Happiness, Faith, Friends, and Fortune—Empirical Evidence from the 1998 US Survey Data

For an extensive review of the literature on the determinants of happiness, see Cummins (2000).

Education is not included in Winkelmann and Winkelmann’s study.

The happiness issue has also been specified in the Singapore National Pledge.

The estimated variance of the regression coefficients is large resulting in insignificant, inconsistent, and biased regression coefficients.

For more information, see Swinyard et al. (2001)

The importance of religion in well-being has also recently been reported by the scientists at the Duke University’s Center for Spiritually Theology and Health (2000–2007). http://​www.​dukespirituality​andhealth.​org.

For the data and detailed questionnaire, see http://​www.​icpsr.​umich.​edu/​GSS/​. The age brackets have been included as separate explanatory variables because age is a focal point in Gerdtham and Johannesson’s study. Although the lower end of AGE1 is generically mentioned in the questionnaire, the survey is aimed at the upper end (adult population). A number of other related explanatory variables such as education at the high school level, trust in each other, confidence in the US. Congress, relation with neighbors, belief in God, and the actual income brackets (as well as income in the logarithmic format, and in squared format) were originally included in the happiness function. The variables were excluded from the final model because they did not pass the model selection criteria (MSC) test. The estimated model has been chosen by the residuals based MSA tests—the Hannan–Quinn (H–Q) and Schwarz Baysian (SC) criteria. The procedure involves adding appropriate variables to a standard happiness model (e.g., that of Gerdtham and Johannesson’s) and minimizing the numerical values of the H–Q and SC criteria. Since the dependent variable is ordinal in nature, to avoid estimation problems, where appropriate, the explanatory variables have been transformed into a series of suitable dummy variables.

The results of an ordered multinomial logit technique (the distribution of e _t is a logistic function) are generally similar to those reported.

Commonly, the ordering is such that the smaller the number, the higher the level of unhappiness and contrariwise—e.g., Gerdtham and Johannesson’s study. However, the happiness data coded by the GSS do not conform to this general rule.

Toward that end, the probability of falling in the lower end (very happy = 1) goes in the opposite direction of the sign of the unadjusted estimated regression coefficient, while the probability of falling in the upper end (not too happy = 3) goes in the same direction as that of the unadjusted estimated coefficient—see Greene (2003) and Johnston and Dinardo (1997).

As is well known, in the OMPM, the estimated significant regression coefficients do not measure the marginal impact of independent variables on the dependent variable. Nevertheless, the first-order conditions [∂P(HAP)/∂explanatory variables] expressly provide unambiguous guidelines for P(HAP = 1) and P(HAP = 3). In that spirit, [∂P(HAP = 1)/∂an explanatory variable] has a sign opposite of the estimated regression coefficient, while [∂P(HAP = 3)/∂an explanatory variable] has the same sign as that of the estimated coefficient. The sign of the middle case [∂P(HAP = 2)/∂an explanatory variable] is ambiguously determined by the other two density functions. For brevity, proofs have not been provided—see Greene (2003).

Fixed individual effects appear to be present.

Some researchers have reported no significant happiness difference between divorced and people who stayed married in unhappy marriages—see for example, Institute for American Values, 2002, Does Divorce Make People Happy? Findings from a Study of Unhappy Marriages By Linda J. Waite, Don Browning, William J. Doherty, Maggie Gallagher, Ye Luo, and Scott M. Stanley. Boston Glob, May 12, 2007, reports that the US divorce rate (still highest among developed and developing nations) dropped to the lowest level since 1970. The reported rate is quite misleading because during the same time period, the marriage rate dropped, the number of couples who postponed marriage by living together (so called “marrying smarter”) increased sharply, and the number of prenuptial agreements (among materialistically endowed couples) increased substantially.

This pattern is observed for widowed in the 65 and older age brackets at which most life goals have already been reached, social pressure is minimal, and family obligations are quite negligible. Although males and females react differently to widowhood, they tend to remain active in their social network with friends, neighbors, and relatives—particularly, the activities that did not require their “significant others.” Consistently, one major source of strength throughout the process of identifying with the role of widowhood, appears to be religious beliefs, activities, and supports. For more information, see Normal Aging II—Report from the Duke Longitudinal Studies, 1970–1973, edited by Erdman Palmore.

The detailed findings are available upon request.