The Pursuit of Happiness: An Analysis of Aggregate Difference in the Distribution of Happiness in the United States

The pursuit of happiness is a not-so-subtle human obsession. We see a notable number of self-help books on how to be happy are published every year and millions of copies of these books are sold worldwide. In addition, hundreds of websites and online courses now offer time-bound happiness projects.[1] This obsession is even more vivid in the United States, where happiness is viewed by most Americans as a fundamental right, on account of the paradigmatic expression given in the 1776 Declaration of Independence of the United States: “We hold these truths to be self-evident, that all men are … endowed by their Creator with certain inalienable rights, that among these are life, liberty and the pursuit of happiness” (quoted in Haller and Hadler 2006: 170). More than half of the states in America adopted the phrase “the pursuit of happiness” in their constitutions, yet a broader public policy to enhance the collective happiness in the country remains absent. Two centuries after independence, the average happiness rating of the US is only 6.6 (out of 10) while its Gross National Product (GNP) per capita is $2,790 (in 1961 US dollars); this is just above Cuba (6.4)  which had a significantly lower GNP per capita at $516 in the same year (Ball and Chernova 2008: 502).

We often hear people say that money cannot buy happiness. Many happiness studies seem to confirm this observation, according to former Harvard University president Derek Bok (2010). Summarizing the major findings of the past happiness research, he examines how governments all over the world could use happiness research in a variety of policy areas to increase wellbeing and improve the quality of life of their citizens. Bok argues that happiness should be a prime aim of all public policies. His suggestion, however, is fundamentally different from what Thomas Jefferson intended when he wrote of the right to pursue happiness in the American Declaration of Independence in 1976. Bok wants to use the power of the state to ensure greater happiness for the greatest number of Americans and not to create a state that, as Jefferson had hoped, frees its citizens to pursue their own version of happiness. He looks at the policy implications for economic growth, equality, retirement, unemployment, health care, mental illness, family programs, education, and governance quality, among other things.

Michalos (2008) asserts that today, with the breakdown of a binding common value system, secularization, rationalization and individualization processes in the US, happiness increasingly comes to be seen as a personal goal rather than a collective goal. To the majority of Americans, the pursuit of happiness is related to the seeking of hedonic enjoyment, a subjective experience that includes “the belief that one is getting the important things one wants, as well as certain pleasant affects that normally go along with this belief” (Kraut 1979: 178). The current American notion of happiness is “something very close to an extended feeling of pleasure or an extended good mood or pleasant affect” (Michalos 2008: 355). Epistemologically, what the ancient Greek wise men meant by “happiness” (the English translation of the Greek word eudaimonia) is not the “extended feeling of pleasure” but closer to the modern notion of “wellbeing”. In Nicomachean Ethics, Aristotle stated that “living well and doing well are the same as being happy” (cited in Michalos 2008: 355). What he meant by “happiness” is not a mere hedonic enjoyment of individuals but a general human wellbeing. According to his view, the happy life of a person, or a group of people, “is a function of the actual conditions of that life and what a person or a community makes of those conditions” (Michalos 2008: 357).

There are variations in the public opinion about the two conceptions of happiness – wellbeing versus hedonic enjoyment (Waterman 1993). Individuals also differ among themselves in their assessment about how happy they are (Green and Elliot 2010). In the US, not all Americans are able to pursue happiness at the same rate due to the unequal environments prevailing in this extent; some aggregates of people are well ahead of others in this pursuit, while others require greater support to have equitable access to happiness. Although the pursuit of happiness is a “fundamental right” of all Americans, the US government does not ensure that the right is equally enjoyed by all its citizens. Making the pursuit of happiness an individual-level goal in this capitalist society where inequality is very high, the government in fact facilitates an unequal distribution of happiness among different aggregates of people. By empirically showing the aggregate difference in the distribution of happiness, we can help present strong arguments for adopting public policies that focus on enhancing citizens’ overall wellbeing – living well and doing well.

The objective of this paper is to analyze which aggregates of American people are reported to be ahead or falling behind in the pursuit and distribution of happiness. Previous happiness research has so far analyzed the effects of various determinants on happiness, but few researchers have tried to show how the effects vary among different aggregates characterized by the determinants such as socioeconomic status, health status, marital status, household type, religious participation and denominations, political and world views, age, gender, and race. This study tries to fill this gap. As such, the focus of this study is not to analyze what factors (mostly continuous variables) are associated with the increase or decrease of individual happiness levels but to convert these factors into aggregate characteristics (categorical variables) and see which aggregates have what levels of happiness.

In previous empirical analyses, researchers predominantly used two measures of happiness – objective and subjective. In the next section, I briefly discuss these two measures and indicate which measure I am going to adopt in this study. Then I analyze the existing research on happiness and draw my own hypotheses about the aggregate difference in the distribution of happiness. I discuss the results derived from the quantitative analyses of my data and finally conclude the study by indicating the policy implications of my major findings.

Measures of Happiness

There is no official happiness index in the US to measure the levels of wellbeing of its citizens. In contrast, Bhutan developed an official happiness index in 1972 and this tiny Asian country was the first in the world to declare that “Gross National Happiness” (GNH) rather than Gross National Product (GNP) would be the nation’s principal yardstick for measuring progress (Bok 2010: 1). Bhutan’s GNH Index is a multidimensional measure of happiness which is itself a multidimensional concept – “not measured only by subjective well-being, and not focused narrowly on happiness that begins and ends with oneself and is concerned for and with oneself” (Ura, Alkire and Zangmo 2012: 1). The pursuit of happiness in Bhutan is collective though it can be experienced deeply personally. So far in the West, France has announced an initiative to measure the collective wellbeing of French people, and in Britain, the prime minister’s strategy unit has prepared a study paper on the implications of happiness research for public policy. As a result, the European Commission has initiated a survey on various measures of wellbeing which is called the “Eurobarometer”.

There are variations in the measures of happiness among researchers. Many economists try to measure happiness in terms of objective economic conditions and utilities. Michalos (2008: 351-52) cites an example of this sort of research, the one conducted by Kahneman (1999) who “resurrected the hedonism from Plato’s fourth century BCE Protagoras…and Bentham’s more familiar treatise of 1789 to construct a concept called ‘objective happiness’, which is rooted in subjective experience and involves only a narrow sense of happiness.” Recent survey measures of subjective happiness are still somewhat novel to economists, although most of them agree on the validity of these data (Ball and Chernova 2008: 499). They also agree that economic measures alone cannot explain the variations in the levels of happiness among different aggregates of people.

Economist Richard Easterlin (1974), along with several other investigators, recently concluded that average levels of happiness in the US have risen very little over the past 50 years, despite substantial growth in per capita incomes. They have shown that although rich people as a whole are happier than the poor, the percentage of Americans who declare themselves as “very happy”, “pretty happy”, or “not too happy” are almost exactly the same as they were fifty years ago. This is perhaps because most Americans think that more money would make them happier, but as income rises they soon become used to their higher standard of living and feel they need even more money to live a good life. These findings challenge the objective measurement of happiness. Michalos (2008: 353) argues that it “makes no sense to measure people’s well-being without asking the people themselves what they think or how they feel.” That is why many psychologists and sociologists nowadays use a subjectivist definition of happiness, one that commonly relies on self-reports, as they believe this is a more appropriate and even necessary approach, given the view that happiness must be defined from the perspective of the person who is the final judge of what makes him or her happy (Myers and Diener 1995). Happiness is thus defined in terms of frequent positive affect, high life satisfaction, and infrequent negative affect (Diener 1984; Diener et al. 1999).

However, it is important to note that individuals may vary in their hedonic profiles; two individuals with similar happiness levels might differ in their relative levels of satisfaction with life, compared to their relative frequency of experiencing positive and negative mood states (Michalos 2008). The fact that self-reported happiness is subjective does not mean that it is unrelated to relatively more objective variables such as actual physical and mental health, living conditions, and work status. “The good life that we must want and achieve for all people is not, just a life in which people feel good, no matter how terrible their real life conditions are,” according to Michalos (2008: 351), “but one in which they feel good with the best of all reasons, because the objectively measurable conditions of their lives merit a positive assessment.”

Many studies have indeed found “high correlations” between subjective and objective measures of happiness (Ball and Chernova 2008: 499). It is nonetheless important to acknowledge my bias toward the subjective measure of happiness, which is to be linked with other relatively objective variables in the survey data, such as age, levels of education, income, work status, marital status, gender, and race.[2] Psychologists would emphasize that personality is the single most important determinant of happiness, such as extroverts are happier than introverts, and that these personality traits are passed on through DNA.[3] This paper is not concerned about psychological or biological determinism of happiness but it includes such variables as people’s world views – whether people are optimists or pessimists about the world – which may be closely related with their personality traits such as confidence and nervousness. The following section draws hypotheses related to this and other factors from the extant research.

Hypotheses Drawn from the Extant Research

One of the major findings of the existing happiness research is that good and close relationships between partners and spouses, parents and children, friends and neighbors, coworkers and associates are a main source of satisfaction with life (Argyle 1987; Veenhoven 1989, 1999; Myers 1993, 2000; Schulz 1995; Eckersley 2000; Land et al. 2001; Michalos et al. 2001). Writing in 1897, Durkheim (1966) had already showed that family life and social relationships were not only a protection against anomie and suicide but also against unhappiness. Many recent studies have also reported that married people and people with children are happier than singles; especially divorced, separated, and widowed people are significantly less happy (Myers 2000; Hayo and Seifert 2003). Moreover, Veenhoven (1983) has shown that the relation of marriage to happiness has not become weaker since the 1950s, but rather stronger. From these findings of the existing happiness research, I draw the following two hypotheses:

      Hypothesis 1: Married (or common-law) people are happier than unmarried (or single), divorced, separated and widowed people.

      Hypothesis 2: People living with children are happier than people living without children.

Social relationships grow with people’s participation in various social activities and networks. Previous research has shown that participation in religious activities has significant effects on happiness: regular church goers are significantly happier than those who seldom or never attend ecclesiastic services (Argyle 1987; Myers 2000). Moreover, people’s religious denominations have also important roles to play in their satisfaction with life. The effect of Protestantism on happiness can be drawn from Weber’s (1930) thesis that Protestants tend to develop an optimistic outlook on life and are inclined to show themselves as successful and happy persons. Thus, I construct the following hypotheses:

      Hypothesis 3: Happiness is higher among participants of religious services than non-participants.

      Hypothesis 4: Happiness is higher among the members of a religious denomination which emphasizes material success in life than those who do not belong to such religious denominations.

Research has also shown that people who view social relations and the world more pessimistically are less content with their life than those who hold a more optimistic and altruistic view of human relations and the world as a whole (Ryan and Dziurawiec 2001). Then, I suggest that:

      Hypothesis 5: Happiness is greater among those who have an optimistic world view than those who hold a pessimistic view about the world and human relations.

World views may be linked with people’s political views. Americans who hold a liberal political view, for example, have fought for years for redistributive programs that would help the sick and the needy, while politically conservative people have long maintained that social legislation often fails to help its intended beneficiaries (Veenhoven 2003). Bok (2010: 7) also mentions that “conservatives continue to extol the virtues of free markets and consumer choice in promoting the welfare of the population” and “liberals worry so much about the distribution of income in America and press so strongly for progressive taxes and expensive government programs to benefit the sick, the needy, and the unemployed.” From these observations, I assume:

      Hypothesis 6: Those who hold a liberal political view are more altruistic, hence happier, than those who hold a conservative political view.

Many researchers have shown that work and income are among the most important determinants of human wellbeing (Diener et al. 1995; Argyle 1999; Frey and Stutzer 2002; Easterlin 2003; Diener and Seligman 2004). Additionally, Easterlin (1974) argues that people care exclusively or almost exclusively about their incomes relative to the national distributions of income in the countries in which they live. Philosophically, Bertrand Russell (1996) has argued that it is through their work that people can be happy. However, by “work”, Russell meant more than just income activities. I can, therefore, say:

      Hypothesis 7: Employed people are happier than unemployed and those who are not currently in formal labor market.

      Hypothesis 8: People with higher incomes are happier than those with lower incomes.

Income and education are also closely correlated and both influence happiness. But how much does education influence happiness? According to Michalos (2008), it depends on how one defines and operationalizes education and happiness as well as the term “influence”. More precisely, if one defines and operationalizes education as highest level of formal education attained including primary, secondary and tertiary education leading to diplomas and degrees, and happiness as measured by standardized single-item or multi-item indexes, then the answer to the question is that education has very little positive influence on happiness. But education has enormous positive influence on happiness, if one defines both education and happiness more broadly. In his broad definition of education, Michalos (2008) includes formal education (leading to diplomas and degrees), non-formal education of the sort that might involve learning through course-work not connected to any diplomas or degrees, and informal education of the sort that might involve learning outside of any course-work such as from news media, works of art and culture, work-related training and experiences, social interaction and life experiences. His broad definition of happiness includes an Aristotelian eudaimonia or general wellbeing ensured by the enjoyment of goods of the mind such as wisdom, moral virtue and pleasure, goods of the body such as physical and mental health, and external goods such as wealth and adequate material resources, good parents and families, good friends, peace and security within and between communities, and well-governed communities. Additionally, Michalos (2008) thinks that the indirect influence of education on happiness is enormous. For example, leisure time creates opportunity for people to learn informally from reading newspapers and books, watching television, visiting places and so on and thus people can enjoy their leisure time more effectively which indirectly but positively influence their levels of happiness. Following Michalos’s arguments, we can then say that:

      Hypothesis 9: More educated people are happier than less educated people.

      Hypothesis 10: Those who have leisure time and opportunity to learn formally, non-formally or informally are happier than those who do not have leisure time and such opportunity to learn.

In every society, age or one’s position in the life cycle, is a socially standardized and evaluated category. In American society, youth are generally more highly evaluated than the elderly, while persons in the middle-age category are mostly involved in work, earning and family maintenance. The young and the old people get more leisure time to enjoy their life than the middle-aged people. Family and work-related responsibilities may also affect the levels of happiness of the middle-aged group. Considering these issues, several studies have shown that levels of happiness vary significantly by age in a nonlinear fashion (Michalos et al. 2001; Hayo and Seifert 2003; Christoph and Noll 2003). Therefore, we get the following hypothesis:

      Hypothesis 11: The younger, but also the older people, tend to be happier than the middle-aged people.

Older people face more health-related problems than younger and middle-aged people. But if the older people are reported to be happier than the middle-aged people, then the common-sense dictum that “a healthy life is a happy life” is dubious. Nevertheless, many studies have shown that health, or subjective health, is one of the most important determinants of happiness (Diener et al. 1998; Hagerty et al. 2001). Categorizing people in terms of health status, I formulate the following hypothesis:

      Hypothesis 12: Those who report to have an excellent or a good health also report to be happier than those who report to have a poor or less than good health.

Cummings and Jackson (2008) claim that despite considerable improvements in the status of Blacks and women over the past 30 years, racial and gender disparities in mortality and morbidity persist in the US. Using General Social Survey (GSS) data from 1974 to 2004, Cummings and Jackson (2008) explore the extent to which race, gender, and socioeconomic status converge to produce differences in self-assessed health and overall wellbeing of the American people. Furdyna et al. (2008) argue that the increased economic prominence of women complicates their wellbeing within marital relationships and that the husband-wife income disparities are more negatively associated with the marital happiness of Black women than White women. Following these arguments, we can say:

      Hypothesis 13: Women report to be less happy than men.

      Hypothesis 14: Self-assessed happiness is greater for Whites than Blacks.

All of the 14 hypotheses are formulated to test whether there is significant difference in the distribution of happiness among different aggregates of people in the US. Some continuous and ordinal variables like age, income, and levels of education are found in the extant literature to have nonlinear effects on the levels of happiness (Michalos et al. 2001; Christoph and Noll 2003; Easterlin 2003; Diener and Seligman 2004). To account for the nonlinear effects, this study uses these variables as categorical ones to characterize different aggregates of people. The division into such aggregates as young, middle and old age or low, average, above-average and high income may not intuitively register as more explanatory than their continuous forms. But, as already stated, the objective of this study is not to examine how the continuous or ordinal variables affect the levels of happiness; rather, we want to see if the aggregates are significantly different from one another in the distribution of happiness. The formation of these categories can be arbitrary and not as self-evident as race, religion or gender. However, it is not uncommon in the existing literature to create these categories in order to account for the nonlinearity issue. Another issue can be raised about the binary formation of some aggregates characterized by, for example, world views and political views. The binary categories – optimist/pessimist and conservative/liberal – do not take into consideration other possible categories such as indifferent and radical. This is a limitation of the available dataset and should not be taken as a limitation of this analysis.

Data and Methods

To test the hypotheses, I use the data from the 1998 wave of the General Social Survey (GSS) conducted by the National Opinion Research Center (NORC) at the University of Chicago (Davis and Smith 1998). The GSS is an independently-drawn national multistage probability sample of non-institutionalized, English-speaking respondents living in the United States. In GSS 1998, a total of 2,832 respondents above 18 years of age were interviewed in person using a standard questionnaire. Happiness was measured in the survey by asking respondents the question: “Taken all things together these days, would you say that you are very happy, pretty happy, or not too happy?” The three response categories from not too happy (0) to very happy (2) have been recoded to create a “happiness dummy” variable for this study with 1 = happy (very to pretty happy) and 0 = not too happy. This dummy variable is taken as the dependent variable for a logistic regression analysis on the independent variables, also an array of dummies. I have followed the existing literature (as reviewed above) to create the dummy variables and select the reference category for each of them.

Independent Variables

      Age: To assess the nonlinear effect of age on happiness, various stages of life cycle have been represented by three dummy variables: young (18-29 years), middle (30-59 years), and old (60-89 years), with “young” as the reference category.

      Marital Status: Three dummy variables have been created for marital status with “currently married” as the reference category. The “previously married” dummy includes divorced, separated, and widowed and equals 1 when other categories equal 0. Similarly, “never married” dummy equals 1 when others equal 0.

      Household Type: Two dummy variables, household with children and household without children, have been created, with “household with children” as the reference category.

      Health Status: Respondents were asked: “Would you say your own health, in general, is excellent, good, fair, or poor?” Two dummy variables have been created from the response categories: “excellent and good” and “less than good/poor” health, with latter as the reference category.

      Work Status: Four dummy variables (currently employed, currently unemployed, retired, and currently in school, keeping household or other) have been created, with “currently employed” as the reference category.

      Total Household Income: Four dummy variables – low income (first quartile with annual income below 20,000 US dollars in 1997), average income (second quartile: 20,000 – 34,999), above average income (third quartile: 35,000 – 59,999), and high income (last quartile: 60,000 and above) – have been created, with “high income” group as the reference category.

      Education: Three dummy variables (less than high school, high school diploma, and college and above) have been created with “less than high school” as the reference category.

      Leisure Time: Two separate items – frequency of reading newspapers (“How often do you read the newspaper?”) with responses ranging from 1 (every day) to 5 (never), and television hours (“On the average day, about how many hours do you personally watch television?”) with responses ranging from 0 to 24 hours – have been combined as a proxy measure of “leisure time” dummy with “no leisure” (never reading a newspaper or watching television in a day) as the reference category.

      Religious Attendance: Two dummy variables, participant and non-participant, have been created, with “non-participant” as the reference category.

      Religious Denomination: Three dummy variables (self-rated Protestant, non-Protestant and no religious denomination) have been created, with the “Protestant” as the reference category.

      World View: Two dummy variables have been created, with “pessimist” as the reference category, from the responses to the question: “On a scale of 1-7, where would you place your image of the world?” Pessimist (scale values 1-4) has been coded as 1 with 0 = optimist (scale values 5-7).

      Political View: Two dummy variables, liberal and conservative, have been created, with “conservative” as the reference category.

      Gender: Two dummy variables, male and female, have been created from respondent’s sex, with “male” as the reference category.

      Race: Two dummy variables, White and Black, have been created, with “White” as the reference category. “Other” race category has been excluded from the analysis to compare happiness levels between Blacks and Whites only.[4]

Results

Zero-order correlations of all variables show that no correlations are above .80 to warrant a multicollinearity problem (results not shown here). The overall variance inflation factor (VIF) in the normal regression analysis is found to be only 1.58, which also confirms that the model has no multicollinearity problem. The heteroskedasticity problem of all the dummy variables used in this study has been corrected using White’s correction.

Odds ratios are reported in five models in Table 1. Model 1 shows the results derived from logistic regression of subjective happiness on socioeconomic variables such as work status, income, education, and leisure time. Adding health related variable to socioeconomic variables, we get Model 2. By adding religion (religious participation and denomination) and personal view (political and world view) related variables to Model 2, we get Model 3. By further adding demographic variables such as age, marital status, and household type to Model 3, we get Model 4. The complete model – Model 5 – includes all of the variables of Model 4 plus two control variables, gender and race. For all of these variables, results are quite consistent through Models 1 to 5.