Determinants of life satisfaction among race groups in South Africa

Abstract

Economic indicators, like gross domestic product per capita, are commonly used as indicators of welfare. However, they have a very limited and narrow scope, excluding many potentially important welfare determinants, such as health, relative income and religion – not surprising since they were not designed to fill this role. As a result, there is growing acceptance, and use of, subjective measures of well-being (called ‘happiness’ or ‘life satisfaction’, often used interchangeably) both worldwide and in South Africa. Happiness economics does not propose to replace income-based measures of well-being, but rather attempts to complement them with broader measures, which can be important in making policy decisions that optimise societal welfare. This paper tests for differences in subjective well-being between race groups in South Africa, and investigates the determinants of self-rated life satisfaction for each group. Using the 2008 National Income Dynamics Study data, descriptive methods (analysis of variance) and an ordered probit model are applied. Results indicate that reported life satisfaction differs substantially among race groups, with black South Africans being the least satisfied group despite changes since the advent of democracy in 1994. Higher levels of educational attainment increased satisfaction for the whole sample, and women (particularly black women) are generally less satisfied than men. As found in many other studies, unemployed people have lower levels of life satisfaction than the employed, even when controlling for income and relative income. The determinants of life satisfaction are also different for each race group: white South Africans attach greater importance to physical health, whereas employment status and absolute income matter greatly for black people. For coloured people and black people, positional status (as measured by relative income) is an important determinant of well-being, with religious involvement contributing significantly to the well-being of Indian people.

Keywords:

• happiness
• subjective well-being
• racial groups
• South Africa

JEL codes:

• I31
• D60

1. Introduction

Despite the progress made in many spheres of economic and social development since the end of apartheid in 1994, South Africa is still one of the most unequal societies in the world, with a Gini coefficient of about 0.58 (United Nations Human Development Report, 2009). Møller's (2007) study of life satisfaction in South Africa, using 2002 Statistics South Africa data, found that the improving material living standards for black South Africans were associated with increases in quality of life, although they were still significantly less satisfied with their life overall compared with coloured, Indian and, especially, white South Africans. Møller (2007) concludes that there is a close relationship between material and subjective well-being for black South Africans, not only because of the improved financial security and living conditions that material possessions bring, but also because they are associated with improved social standing. As such, ‘a well-appointed home tells the world that black South Africans are no longer second-class citizens in their own country’ (Møller, 2007:412).

In a later study, which measures quality of life among black South Africans in a lower-income suburb of Grahamstown, Møller & Radloff (2010) find that it is not only material living standards that have an impact on quality of life, but also symbolic progress (‘sense of place and belonging’), explored through the willingness of participants to change the name of Grahamstown to ‘Rhini’. In addition, life satisfaction, especially among black South Africans, is substantially affected by the provision of services (Møller & Jackson, 1997) as well as place of residence (Møller, 2001).

In a recent study of the difference between needs and wants in quality of life indicators in developing countries, McGregor et al. (2009) point out that the significance of variables may differ for people living in different contexts. For example, those living in rural areas judge access to food and electricity as more important to their quality of life than those who live in cities, because food and electricity are readily available in urban communities. Likewise, communities with much migrant labour list good family relations as more important than communities where most people work close to home (McGregor et al., 2009:148). These findings suggest that, for development policies to be effective, differences between communities need to be taken into account.

An understanding of the determinants of life satisfaction for different race groups in South Africa is important as the information can be used to design future policies aimed at improving well-being. A number of studies on quality of life in South Africa test which race group is happiest in comparison with others (Powdthavee, 2003; Hinks & Gruen, 2007; Mahadea & Rawat, 2008). Race is included as a control variable, but the determinants of happiness for each group are not explicitly analysed. There is thus a distinct lack of research in South Africa on the relationship between happiness and race as well as the determinants of subjective well-being for each race group. Sixteen years after the democratisation of South Africa, it can reasonably be expected that happiness differences among race groups should be less substantial than before the 1994 elections (Powdthavee, 2003).

Another important reason to examine the differences between race groups is that social cohesion is a determinant of how well a democracy functions in the long run (Kunene, 2009). A society in which large material and welfare differences exist between social groups is unlikely to be one in which citizens feel solidarity with each other or pursue common goals (Chipkin & Ngqulunga, 2008). From both short-term and long-term developmental perspectives, therefore, welfare differences across race groups, and their determinants, are important.

2. Theoretical framework and previous research

The following section gives a brief review of social cohesion theory as the overarching theoretical framework of the paper. Previous research is described, with particular emphasis on the determinants of subjective welfare ratings by various social groups.

2.1 Social cohesion theory in a South African context

Social cohesion theory argues that, for a democracy to work well and achieve stability, a society needs to be united by common values or goals, which generate feelings of solidarity (Kunene, 2009). Early theorists assumed that this would necessarily mean that more culturally (ethnically, racially, or linguistically) homogeneous societies would be needed to achieve long-lasting social cohesion (Jenson, 2010). Chipkin & Ngqulunga (2008) argue, however, that a shared understanding of values and moral codes can potentially overcome cultural diversity – as may be happening in the South African case, where the constitution is seen as the unifying factor.

Social cohesion, a feeling of empathy and solidarity with fellow citizens, is necessary if choices on how society should allocate public resources, and the trade-offs they imply, are to be agreed on. As Chipkin & Ngqulunga (2008:63) put it: ‘one might be prepared to delay the realisation of one's needs for the benefit of another if one feels a sense of solidarity or empathy for them’. In addition, the more unequally society's resources are divided, the less likely is social cohesion because the wealthier members feel over-taxed for little benefit and the poorer members feel trapped in a cycle of poverty.

Jenson (2010) defines two major dimensions of social cohesion: social cohesion as social inclusion, involving social policy and active citizenship; and social cohesion as social capital, referring to the strength of social relationships and interaction in society. Both Chipkin & Ngqulunga (2008) and Kunene (2009) argue that while social capital indicators, like feelings of belonging, citizenship commitment, voter participation and confidence in the future, are relevant, concrete or ‘positive’ factors directly affecting quality of life are more significant in a developing country like South Africa, coming out of the racially divisive period of apartheid. For example, Kunene (2009) assesses South African social cohesion through indicators such as income distribution, employment and unemployment, housing quality, gender equality, household debt, education levels, and so on. Chipkin & Ngqulunga (2008) explore the concept through research into the socio-economic distribution of violent crime and gender-based violence.

These models assume that development indicators have a positive relationship with overall quality of life and welfare. Thus, more education, employment, housing, and so forth, will result in increased quality of life. Development economists have a similar line of thinking, using indicators such as the World Bank's ‘Development Diamond’ (including life expectancy at birth, school enrolment, access to safe water, and gross national product per capita) to track and compare progress (World Bank, 2012).

In a South African context, a study by Posel & Hinks (2011) investigated variations in levels of trust of neighbours and strangers in different race groups, using the NIDS (2008) data. Trust was measured by asking respondents the following questions: imagine you lost a wallet or purse that contained R200 and it was found by someone who lives close by/a complete stranger. Is it very likely, somewhat likely or not likely at all to be returned with the money in it? Given the high levels of unemployment and crime in South Africa, the authors were not surprised to find generally low levels of trust, even among neighbours, although respondents were more likely to trust neighbours than strangers. They also found that white people were far more trusting of their neighbours than black people, but that these differences were considerably reduced when controls for income and neighbourhood characteristics were included. This demonstrates that the socio-economic situation in which people find themselves can have a significant impact on levels of trust and social cohesion.

This study takes the analysis further by investigating the marginal impacts of development indicators on subjective welfare, as well as commenting on subjective welfare differences between ethnic and gender groups in South Africa. Overall welfare shows improvement, but statistically significant differences between racial groups could indicate an ongoing threat to social cohesion and societal stability.

2.2 Conceptual clarification of terms and determinant variables

Interest in examining the causes of, and reasons for, happiness has become popular in economics in past decades, as can be seen by the exponential increase in literature regarding life satisfaction (Clark et al., 2008). Veenhoven (1991) views happiness as the degree to which an individual judges the overall quality of his or her life as favourable. To increase the happiness of all members of society, therefore, factors that contribute to happiness need to be better understood (Veenhoven, 1996). As mentioned earlier, although traditional economic indicators such as gross domestic product have long been employed as indicators of well-being, there is growing consensus that such measures are inappropriate for assessing individual levels of well-being and progress, since important non-monetary measures are not taken into account (Bleys, 2012; Natoli & Zuhair, 2011; Tsai, 2011). Analyses of subjective well-being data are ideally suited to study progress made in enhancing the well-being of individuals.

Some researchers distinguish between the terms ‘happiness’, ‘well-being’ and ‘life satisfaction’, but others use the terms interchangeably (Veenhoven, 1996; Schyns, 1998; Posel & Casale, 2011). For example, Schyns (1998) found a high correlation between mean happiness and mean life satisfaction, and suggests that happiness and life satisfaction are very similar concepts. As such, this study uses these terms interchangeably.

Happiness levels can change significantly in response to many different factors (Schyns, 1998; Hagerty & Veenhoven, 2003). Happiness surveys can examine the effects of non-income factors, such as education, race and gender, as well as more traditional economic indicators, such as inflation, real gross domestic product and price stability (Frey & Stutzer, 2002).

In terms of determinants, many studies have found unemployment to be negatively associated with subjective well-being (Clark & Oswald, 1994; Oswald, 1997; Ravalion & Lokshin, 2001; Stutzer, 2001; Powdthavee, 2003; Møller & Radloff, 2010). Graham (2008) notes that unemployment has a greater negative effect in countries where there are no measures to counter the effects of unemployment. Clark & Oswald (1994) found that unemployment has a greater effect on happiness than income. In a South African context, Hinks & Gruen (2007) found a negative relationship between happiness and unemployment, and Møller & Radloff (2010) conclude that, despite improved living conditions, lack of income and employment resulted in negative perceptions of a household's situation.

The relationship between age and well-being is generally non-linear and ‘U-shaped’ in nature (Clark et al., 1996; Blanchflower & Oswald, 2000; Ferrer-i-Carbonell & Gowdy, 2007), suggesting that happiness is high at a young age, decreases over time until it reaches the lowest level of happiness (between 30 and 50 years of age) and then increases again. For South Africa, Powdthavee (2003, 2005) found a significant U-shaped relationship between happiness and age, but Hinks & Gruen (2007) found no evidence of this U-shaped association.

Education and subjective well-being are mostly positively related (Diener et al., 1993; Oswald, 1997; Blanchflower & Oswald, 2000; Møller, 2007). The positive relationship between well-being and education level is generally attributed to the higher income, productivity, and social status brought about by more education (Witter et al., 1984). More education, however, may lead to greater aspirations, and if these aspirations are not met can lead to dissatisfaction (Clark & Oswald, 1994; Diener et al., 1999). Powdthavee (2003) found the same in South Africa and reported a negative relationship between happiness and education. Hinks & Gruen (2007), in contrast, found a weak positive association between happiness and education in South Africa, while Mahadea & Rawat (2008) report no significant relationship between years of completed education and happiness in South Africa.

Marital status is another important determinant of life satisfaction found mainly in research conducted in the developed world. Married people are consistently found to be more satisfied than those who are divorced, separated, single or widowed (Clark & Oswald, 1994; Easterlin, 2001; Layard, 2006). In addition, cohabitants generally report higher levels of happiness than those who are single (Dolan et al., 2008:106). Powdthavee (2003) and Hinks & Gruen (2007) found no significant relationship between happiness and marital status in South Africa, but Møller (2007) found a weak positive relationship, suggesting that well-being was marginally greater among married individuals when compared with other marital status groups. However, in a later study, Powdthavee (2005) found that South Africans in civil marriages were significantly happier than people who were single.

Men and women generally report different levels of subjective well-being. Clark & Oswald (1994) and Blanchflower & Oswald (2000) found that women are happier than men, while Stevenson & Wolfers (2009) found that, despite improvements in their lives, the happiness of women has declined relative to that of men in the USA. In contrast, Graham (2008) found no significant happiness differences among gender groups in Latin America. In South Africa, both Hinks & Gruen (2007) and Mahadea & Rawat (2008) report no significant happiness differences among gender groups.

Health is an important determinant of well-being. Those in good health generally report a higher level of life satisfaction than those with poorer health (Veenhoven, 1996; Gerdtham & Johannesson, 2001). The importance of health for higher levels of happiness is consistent in both developed and developing nations, and is the most important factor affecting well-being in Latin America (Graham, 2008). Gerdtham & Johannesson (2001) found a positive relationship between health and happiness using data from Sweden.

Although the impact of religion is an under-researched area (Peterson & Webb, 2006), some studies show that individuals who view religion as important in their lives are more satisfied and report higher levels of well-being than those who attach no value to religion (Ferriss, 2002; Rule, 2007). For South Africa, previous research suggests a positive association between the importance of religion and well-being (Rule, 2007).

Children and happiness are related, but the results are mixed (Dolan et al., 2008). Some studies find that having children exhibits a positive relationship with happiness (Haller & Hadler, 2006), while others find a negative association between children and happiness in, for example, single parents (Frey & Stutzer, 2000), poor families (Alesina et al., 2004) and divorced mothers (Schoon et al., 2005).

Individuals living in different geographic locations report different levels of well-being. For instance, Gerdtham & Johannesson (2001) and Graham & Felton (2006) found that people residing in rural areas were more satisfied with their life than those living in urban areas, in Sweden and Latin America, respectively. In South Africa, however, Møller (2001) found that urban dwellers were more satisfied than those living in informal or rural areas.

The positive relationship between subjective well-being and absolute income is one of the best known findings in the literature (Schyns, 1998; Graham, et al., 2004; Blanchflower & Oswald, 2004). The effect of income on well-being is greater in developing countries than in developed countries; once a certain income threshold is reached, higher income does not facilitate higher levels of life satisfaction (Clark et al., 2008). The positive association between absolute income and happiness has been confirmed in the South African case (Powdthavee, 2003; Hinks & Gruen, 2007; Møller, 2007; Mahadea & Rawat, 2008).

Studies including relative income suggest that individual well-being is strongly affected by positional status in society (Clark et al., 2008; Posel & Casale, 2011) with the relationship between relative income and well-being dependent on the reference group (Ferrer-i-Carbonell & Gowdy, 2007). Clark et al. (2008) suggest that an increase in income relative to others raises the level of well-being of that individual. People care about their status in society. This is also the case in South Africa, with Powdthavee (2003) and Posel & Casale (2011) reporting higher relative income associated with higher levels of well-being.

2.3 Previous research on subjective well-being and race

Previous research has shown that happiness differs substantially between race groups in both developed and developing countries (Graham, 2005). For example, in both the USA and Latin America, studies have found black people to be the least happy race group (Oswald, 1997; Hughes & Thomas, 1998; Di Tella et al., 2001; Graham, 2005).

Using data from the 1993 Southern African Labour and Development Research Unit survey, Powdthavee (2003) found black people to be the least happy race group, which he attributes to the apartheid legacy of South Africa. In another national study using 2002 data, Møller (2007) confirmed this finding. There is some evidence of progress, however. Harris (2007) conducted a study on the changes South Africans experienced with the end of apartheid and found that black people were happier at the advent of democracy than they were before. However, the percentage of black people that reported being relatively happy was much lower than that of the other race groups.

Smaller regional South African studies have produced similar findings: From studies conducted in Durban, Hinks & Gruen (2007) found that whites were happiest followed by Indians and coloureds. In Pietermaritzburg, Mahadea & Rawat (2008) found that black people were happier than coloured people were but that whites remained the happiest.

3. Data and method

The data were obtained from the NIDS, which was conducted by the Southern African Labour and Development Research Unit in 2008 (NIDS, 2008). NIDS is a nationally representative survey examining 7300 households carried out by approximately 300 fieldworkers. The aim of NIDS is to examine income, consumption, expenditure and well-being over time. NIDS is the most recent dataset to contain a question on life satisfaction and was therefore chosen for this study. The question regarding life satisfaction in the survey states: Using a scale of 1 to 10 where 1 means very dissatisfied and 10 means very satisfied, how do you feel about your life as a whole right now?

The study's analysis comprises both descriptive and regression based methods. In the descriptive section, analysis of variance and median tests are conducted. The former tests for the equality of mean life satisfaction between race groups, while the latter tests whether reported median life satisfaction between race groups is equal.

Given the ordinal nature of happiness, the common method for estimating happiness equations is to adopt an ordered probit model (Gerdtham & Johannesson, 2001; Hinks & Gruen, 2007). The following model is estimated:

where y represents reported happiness at the ith scale, β _i refers to the coefficients to be estimated, X _i represents the vector of explanatory variables and ϵ _i represents the error term.

The explanatory variables, informed by the literature review, include race, age, age squared, education, marital status, gender, health, religion, children, absolute income, employment status, relative income, and area of residence. Race is classified into four categories; black (base), white, Indian and coloured. Age refers to the age of the respondent. Age squared is included to test for the presence of non-linearity in the relationship between age and happiness, as suggested by previous studies. Education is divided into four categories: ‘no schooling’ (base), ‘primary school’, ‘secondary school’ and ‘post-secondary school’. Marital status is classified into five categories: ‘single’ (base), ‘married’, ‘cohabitant’, ‘widowed’ and ‘divorced/separated’. Gender is classified in two categories: male (base) and female. The health variable measures the individual's own assessment of current health and consists of five categories; namely, ‘poor’ (base), ‘fair’, ‘good’, ‘very good’ and ‘excellent’. The religious importance variable is measured through questioning the importance of religion in an individual's life, and is separated into four categories, ranging from ‘not important at all’ (base), ‘unimportant’, ‘important’ to ‘very important’. The children variable refers to the respondent's number of children. Absolute income is net income per month as a logarithm. Employment status consists of two categories, namely unemployed (base) and employed. To measure relative income, individuals were asked to classify their household income in comparison with other households in their area. Relative income is classified into five categories; ‘much below average income’ (base), ‘below average income’, ‘average income’, ‘above average income’ and ‘much above average income’. Area of residence is classified as rural (base) and urban.

Five ordered probit regressions were estimated. The first regression includes the entire sample while the remaining regressions report results for each individual race group. In terms of regression diagnostics, the Pseudo-R² and Wald chi-squared tests are used. The former is a standard measure of goodness of fit, while the latter reports the joint significance of the explanatory variables in explaining the variation in the dependent variable.

4. Results and discussion

Summary statistics are presented in Table 1.4 Mean education is substantially higher among whites, with roughly 11 years of completed education on average. Blacks and coloureds possess the lowest levels of average completed education. Women make up about 61% among black, coloured and Indian people in their respective samples, while white women make up 56% of the white sample. Reported health status is greater among whites, while religious importance is greater for coloureds and Indians. White people possess by far the highest level of mean absolute income (R8503.64); average income is far lower for the black (R2520.53) and coloured (R2416.31) groups. Unemployment is most prevalent among the black population group, with about 76% of black people in the labour force unemployed. Whites reported the highest levels of positional status followed by Indians, with blacks having the lowest levels of perceived relative income. Finally, roughly 41% of black individuals reside in urban areas, compared with 89% of white individuals.

Table 1: Summary statistics, by race group

	Mean	Standard deviation	Minimum	Maximum
Age
Black	39.53	17.24	18	105
Coloured	41.53	15.61	18	94
Indian	43.04	15.27	18	78
White	48.64	16.46	18	87
Education level
Black	1.47	0.87	0	3
Coloured	1.55	0.78	0	3
Indian	1.93	0.79	0	3
White	2.43	0.52	0	3
Marital status
Black	2.46	1.71	1	5
Coloured	2.96	1.78	1	5
Indian	3.96	1.57	1	5
White	4.04	1.50	1	5
Gender
Black	0.61	0.49	0	1
Coloured	0.61	0.48	0	1
Indian	0.61	0.49	0	1
White	0.56	0.50	0	1
Health status
Black	3.45	1.28	1	5
Coloured	3.49	1.26	1	5
Indian	3.30	1.27	1	5
White	3.72	1.08	1	5
Religious importance
Black	3.23	0.79	1	4
Coloured	3.61	0.62	1	4
Indian	3.57	0.71	1	4
White	3.37	0.81	1	4
Number of children
Black	1.49	2.00	0	12
Coloured	1.47	1.87	0	13
Indian	1.21	1.61	0	8
White	1.21	1.57	0	8
Absolute income (imputed, in Rands)
Black	2 520.53	4 068.79	20	90 000
Coloured	2 416.31	3 039.05	55	42 000
Indian	7 468.41	7 978.63	920	40 000
White	8 503.64	8 175.03	100	70 000
Employment status
Black	0.24	0.43	0	1
Coloured	0.40	0.49	0	1
Indian	0.33	0.47	0	1
White	0.40	0.49	0	1
Perceived relative income
Black	2.32	0.95	1	5
Coloured	2.56	0.85	1	5
Indian	2.80	0.80	1	5
White	2.97	0.76	1	5
Geographical location
Black	0.41	0.49	0	1
Coloured	0.79	0.41	0	1
Indian	0.65	0.48	0	1
White	0.89	0.32	0	1
Notes: Education = 0 if ‘none’, 1 if ‘primary’, 2 if ‘secondary’, 3 if ‘post-secondary’; marital status = 1 if ‘single’, 2 if ‘cohabiting’, 3 if ‘widowed’, 4 if ‘divorced/separated’, 5 if ‘married’; gender = 1 if female, 0 otherwise; health status = 1 if ‘poor’, 2 if ‘fair’, 3 if ‘good’, 4 if ‘very good’, 5 if ‘excellent’; religion = 1 if ‘not at all important’, 2 if ‘unimportant’, 3 if ‘important’, 4 if ‘very important’; employment status = 1 if unemployed, 0 otherwise; relative income = 1 if ‘much below average income’; 2 if ‘below average income’; 3 if ‘average income’; 4 if ‘above average income’; 5 if ‘much above average income’; geographical location = 1 if ‘urban’, and 0 if ‘rural’.

Table 2 illustrates the reported satisfaction levels for people across the different race groups, with the Pearson chi-squared test indicating that the relationship between happiness and race is statistically significant (p < 0.001). Of the total sample of black respondents, about 61% report a level of satisfaction of 5 or less while only 21% of white respondents report a level of satisfaction of 5 or less. Only 7% of black respondents report a satisfaction level of 10, while the majority of blacks and coloureds report a satisfaction level of 5. This is different from results for whites and Indians where the majority of both groups report a satisfaction level of 8. Consistent with the South African study by Hinks & Gruen (2007), therefore, the majority of black South Africans report the lowest levels of well-being.

Table 2: Reported life satisfaction (%), by race group

Satisfaction level of life currently	Black	Coloured	Indian	White	Total
Satisfaction scale: 1	8.83	2.86	3.54	0.59	7.33
Satisfaction scale: 2	6.30	1.65	1.01	1.19	5.22
Satisfaction scale: 3	10.93	3.13	2.53	1.66	9.03
Satisfaction scale: 4	15.86	7.86	5.56	4.16	13.67
Satisfaction scale: 5	19.27	19.89	19.19	13.66	18.97
Satisfaction scale: 6	13.17	13.30	11.11	12.59	13.12
Satisfaction scale: 7	10.09	15.16	21.21	18.29	11.54
Satisfaction scale: 8	6.49	14.34	24.75	29.45	9.52
Satisfaction scale: 9	2.40	5.27	2.02	9.26	3.29
Satisfaction scale: 10	6.67	16.54	9.09	9.14	8.33
Total	100.00	100.00	100.00	100.00	100.00
Number of observations	9 307	1 820	198	842	12 230
Pearson χ^2 (27)	1 500.0***
Mean satisfaction level	5.04	6.60	6.54	7.07	5.44
Note: ***p < 0.001.

The analysis of variance conducted indicates that mean life satisfaction is significantly different across all race groups (F = 394.53, p < 0.001). The median test also shows that median satisfaction among race groups is not equal (p < 0.001). The majority of black respondents report a satisfaction level that is lower than the median of the entire sample, while the majority of people in all other race groups report a level of satisfaction that is higher than the overall median. White people have the highest mean level of life satisfaction followed by coloured and then Indian people. Black people report the lowest mean level of satisfaction, as is displayed in Table 2. The descriptive statistics in Table 2 provide firm evidence of distinct well-being differences between race groups.

The ordered probit regression results are reported in Table 3. For all models, the Wald chi-squared statistic indicates that the explanatory variables are jointly significant in explaining the variation in life satisfaction. The Pseudo-R² values hover between 3 and 11%, consistent with other cross-sectional happiness studies employing an ordered probit model (Powdthavee, 2003).

Table 3: Ordered probit regression results (dependent variable = life satisfaction)

	Overall sample	Black	Coloured	Indian	White
Age	–0.0190 (0.0039)***	–0.0219 (0.0044)***	–0.0046 (0.0102)	–0.0971 (0.0403)**	–0.0321 (0.0169)*
Age squared	0.0002 (0.0000)***	0.0002 (0.0000)***	0.0001 (0.0001)	0.0011 (0.0004)***	0.0005 (0.0002)***
Primary education	0.1181 (0.0367)***	0.0907 (0.0402)**	0.2266 (0.0941)**	0.3639 (0.3723)	–0.7239 (0.9691)
Secondary education	0.1734 (0.0386)***	0.1357 (0.0443)***	0.2816 (0.0957)***	0.6025 (0.4326)	–0.2596 (0.6526)
Post-secondary education	0.1936 (0.0522)***	0.1305 (0.0643)**	0.3584 (0.1365)***	0.1942 (0.4882)	–0.1740 (0.6565)
Cohabitation	0.0065 (0.0383)	0.0197 (0.0423)	–0.1240 (0.0947)	–0.1487 (1.2799)	0.1052 (0.2465)
Widowed	–0.0073 (0.0459)	0.0349 (0.0521)	–0.1423 (0.1218)	–0.4634 (0.3783)	–0.1770 (0.2224)
Divorced/separated	–0.1086 (0.0688)	–0.1383 (0.0921)	–0.0061 (0.1498)	–0.2048 (0.4614)	–0.1422 (0.2054)
Married	0.0644 (0.0304)**	0.0702 (0.0347)**	–0.0381 (0.0776)	0.4587 (0.3223)	0.1029 (0.1537)
Female	–0.0745 (0.0295)**	–0.1179 (0.0341)***	0.0249 (0.0749)	0.1523 (0.2297)	0.1767 (0.1221)
Health: fair	0.1829 (0.0480)***	0.1466 (0.0534)***	0.2575 (0.1230)**	0.0852 (0.3368)	0.7255 (0.2757)***
Health: good	0.2063 (0.0453)***	0.1582 (0.0508)***	0.2972 (0.1178)**	0.1819 (0.3114)	0.8983 (0.2586)***
Health: very good	0.3342 (0.0471)***	0.3034 (0.0532)***	0.3629 (0.1205)***	0.0505 (0.3378)	1.0942 (0.2614)***
Health: excellent	0.2323 (0.0484)***	0.1720 (0.0549)***	0.2628 (0.1223)**	0.2851 (0.3229)	1.2439 (0.2673)***
Religion: unimportant	0.0125 (0.0715)	0.0062 (0.0778)	–0.1060 (0.2645)	1.7867 (0.6047)***	0.0783 (0.2870)
Religion: important	0.1047 (0.0602)*	0.1127 (0.0649)*	0.1629 (0.2301)	1.2687 (0.4640)***	–0.0590 (0.2548)
Religion: very important	0.3792 (0.0613)***	0.4011 (0.0667)***	0.3122 (0.2286)	1.6255 (0.4512)***	0.2393 (0.2547)
Children	0.0152 (0.0079)*	0.0197 (0.0089)**	0.0037 (0.0208)	0.1522 (0.0596)**	–0.0141 (0.0405)
Absolute income	0.0609 (0.0191)***	0.1138 (0.0257)***	–0.0350 (0.0429)	0.4884 (0.1755)***	0.0939 (0.0730)
Employed	0.3253 (0.1429)**	0.6789 (0.1871)***	–0.2894 (0.3167)	3.9998 (1.4781)***	0.5893 (0.6379)
Relative income: Below average	0.3013 (0.0335)***	0.3153 (0.0363)***	0.2920 (0.0931)***	–0.1770 (0.3368)	–0.0832 (0.3315)
Relative income: Average	0.7857 (0.0346)***	0.7938 (0.0379)***	0.7886 (0.0947)***	0.5610 (0.2868)**	0.3242 (0.3324)
Relative income: Above average	1.0399 (0.0502)***	1.1585 (0.0579)***	0.7765 (0.1310)***	0.5650 (0.4908)	0.4519 (0.3317)
Relative income: Much above average	1.2630 (0.0898)***	1.2470 (0.0934)***	1.4212 (0.3309)***	1.1080 (0.8603)	1.0891 (0.6510)*
Urban	0.0583 (0.0240)**	0.0577 (0.0262)**	–0.0582 (0.0768)	0.1619 (0.2297)	–0.0037 (0.1547)
Coloured	0.5424 (0.0325)***
Indian	0.3840 (0.0772)***
White	0.4720 (0.0407)***
Observations	9437	7137	1514	156	630
Wald χ^2	2 384.2***	1 304.4***	182.3***	95.7***	103.1***
Pseudo R^2	0.0601	0.0444	0.0312	0.1097	0.0429
Notes: Robust standard errors are shown in parentheses. ***p < 0.001, **p < 0.05, *p < 0.10.

Results from the entire sample confirm the existence of life-satisfaction differences among race groups. Coloureds, whites and Indians are all significantly happier than blacks (p < 0.001). Post-estimation chi-square tests indicate that coloureds are happier than Indians (p < 0.05) but not significantly happier than whites (p = 0.1598). There is also no significant difference between Indians and whites (p = 0.1954). Powdthavee (2003, 2005) and Hinks & Gruen (2007) also found black individuals to be the least satisfied race group in South Africa. The results are consistent with findings by Blanchflower & Oswald (2004) for the USA. What the results of this paper suggest is that, in spite of democratisation in 1994 and national policy favouring development as well as employment provision for black, Indian and coloured people, black people remain the least satisfied race group in South Africa while white people are the happiest.

There is a significant U-shaped relationship between age and life satisfaction in the overall sample as well as for all race groups, except coloureds. In the overall sample, satisfaction decreases until the age of roughly 42 years, after which individual satisfaction starts to increase. The turning points for blacks, Indians and whites are 46, 35 and 44 years, respectively. A possible explanation for this is the prevalence of HIV/AIDS among adults in South Africa, which has resulted in a significant decrease in the black African population between 35 and 39 years old (South African Institute of Race Relations, 2009:11). The existence of a U-shaped relationship between well-being and age is in line with the findings of international studies such as Clark et al. (1996) and Gerdtham & Johannesson (2001), as well as the South African studies of Powdthavee (2003, 2005). Hinks & Gruen (2007), however, found no evidence of a U-shaped relationship in South Africa.

Reported well-being increases with education in the overall sample. Being educated at post-secondary level contributes the most to well-being levels relative to those who have no education (p < 0.001). Similar findings are found from evidence in developed countries (Diener et al., 1993). Educational attainment was, in general, not a significant determinant of well-being for white and Indian people. For black and coloured people, life satisfaction was strongly positively associated with educational attainment.

The results need to be interpreted within the context of educational attainment levels by race group in South Africa. According to the South African Institute of Race Relations 2008/9 Survey (South African Institute of Race Relations, 2009:381–3), the percentage of whites 20 years and older with no schooling is less than 1%, and is 3.4% for coloureds, while 11.5% of blacks are still in this position (much improved from 17.4% in 1997, however). Similarly, while most whites (73.2%) have completed Grade 12 and 16.8% go on to complete higher education, only one-quarter (26%) of Africans complete high school and only 1.8% higher education. In this context, it is not surprising that for race groups where the usual position is high school completion with a relatively high proportion possessing some post-secondary qualification, the impact of educational attainment on welfare is less than for groups where the educational attainment levels of most are lower. The latter is confirmed in this sample, as black and Coloured people possess the lowest average levels of educational attainment relative to the Indian and white groups.

Married people are significantly more satisfied than those who are single (for the entire sample and black individuals). In the overall sample, the divorced seem to be less satisfied than singles, although the difference is not statistically significant (p = 0.115). Powdthavee (2005) and Mahadea & Rawat (2008) found similar results regarding divorce while these findings are broadly in contrast to Hinks & Gruen (2007).

For the overall sample, the results show that women are significantly less satisfied than men (p < 0.05). This finding is inconsistent with that of Hinks & Gruen (2007), who found no significant differences in happiness between men and women in South Africa. Black men are more satisfied than black women (p < 0.01), but there are no significant gender differences in the other race groups. The fact that black women are less satisfied relative to black men could be attributed to the more patriarchal societal structure in many black cultures, with women still required to fulfil traditional roles.

The number of children is positively and significantly associated with individual well-being in the overall sample (p < 0.10) and among blacks (p < 0.05) and Indians (p < 0.05). This is in contrast to Mahadea & Rawat (2008), who found children and happiness to be negatively associated. One possible explanation for the findings of this paper is that having children may be a means of signalling higher levels of status in certain communities.

The differences between men and women, and possibly the effect of children, can be understood in the context of the proportion of urban single parents by race and gender (South African Institute of Race Relations, 2009:46–7): 52% of black urban parents are single, while only 24% of white parents are in this group. Of African single parents, 79% are women (69% for white people). Only 37% of African fathers of children aged 15 years or younger were present (living with the family), compared with 55.4% of coloured, 86.6% of Indian, and 86.7% of white families.

As expected, health status is both significant and positively associated with reported well-being in the overall sample and is in accordance with the findings of, for example, Gerdtham & Johannesson (2001). Those who report fair health and above-average health have a higher level of well-being than those who report a poor health status. Besides being true of the overall sample, it also holds for each individual race group with the exception of Indians where the results are insignificant across all categories. It should be noted, however, that the small sample size of the Indian population group (n = 156) may offer some explanation for the insignificant finding.

Consistent with the findings of Rule (2007), religious importance exhibits a positive relationship with individual life satisfaction in the overall sample as well as for blacks and Indians. For the latter groups, individuals who view religion as very important are more satisfied than those who attach no importance to religious activity. Post-estimation chi-square tests for coloureds indicate that those viewing religion as important are not significantly more satisfied than those who report religion as unimportant (p = 0.4492), with the results for whites indicating that those who report religion as very important are more satisfied than those who report religion as being unimportant (p < 0.01). The link between well-being and religion is strongest for Indian people, as those who attach great value to religious activities are far more satisfied than those who regard them as unimportant.

Absolute income is positively related to life satisfaction but is only significant in the total sample (p < 0.01) and for black (p < 0.001) and Indian (p < 0.01) people. Historically, black people have been poorer relative to other race groups, especially whites. This is clearly demonstrated by the much higher percentage of black South Africans living in poverty (49%) when compared with white people (3.6%) (South African Institute of Race Relations, 2009:306). Black people reported the lowest levels of mean income in the sample (R2520.53), which further supports the argument that black people are still the poorest race group. The strong positive association between absolute income and life satisfaction in the black population may therefore indicate that extra income adds substantially more to well-being compared with other racial groups, where most may already have reached an income threshold above which subjective well-being does not increase with additional income. These results are broadly in line with Powdthavee (2003), Hinks & Gruen (2007) and Mahadea & Rawat (2008), and confirm the well-known finding of a positive relationship between happiness and absolute income in cross-sectional studies.

In a similar vein, the level of relative income displays a strong positive association with life satisfaction. The importance of positional status is especially strong in the black and coloured groups and to a much lesser extent in the white group. These results confirm the idea that individuals measure their happiness relative to those in their reference group (Clark et al., 2008) and are consistent with the South African findings of Hinks & Gruen (2007) as well as Posel & Casale (2011).

Results from the complete sample regression suggest that the unemployed are less satisfied than the employed, even when controlling for relative and absolute income (p < 0.05). The finding is supported by the many international studies in both developing and developed countries that report a negative association between unemployment and happiness (Clark & Oswald, 1994; Oswald, 1997; Ravalion & Lokshin, 2001; Stutzer, 2001; Graham, 2008). For the individual race groups, the employment coefficient is statistically significant for blacks (p < 0.001) and for Indians (p < 0.01).

According to the South African Institute of Race Relations (2009:232), 27% of African people were unemployed in 2008 (using the narrow definition of unemployment) while 4.6% of white people, 19.5% of coloured people, and 12.7% of Indian people were in this category. (Using the broad definition of unemployment, the gap between African unemployment and that of other race groups widens even further.) It is conceivable that provision of employment to black individuals, even at lower wages, would improve the well-being of this population group significantly. Since employment would facilitate higher income and since both are positively associated with greater life satisfaction, eradication of unemployment among black people is sure to affect the general well-being of this group positively.

Finally, the area of residence dummy is positive and statistically significant in the overall sample and for black people (both p < 0.05). This suggests that on the whole South Africans who reside in urban areas are more satisfied with their lives than those in rural areas. That black urban dwellers are more satisfied than black rural dwellers is not surprising as the majority of black people in this sample live in rural areas. Lack of service delivery in rural areas especially, evident from the many service delivery protests in rural parts of South Africa, is a likely factor for the lower levels of reported life satisfaction among residents in rural areas.

5. Conclusion

Subjective well-being measures can be an important tool in designing effective development policies (Graham, 2008) and are also important for social cohesion and a sustainable democracy (Kunene, 2009). Although South Africa has made significant progress in the provision of material goods to previously disadvantaged citizens, which could enhance feelings of solidarity and empathy among South Africans, the society is still split largely along racial lines with respect to such matters as the incidence of poverty, unemployment and educational attainment.

This paper investigated life satisfaction among different race groups in South Africa using data from the NIDS conducted in 2008. The overall results indicate that black South Africans still report much lower levels of well-being than other race groups, supporting the findings of previous research (Powdthavee, 2003; Hinks & Gruen, 2007; Møller, 2007; Mahadea & Rawat, 2008), which continues to pose a threat to social cohesion. Higher levels of educational attainment increase satisfaction for the whole sample. Women, particularly black women, are generally less satisfied than men. As found in many other studies, unemployed people have lower levels of life satisfaction than the employed, even when controlling for income and relative income.

The determinants of life satisfaction are different for each race group: while whites attach greater importance to physical health, employment status and absolute income matter greatly for blacks. For coloureds and blacks, positional status (as measured by relative income) is an important determinant of life satisfaction, with religious involvement contributing significantly to the well-being of Indians. Black people in rural areas are significantly less satisfied than their urban counterparts.

Such findings have important implications for development policies that seek to reduce the differences in subjective welfare between South Africans. For example, for black South Africans, the results imply that more employment opportunities, even at lower wage rates, would improve welfare. Strongly unionised labour, which has driven up wage rates, even when unemployment has been increasing, could be seen in this context as counterproductive.

Acknowledgements

The authors are grateful to participants at the 2011 Biennial Conference of the Economic Society of South Africa, 5–7 September in Stellenbosch, as well as four referees for comments and suggestions made on an earlier version of this paper.

Notes

⁴ It should be noted that unemployment prevalence in this sample is considerably higher than official statistics would suggest. Examination of cross-tabulations for employment status (using a derived variable indicating whether someone is employed, unemployed according to the strict definition of unemployment, unemployed according to the broad definition of unemployment, or not economically active) in the NIDS dataset revealed that, for people younger than 65 years old, roughly 32% were classified as not economically active. This is highly unlikely, and probably stems from confusion between definitions of discouraged workers (who would be counted as unemployed in a broad definition of unemployment) and not economically active persons (who would not be counted as part of the labour force).