is not the point of life.  But if they say they are “completely satisfied with their life” -- that is equivalent to saying all of their life preferences are being met. That seems like a pretty good operating definition of happiness.

The WVS question on life satisfaction asks respondents to grade their satisfaction across on a scale of 1 to 10.  Answers to the question are somewhat better distributed than the happiness question, with 17% of respondents in the bottom 50% of the distribution

Thus the balance of this essay will examine correlations with life satisfaction, not “happiness.”  Table 2 shows the WVS questions with the highest correlation to life satisfaction.

Table 2: World Values Survey: US Correlations With Life Satisfaction (variables in red text have much lower correlations once we control for their strong relationship to "Satisfaction with financial situation of household" below)

WVS Variable	Correlation	Question
V10	-0.61	Feeling of happiness
V59	0.55	Satisfaction with financial situation of household
V55	0.51	How much freedom of choice and control over own life
V11	-0.41	State of health (subjective)
V56	0.33	Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?
V239	0.28	Scale of incomes
V170	-0.25	Secure in neighborhood
V102	-0.25	How much you trust: Your family
V190	0.23	In the last 12 month, how often have you or your family: Gone without needed medicine or treatment that you needed
V238	-0.22	Social class (subjective)
V191	0.22	In the last 12 month, how often have you or your family: Gone without a cash income
V237	-0.21	Family savings during past year
V104	-0.21	How much you trust: People you know personally
V192	0.21	Science and technology are making our lives healthier, easier, and more comfortable
V213	-0.20	I see myself as part of my local community
V145	-0.20	How often do you attend religious services
V113	-0.20	Confidence: The police
V152	0.20	How important is God in your life
V95	0.19	Self positioning in political scale
V103	-0.19	How much you trust: Your neighborhood

Two issues arise before we can take the results of Table 2 at face value.  One issue is that correlation is not causation. Academics consistently find that income or wealth is one of the characteristics most highly correlated with self-reported happiness.  While many might assume money causes happiness (there is quite a bit of conflicting research on this), it is also possible that causation runs the other way and innately happy people are easier to get along with and thus are more likely to be hired, promoted and prosper.

Another issue is that many of the questions asked of respondents are related to each other. As noted, the most correlated question to life satisfaction (other than the equivalent happiness question) in Table 2 is "Satisfaction with financial situation of household."  Not surprisingly the echo of this relationship shows up in related topics like "Where are you on the scale of income?", “Family savings in the past year” and "How often have you gone without medical care in the past 12 months?"  Is each of these questions separately important to happy people?

The statistical technique called partial correlation attempts to correct for the fact that many questions are related to another.  Table 3 shows the strongest partial correlations between happiness and other questions after attempting to control for “satisfaction with financial situation of household.”  As expected, topics highly related to financial satisfaction (shown in red in Table 2) are much less important in Table 3.

Table 3: World Values Survey: US Data Partial Correlations With Life Satisfaction After Controlling For "Satisfaction with financial situation of household"

WVS Variable	Partial Correlation	Question
V23	-1.00	Satisfaction with your life
V10	-0.51	Feeling of happiness
V55	0.39	How much freedom of choice and control over own life
V11	-0.31	State of health (subjective)
V152	0.20	How important is God in your life
V56	0.19	Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?
V102	-0.18	How much you trust: Your family
V4	-0.16	Important in life: Family
V145	-0.15	How often do you attend religious services
V163	0.15	People over 70: viewed with respect
V9	-0.14	Important in life: Religion
V147	-0.14	Religious person
V104	-0.13	How much you trust: People you know personally
V28	0.13	Active/Inactive membership: Labor Union
V213	-0.13	I see myself as part of my local community
V33	0.13	Active/Inactive membership: Consumer organization
V161	0.13	People over 70: are seen as friendly
V211	-0.13	How proud of nationality
V49	-0.12	One of my main goals in life has been to make my parents proud
V215_07	-0.12	I see myself as part of [North America]
V25	0.12	Active/Inactive membership: Church or religious organization

Happiness for whom?

Before proudly announcing you have discovered the secret recipe that makes people happy, it’s worth asking if drivers of happiness change with social characteristics like age, gender or culture. Table 4 performs the same partial correlation exercise as Table 3, but is limited to respondents who were older than 50.  Many of the highly correlated questions are the same as the overall population.  However, as shown by the questions in Table 4 in red, older respondents’ happiness (not surprisingly) showed greater correlation with personal safety, leisure and belief that the elderly are competent.

Table 4: World Values Survey: US data For People Older than 50 -- Partial Correlations With Life Satisfaction After Controlling For "Satisfaction with financial situation of household" (variables in red are more important correlations for people older than 50)

WVS Variable	Partial Correlation	Question
V23	-1.00	Satisfaction with your life
V59	0.56	Satisfaction with financial situation of household
V10	-0.51	Feeling of happiness
V55	0.37	How much freedom of choice and control over own life
V11	-0.30	State of health (subjective)
V56	0.20	Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?
V152	0.16	How important is God in your life
V28	0.16	Active/Inactive membership: Labor Union
V163	0.16	People over 70: viewed with respect
V102	-0.14	How much you trust: Your family
V211	-0.13	How proud of nationality
V104	-0.13	How much you trust: People you know personally
V6	-0.13	Important in life: Leisure time
V4	-0.13	Important in life: Family
V192	0.12	Science and technology are making our lives healthier, easier, and more comfortable
V145	-0.12	How often do you attend religious services
V162	0.12	People over 70: are seen as competent
V9	-0.12	Important in life: Religion
V34	-0.12	Active/Inactive membership: Self-help group, mutual aid group
V5	-0.12	Important in life: Friends
V170	-0.12	Secure in neighborhood

Another important question is whether feelings of happiness are persistent. Was the respondent feeling “completely satisfied” at the time they took the survey because their team had just won the World Series? How would they have responded one year later? Much better insights would be gained by following the same individuals over time.  Ideally you would have them answer the questionnaire frequently to see how stable their feelings were and what changes in their situation caused changes in their happiness.

What did their answers really mean?

A final conclusion is that one shouldn’t make facile policy conclusions from these happiness relationships because you don’t know why people responded the way they did. Take the question that was highly related to happiness “How much freedom of choice and control over your life do you have?”  An individualist might be quick to take it as confirmation of their belief that freedom and limited government are correlated with happiness.

However the respondent who said they valued “freedom of choice and control” might believe access to publicly subsidized health care is an important contributor to that sense of control when compared to having no access to health care.  Especially since health is so highly related to happiness.  

Based on the WVS data, both individual and communal factors -- freedom, family, health, community, and religion -- seem to play a role in happiness.  Very few of the correlations in the data are particularly strong, suggesting that we are complex beings whose happiness will not come from simple policy prescriptions.

Note: As noted above, the WSV data is freely available.  If you wish to generate Tables 1-4 yourself using the R language, my code can be downloaded at: WorldValueSurveyRCode/WorldValueSurvey.r