Self-assessed well-being and economic rank in South Africa

Abstract

This study explores subjective measures of well-being in South Africa collected in the first two waves of the National Income Dynamics Study. These subjective measures include individual life satisfaction, current self-assessed economic rank and expected economic rank in the future. The paper describes how the distributions of these measures have changed over the course of the panel and it investigates the relationship between life satisfaction and perceived economic rank in a multivariate context, controlling for individual fixed effects. The panel data suggest a leftward shift in the distribution of life satisfaction over the two waves. Moreover, the majority of adults did not perceive their economic rank as having improved and they reported lower expectations of future upward economic mobility. Perceptions of current and future economic rank are key correlates of life satisfaction, findings that remain robust to controls for unobserved individual heterogeneity.

Keywords:

• subjective well-being
• life satisfaction
• relative income
• perceived economic rank

JEL codes:

• D31
• D63
• I31

1. Introduction

Most nationally representative household surveys in South Africa collect data on money-metric measures of well-being, which are then used to generate statistics on poverty and inequality. However, these measures may be limited in several ways. First, they typically are not able to identify differences in economic well-being within the household when all resources in the household are not equally shared. Second, income received or spent captures only one aspect of economic status specifically and of well-being more generally, and a wide range of other factors will also affect an individual's quality of life.

In recent decades, subjective indicators increasingly have been used to complement money-metric indicators of well-being. There is by now a substantial international literature that explores the measurement and correlates of self-reported happiness (or life satisfaction) and perceived economic status (for reviews, see Stutzer & Frey, 2010; Ravallion, 2012). However, although research on income poverty and inequality in South Africa is very well established, there has been little work on subjective indicators, at least partly because this information was not regularly available in national household survey data. Data collected in the National Income Dynamics Study (NIDS) provide a unique opportunity to augment income measures of well-being for South Africa because NIDS is the first household panel survey that includes a range of questions asking respondents to provide subjective assessments of their individual well-being, in addition to collecting detailed information on income and expenditure.

This study analyses two sets of self-assessed measures of well-being: life satisfaction and perceived economic status. Section 2 outlines the nature of information collected in NIDS and considers how the distributions of life satisfaction and perceived economic status have changed across the two waves. The section also probes how self-assessed economic status differs from economic status measured using reported income. Section 3 estimates the correlates of self-assessed life satisfaction using pooled and fixed-effects regression analysis. A key objective is to explore the relationship between life satisfaction and self-assessed economic status. Section 4 summarises the main findings of the study.

2. Subjective measures of well-being

2.1 Subjective well-being or life satisfaction

Many international studies have found that when people are asked to assess how satisfied or happy they are with their lives, their responses provide meaningful and useful measures of their quality of life (for reviews of these studies, see Kahneman & Krueger, 2006; Stutzer & Frey, 2010). Self-reported happiness or life satisfaction, commonly referred to as subjective well-being (SWB) in the literature, is well correlated with a range of other characteristics or factors that would be expected to influence an individual's quality of life, such as unemployment or ill-health, and follow-up studies have found that low levels of subjective well-being predict mortality, particularly in men, and suicide (Koivumaa-Honkanen et al., 2000, 2001; Stutzer & Frey, 2010).

In South Africa, data on subjective well-being have been collected in a few nationally representative household surveys, but the question on life satisfaction typically has been asked about the household: ‘how satisfied is the household with how it lives these days’.² This framing of the question assumes not only that a respondent would be able to report objectively on a household's level of satisfaction, but more fundamentally that there is a unified SWB function at the household level.³

In contrast, NIDS collects information on life satisfaction at the level of the individual. In both waves 1 and 2 of NIDS, all adults were asked the following question: ‘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?’ (See Question M5 of the adult questionnaire, Southern Africa Labour and Development Research Unit, 2008:23.)

Figure 1 compares responses among all adults (aged 17 years and older) in the two waves of NIDS. Only those adults who were present in both waves as resident household members are included in the sample for each wave. In both years, the modal level of reported satisfaction was five. However, the distribution in reported life satisfaction has clearly shifted to the left over the two-year period. The percentage of all adults who report being dissatisfied increased from 2008 to 2010, while the percentage who report being satisfied has fallen. For example, whereas 21% of all adults reported a satisfaction level of three or lower in 2008, this increased by 10 percentage points to 31% in 2010.

Figure 1: Subjective well-being in South Africa, 2008 and 2010

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include adults who were older than 16 years in wave 1. The data are weighted.

The overall distribution in reported life satisfaction masks sharp cleavages by race, as illustrated in Figure 2. Although the modal level of satisfaction reported by African adults was five (mirroring the national distribution), it was eight among white adults. Furthermore, race differences widened over the two-year period. Both whites and Africans reported lower levels of satisfaction overall in 2010 than in 2008, but the decline was more pronounced among Africans. By 2010 the majority of African adults in the panel (53%) reported a satisfaction level of four or lower, compared with 39% in 2008. The comparable figures among whites are 9% in 2010 and 7% in 2008.

Figure 2: Subjective well-being among Africans and whites: (a) 2008 and (b) 2010

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include adults who were older than 16 years in wave 1. The data are weighted.

The data presented in Figures 1 and 2 consider the two waves of NIDS as individual cross-sections. Table 1 takes advantage of the panel nature of the data and describes changes in subjective well-being for each individual. Of the 8870 adults who provided information on their subjective well-being in both waves of the panel, only about 15% reported the same level of satisfaction in 2010 as in 2008, although the percentage among whites specifically is considerably higher (25%). Over one-half of all adults reported lower levels of subjective well-being in 2010 than in 2008, a finding that is driven particularly by the decline in satisfaction among Africans. Slightly more than one-half (53%) of all African adults in the panel reported being less satisfied in 2010 than in 2008, compared with 35% of white adults.

Table 1: Differences in subjective well-being among individuals in 2008 and 2010

Difference in satisfaction level (2010 value – 2008 value)	Percentage of all adults	Percentage of African adults	Percentage of White adults
–9	0.62	0.71	–
–8	0.84	1.01	0.08
–7	2.12	2.43	0.71
–6	2.78	3.10	1.74
–5	4.51	4.86	1.17
–4	6.01	6.81	1.89
–3	9.57	10.08	7.90
–2	11.60	12.18	10.45
–1	11.77	11.86	10.71
0	14.60	12.76	24.93
1	10.05	9.18	14.92
2	8.63	8.03	9.84
3	6.58	6.15	9.09
4	4.65	5.06	3.04
5	2.96	2.91	3.52
6	1.29	1.31	–
7	0.68	0.76	–
8	0.24	0.25	–
9	0.48	0.56	–

Source: Author's own calculations, NIDS 2008 and 2010.

Notes: The samples include adults who were older than 16 years in wave 1. The statistics in bold represent the percentage of adults whose measure of life satisfaction was unchanged across the two waves of NIDS. The data are weighted.

2.2 Subjective measures of economic status

In addition to collecting subjective measures of life satisfaction, the adult module in waves 1 and 2 of NIDS includes a number of questions asking individuals to assess their economic status. Using a six-step ladder, with the bottom step representing the poorest people in South Africa and the top step representing the richest, adults were asked to identify on what step they thought their household ranked today. They were also asked on what step they thought they ranked when they were aged 15, and on what step they expected to rank two years (and five years) in the future. (See questions M1 to M4 of the adult questionnaire.) These questions capture information on perceived relative economic status – relative to others in South Africa, or relative to where the individual ranked in the past or expects to rank in the future.

Figure 3 compares the distribution of current perceived economic ranking in South Africa, in each of the waves of the panel. In 2008, the modal ladder step was two and a little under one-half of all adults (48%) thought that they ranked among the poorest third (steps 1 and 2) of South Africans. By 2010, the modal ladder step had increased to three, and a smaller share of adults (44%) assessed their economic status as being on steps 1 and 2 of the economic ladder. The percentage of adults who perceived their economic status as being in the middle of the economic ladder (steps 3 and 4) also rose slightly (by three percentage points) to 52%. However, in both years, less than 4% of adults in the sample thought that they were among the richest third of South Africans.

Figure 3: Perceived economic ranking in South Africa, 2008 and 2010

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include adults who were older than 16 years in wave 1. The data are weighted.

Of course it is not possible for less than 4% of all adults to be among the richest one-third of South Africans, and for more than 40% to be among the poorest third. Rather, the statistics presented in Figure 3 suggest that a considerable share of South African adults underestimates their relative economic position – in comparison with others in South Africa, a sizeable group of people are actually better off than they perceive themselves to be.

Table 2 explores this further by comparing the perceived economic rank of adults with how adults would rank using a measure of economic well-being derived using reported income.⁴ To identify an income rank, the distribution of per-capita household income is divided into thirds. Similarly the six ladders steps are grouped into thirds (with the bottom two steps representing the lowest third, for example) (see Posel & Casale, 2011).

Table 2: Income versus perceived economic rank in South Africa, 2008 and 2010

Income rank	Perceived richest	Perceived middle	Perceived lowest	Total
2008
All
Richest third	0.07 (0.07)	0.65 (0.01)	0.28 (0.01)	1.00
Middle third	0.03 (0.00)	0.45 (0.01)	0.53 (0.01)	1.00
Lowest third	0.01 (0.00)	0.33 (0.01)	0.66 (0.01)	1.00
Africans
Richest third	0.05 (0.01)	0.61 (0.02)	0.34 (0.02)	1.00
Middle third	0.02 (0.00)	0.45 (0.01)	0.53 (0.01)	1.00
Lowest third	0.01 (0.00)	0.33 (0.01)	0.65 (0.01)	1.00
Whites
Richest third	0.10 (0.02)	0.73 (0.03)	0.17 (0.02)	1.00
Middle third	0.02 (0.02)	0.50 (0.11)	0.48 (0.11)	1.00
Lowest third	0 (0)	0.57 (0.23)	0.43 (0.23)	1.00
2010
All
Richest third	0.08 (0.01)	0.68 (0.01)	0.25 (0.01)	1.00
Middle third	0.02 (0.00)	0.46 (0.01)	0.52 (0.01)	1.00
Lowest third	0.01 (0.00)	0.39 (0.01)	0.60 (0.01)	1.00
Africans
Richest third	0.05 (0.01)	0.64 (0.02)	0.31 (0.02)	1.00
Middle third	0.02 (0.00)	0.46 (0.01)	0.53 (0.01)	1.00
Lowest third	0.01 (0.00)	0.39 (0.01)	0.60 (0.01)	1.00
Whites
Richest third	0.14 (0.02)	0.74 (0.03)	0.12 (0.02)	1.00
Middle third	0.0 (0.0)	0.22 (0.13)	0.78 (0.13)	1.00
Lowest third	0 (0)	0.08 (0.09)	0.92 (0.09)	1.00

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include adults who were older than 16 years in wave 1. The data are weighted. Standard errors are in parentheses. An individual's income rank is calculated with reference to average per-capita total household income. The statistics in bold represent the proportion of adults for whom perceived economic rank coincided with income rank.

The table describes considerable divergence between perceived relative status and income ranking. For example, in 2008 only 7% of adults who ranked among the richest one-third in terms of reported income perceived their economic status as corresponding to the upper two steps of the economic ladder. The majority (65%) perceived their relative economic status to be in the middle of the economic ladder (steps 3 and 4). The largest correspondence between the income rank of individuals and their perceived economic rank occurs among adults in the bottom third of the income distribution. In 2008, 66% of adults who were placed in the bottom third of the income distribution also perceived that they ranked on the bottom two steps of the economic ladder.

The table suggests further that among adults in the upper third of the income distribution, the divergence between income rank and perceived rank is considerably larger among Africans than among whites. In 2008, for example, only 5% of Africans in the upper third of the income distribution perceived their relative economic status as corresponding to the richest third of South Africans; while 34% ranked their economic status on the lowest two steps of the ladder (the comparable percentages for whites in the upper third of income are 10% and 17% respectively).

What explains this poor match between where individuals think they rank on the economic ladder and where they actually rank in the income distribution, and why is the divergence larger among richer Africans than whites? One explanation is that the level of income associated with a ‘middle-class’ lifestyle may far exceed the middle of South Africa's income distribution, leading respondents to underestimate their ranking on the income ladder. It is also possible that people base their assessments in the ladder question on accumulated income and expected future income (or permanent income), while the income rank is based on current monthly income. In the context of large historical inequalities in access to resources, current monthly income may not be a good predictor of permanent income, and particularly among Africans. Even though Africans may rank in the upper third of the income distribution, their economic status in terms of permanent income may be lower.

A further explanation is that people do not have complete or accurate information about the economic status of others. Given racially differentiated opportunities in the past, whites have higher levels of education and are more likely to be proficient in English, the dominant language of business, politics and communication in the country (Casale & Posel, 2011). Consequently whites may have access to more information when assessing their relative economic status, helping to explain why the divergence is smaller for this sub-sample. In addition, one of the legacies of apartheid may be that even relatively rich Africans still perceive their economic status as being inferior, particularly when compared with whites (Posel & Casale, 2011).

A comparison of the two years of data presented in Table 2 suggests that, over the two waves of NIDS, perceived economic status increased primarily among adults in the bottom third of the income distribution. In particular, a growing share of adults in the lowest income third thought that they ranked in the middle of the economic ladder (39% in 2010, compared with 33% in 2008). To explore changes in self-assessed economic status further, Figure 4 describes the difference in reported ladder steps in 2010 and 2008, for each individual in the panel. In 2010, a little over one-third of all adults reported being on the same ladder step as in 2008; 31% reported being on a lower step, and 35% on a higher step. However, Figure 5 shows that whites were more likely than Africans to perceive their economic ranking in South Africa as unchanged across the waves, and Africans were more likely to view their economic ranking as having declined.

Figure 4: Differences in perceived economic rank (2010–2008)

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include adults who were older than 16 years in wave 1. The data are weighted.

Figure 5: Differences in ladder steps (2010–2008), Africans and whites

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include adults who were older than 16 years in wave 1. The data are weighted.

One of the more striking changes in the responses to questions about perceived economic status concerns expectations of future mobility. In 2008, 72% of adults anticipated being on a higher rung in two years time compared with at the time of the survey; by 2010 this had fallen to 51%. Table 3 shows that expectations of future mobility declined particularly among Africans (from 77% in 2008 to 50% in 2010). In comparison with other groups, whites are distinctive – although they were the least likely to anticipate being on a higher ladder rung in the future, their expectations of future mobility did not decline across the waves and rather increased.

Table 3: Anticipated upward mobility (two years in the future)

	2008	2010
African	0.77 (0.01)	0.50 (0.01)
Coloured	0.71 (0.02)	0.56 (0.03)
Indian	0.69 (0.05)	0.55 (0.07)
White	0.40 (0.03)	0.53 (0.03)

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include adults who were older than 16 years in wave 1.The data are weighted. Upward mobility is defined as an individual expecting to be on a higher ladder step in two years time compared with at the time of the survey.

In sum, descriptive statistics identify three broad changes in the subjective assessments of well-being among adults in South Africa. First, over one-half of the resident adults reported being less satisfied with their lives in 2010 than in 2008; and Africans were considerably more likely than whites to report lower levels of satisfaction. Second, almost two-thirds of the adults did not perceive that their economic ranking in South Africa had improved over the period, although there is also evidence that richer individuals considerably underestimate their relative class position. Third, expectations of future upward mobility declined markedly among adults, and particularly among Africans. These changes in subjective perceptions of economic rank may partly reflect the effects of the economic recession in South Africa (2008–09), where economic output contracted and employment levels declined significantly (Verick, 2012). However, they suggest further that the effects of this recession have been experienced more strongly by Africans than by whites. The next section explores the relationship between changes in self-assessed economic rank and changes in satisfaction among Africans between the two waves of the panel.

3. Predicting life satisfaction: the role of perceived economic status

There is a large literature from both psychology and economics that investigates what makes people more or less satisfied with their lives. One of the main themes to emerge from the economics literature in particular is how the economic status of individuals affects their subjective well-being. Several studies have shown that self-assessed satisfaction is influenced not simply by how rich or poor individuals are, but also by how their economic status compares relative to others (cf. Easterlin, 1974, 1995; McBride, 2001; Ferrer-i-Carbonell, 2005; Luttmer, 2005; Kingdon & Knight, 2007; Bookwalter & Dalenberg, 2009). To identify this relative status, or relative income, studies have commonly used the individual's position in the distribution of reported income (or expenditure), or compared the individual's reported income with the mean or median income of a selected comparison group.

However, as the discussion in the previous section has illustrated, there can be considerable divergence between an individual's actual and perceived ranking in the income distribution. In particular, richer South Africans typically underestimate how well off they are in comparison with others. If relative standing affects individual subjective well-being because of perceptions of relative deprivation or relative advantage, then a subjective measure of relative standing may provide a better measure of relative income than where individuals are actually rank based on their income.

An earlier study explored the relationship between relative economic standing and an individual's level of satisfaction in South Africa, using wave 1 of the NIDS data (Posel & Casale, 2011). The results from this study suggest that, in addition to absolute income, relative economic standing, measured by the individual's rank in the income distribution, is a significant predictor of how satisfied individuals are with their lives. However, individual perceptions of relative standing are an even stronger predictor. For example, individuals in the richest third of the income distribution reported far higher levels of satisfaction if they also thought that they ranked among the richest third of South Africans.

One of the limitations of this study is that with data from only one wave of NIDS, it was not possible to control adequately for an individual's personal traits or attitudes to life. If these unobserved individual characteristics are correlated with both reported levels of satisfaction and perceptions of relative standing, then this endogeneity will produce bias in the estimates. With the release of the second wave of NIDS, there is now information for each individual at two points in time, and it is therefore possible to control for those unobserved individual characteristics that are time invariant using fixed effects (or first differencing) estimation techniques, and in so doing to test whether the earlier findings from Posel & Casale (2011) are robust to heterogeneity bias.

To estimate the predictors of subjective well-being, I estimate three regressions:

where SWB_it is individual i's reported level of satisfaction in time t; Y_i is a vector of income variables; X_i is a vector of individual demographic and employment-related variables; SWB_i*, Y_i* and X_i* represent the mean values of the variables over the two waves; δ_i is the time-invariant error capturing unobserved individual-specific characteristics; and ν_it is the idiosyncratic or time-varying error.

The first regression (Equation 1) is an ordinary least squares (OLS) regression estimated only for wave 1 of NIDS.⁵ The second regression (Equation 2) is a pooled OLS regression for both waves of NIDS, and includes a wave dummy variable, W_t for t = 1. The third regression (Equation 3) is the fixed-effects estimation, which exploits differences in the independent and explanatory variables for each individual across the two waves.

The income variables measure an individual's income rank, as described in the previous section (whether in the upper or middle third of the income distribution, with the poorest third as the omitted category); an individual's perceived economic rank (on the upper two steps of the economic ladder, or the middle two steps, with the lowest steps as the omitted category); whether an individual expects to be on a higher step of the economic ladder in the future (upward mobility); and the individual's absolute income (measured as the log of per-capita household income). Individual characteristics include: age; education (completed secondary or tertiary education, with less than completed secondary as the omitted category); marital status (married, cohabiting, widowed or divorced, with never married as the omitted category); employment status (employed or not economically active [including students, housewives and pensioners], with unemployed as the omitted category); health status (measured by whether the individual reports being in very good health); and whether the individual reports difficulty with daily activities such as dressing, bathing or eating.

Table 4 reports the results of the estimations for all African adults (17 years and older) in the NIDS panel.⁶ Overall, the fixed-effects estimation confirms earlier findings in Posel & Casale (2011), based only on wave 1 of NIDS. Absolute income is a positive and significant correlate of subjective well-being among Africans in both the pooled and panel regressions. In addition to absolute income, an increase in the individual's relative economic standing is also associated with significantly higher satisfaction levels. However, it is specifically perceived economic rank rather than actual income rank that matters. In fact, actual income rank is not a significant predictor of satisfaction levels in the fixed effects or pooled regressions. In comparison with the coefficients in the OLS pooled regression, the coefficients for perceived economic rank fall,⁷ but the relationship between the estimated coefficients remains unaltered: an increase in perceived economic rank has an increasing effect on reported levels of satisfaction. Positive expectations about future relative economic standing also positively affect self-assessed satisfaction. Part of the explanation for falling levels of satisfaction, particularly among Africans, may therefore lie with the decline in perceived economic rank and the fall in positive expectations of upward mobility in the future.

Table 4: Predicting subjective well-being among Africans: OLS and fixed effects

	NIDS Wave 1 (2008)	NIDS Panel (2008–10)
	OLS	OLS pooled data	Fixed effects
Perceived richest third	2.405***	1.870***	1.683***
	(0.185)	(0.117)	(0.211)
Perceived middle third	0.969***	0.821***	0.718***
	(0.050)	(0.032)	(0.058)
Richest income third	0.263**	–0.038	–0.091
	(0.119)	(0.082)	(0.154)
Middle income third	0.239***	–0.005	0.059
	(0.072)	(0.048)	(0.091)
Expect upward mobility	0.356***	0.267***	0.212***
	(0.055)	(0.032)	(0.058)
Log (per-capita household income)	0.166***	0.277***	0.136**
	(0.045)	(0.031)	(0.062)
Age^2	–0.000	0.000***	0.001***
	(0.000)	(0.000)	(0.000)
Matric (Grade 12)	0.052	0.019	–0.131
	(0.071)	(0.044)	(0.133)
Diploma or degree	0.195**	0.285***	0.181
	(0.097)	(0.061)	(0.186)
Married	0.201***	0.114***	0.557***
	(0.064)	(0.042)	(0.205)
Cohabiting	–0.019	–0.014	0.013
	(0.084)	(0.058)	(0.202)
Divorced or widowed	0.044	–0.031	0.409*
	(0.092)	(0.061)	(0.245)
Employed	0.110*	0.089*	0.327***
	(0.066)	(0.047)	(0.089)
Not economically active	0.200***	0.135***	0.272***
	(0.066)	(0.044)	(0.083)
Daily activities difficult	–0.628***	–0.504***	–0.448***
	(0.133)	(0.079)	(0.146)
Very good health	0.446***	0.549***	0.495***
	(0.050)	(0.034)	(0.061)
Wave 1 dummy	–	0.743***	0.926***
		(0.033)	(0.092)
R^2	0.098	0.108	0.077 (within)

Source: Author's own calculations, NIDS 2008 and 2010. Notes: The samples include African adults who were older than 16 years in wave 1. Standard errors are in parentheses. All estimations also included a quadratic in age and a wave dummy variable. ***Significant at the 1% level **Significant at the 5% level *Significant at the 10% level.

In addition to subjective measures of current and future economic status, changes in the individual's demographic characteristics also significantly affect satisfaction in ways that are consistent with many other studies (for reviews of findings in the subjective well-being literature, see Dolan et al., 2008; Blanchflower, 2009). Life satisfaction increases significantly among individuals who acquire tertiary education, who marry, become employed or not economically active (rather than unemployed), and who report being in very good health.⁸ In contrast, individuals who experience difficulty with basic daily activities become significantly less satisfied.⁹

4. Conclusion

Similar to the distribution of income in South Africa, the distribution of life satisfaction is highly skewed by race. Whites are not only richer than Africans, but they also report higher levels of life satisfaction. Differences in access to resources help explain these differences in subjective well-being. However, satisfaction levels are influenced not only by the level of absolute income, but also by relative income, and in particular by how individuals think their income compares with the income of others.

There is considerable divergence between where individuals think they rank on an income ladder for South Africa and where they are actually ranked based on their reported income. Self-assessed rankings typically are far lower than ‘actual’ income rankings. This disparity may arise if perceived rankings incorporate assessments of income flows over time, while income rankings are based only current monthly income. Richer individuals probably also underestimate how well-off they are in comparison with other South Africans, and associate what they perceive to be a ‘middle-class’ lifestyle with being in the middle of South Africa's income distribution. Perceptions of relative income status therefore provide a more meaningful measure of relative deprivation or relative advantage, and perceived ranking is a stronger predictor of subjective well-being than a ranking based on reported income.

The period spanned by waves 1 and 2 of NIDS coincides with an economic recession in South Africa. Over this period, reported levels of life satisfaction declined significantly among adults, but by considerably more among Africans than among whites. More than one-half of all African adults in the longitudinal sample reported lower levels of satisfaction in 2010 than in 2008 (compared with less than 40% of whites). Key to explaining these trends in life satisfaction is that Africans were also more likely than whites to perceive their economic standing relative to others in the country as having fallen over the two-year period, and to lower their expectations of future upward mobility.

Notes

2 See, for example, the 1993 Project for Statistics on Living Standards and Development and the 1998 October Household survey.

3 Most of the earlier literature using these data overlooked this latter concern, and instead dealt with whether a single respondent was able to report reliably on the household's subjective well-being (Bookwalter et al. 2006; Kingdon & Knight 2006, 2007).

4 In NIDS, detailed information is collected on both labour and non-labour income received by individuals (including earnings, social grants, remittances and private pensions). Non-labour income is collected as point values. Wages and earnings are also reported as point values except where respondents did not or would not provide this information, in which case earnings were reported in parentheses. To generate a continuous income variable, earnings in parentheses were assigned the mid-point of the parenthesis.

5 Life satisfaction is often also estimated using ordered probit regressions. Because the cut-off points in the probit regressions for wave 1 and for the pooled sample are relatively equally spaced, OLS regressions are used (treating the dependent variable as a linear measure of subjective well-being).

6 A Hausman test rejected the null hypothesis of no systematic differences between the coefficients from a random and fixed-effects model (χ² = 57.56), suggesting that a fixed-effects model is more appropriate.

7 A fall in the estimated coefficients may derive from both endogeneity bias and measurement error.

8 Although not reported in Table 4, a quadratic in age is significant in both the Wave 1 and pooled regressions, with subjective well-being initially decreasing with age and then increasing, a common finding documented across a wide range of empirical studies on subjective well-being.

9 The positive and significant coefficient for the wave 1 dummy variable captures the aggregate decline in reported subjective well-being across the waves.