Volume 72, Issue 3 p. 842-849
BRIEF REPORT
Open Access

Global evidence of inequality in well-being among older adults

Michael D. Smith PhD

Michael D. Smith PhD

National Oceanic and Atmospheric Administration, Economics and Social Sciences Research, Alaska Fisheries Science Center, Seattle, WA, USA

Search for more papers by this author
Dennis Wesselbaum PhD

Corresponding Author

Dennis Wesselbaum PhD

Department of Economics, University of Otago, Dunedin, New Zealand

Correspondence

Dennis Wesselbaum, Department of Economics, University of Otago, Dunedin, New Zealand.

Email: [email protected]

Search for more papers by this author
First published: 01 December 2023
Citations: 1

The findings and conclusions in the article are those of the author(s) and do not necessarily represent the views of the National Marine Fisheries Service, NOAA.

Abstract

Background

By 2030, the UN expects 1.4 billion older adults and 2.1 billion by 2050. By 2050, 80 percent of older persons will live in developing nations. This demographic shift will present new challenges and opportunities in several areas, including health, migration, employment, and social safety nets. This study's aims were to: (1) present novel evidence on the trends and determinants of well-being and well-being inequality among older people around the world; and (2) highlight variation across World Bank development groups.

Methods

The study utilizes individual-level survey data from nine waves of the Gallup World Poll (2009–2017), which is representative of about 99.5% of the global population. First, we report country-level panel evidence on well-being and well-being inequality for adults over 60 years of age. Second, we estimate regressions to identify the individual-level determinants of well-being and well-being inequality.

Results

Our results indicate that average levels of happiness vary little over time. This holds for all World Bank development groups. In contrast, we show that inequality in well-being increases for all categories except in high-income countries. Examining the factors that influence well-being and well-being inequality reveals the particular importance of income, social ties, and health. We also reveal gender differences in global well-being; women tend to be happier than men. Lastly, whereas variations in inequity-causing factors are minimal when comparing older to younger individuals, they vary substantially when comparing across development groups.

Conclusions

Our findings suggest that rather than focusing on the average level of well-being among older people, governments should consider the full distribution of well-being. This requires a special emphasis on health, social networks, and education, as well as the assessment of distributional impacts in policy proposals.

Key points

  • Mean well-being does not vary over time, but well-being inequality increases in all countries other than high-income ones.
  • Comparing older and younger adults reveals similarities in the predictors of inequality.
  • When comparing across development groups, the predictors of inequality differ greatly.

Why does this paper matter?

Promoting well-being for all ages is a central objective of global policy. The share of older adults is rising rapidly, particularly in the developing world. As a holistic indicator of development, the distribution of well-being is correlated with other factors of life, such as trust, health, and social interactions. We present novel evidence on the global patterns and determinants of well-being and well-being inequality among older adults, as well as differences along the path of economic development.

INTRODUCTION

This decade has been designated by the World Health Organization (WHO) as the “decade of healthy ageing,” which is defined as “the process of developing and maintaining the functional ability that enables well-being in older age,” in accordance with Sustainable Development Goal 3 of the United Nations to promote well-being for all ages.1 The United Nations projects that the population of older adults will climb to 1.4 billion by 2030 and to 2.1 billion by 2050. Concurrently, the share of older people is growing rapidly, particularly in developing countries: it is anticipated that by 2050, over 80% of older persons will reside in developing countries.2 This demographic shift will present challenges in areas like health care, social safety nets, and economic development.3-6

Over the last decade, politicians and scholars have questioned gross domestic product (GDP) as the primary measure of development and rethought social progress and how it should be measured.7, 8 Subjective well-being (SWB) has emerged as a potential alternative to GDP.9, 10

SWB integrates a person's subjective evaluations and mental comparisons with multiple reference points, such as aspirations, their own past, and other people, into a single indicator that captures a person's objective welfare.11, 12 SWB includes hedonic well-being and life satisfaction components to assess life satisfaction. As SWB is a more holistic indicator of development, its distribution is related to other facets of life, such as trust, health, and social interactions.13 Inequality in SWB is “the ethics of social arrangements” and reveals resilience, living arrangements, risk, and opportunity.14, 15

Recent research on inequality in well-being has increased, however these studies are either limited to a few countries or do not examine older adults.16-18 In this article, we present novel evidence on the global patterns and determinants of well-being and well-being inequality among older adults, as well as differences along the path of economic development.

METHODS

Data set

The study employs individual-level survey data of 179,075 individuals from nine waves (2009–2017) of the Gallup World Poll (GWP).19 The GWP uses randomly selected (via Kish grid method) and nationally representative samples. Overall, it is representative of ~99% of the worldwide population.20 In most countries, GWP interviews 1000 adults. If the country has telephone coverage of at least 80%, the interviews are conducted via telephone; otherwise, they are conducted face-to-face. Despite efforts to limit measurement error at every stage of the cross-country survey process, survey error can still arise due to difficulties such as translation, framing, or strategic responses, but error margins are minimal.20 We follow the United Nations and define an “older person” if the respondent is 60 years or older.

Key outcome

The key outcome variable is subjective well-being. The GWP follows the guidelines by the Organisation for Economic Co-operation and Development (OECD) on measuring SWB using the question: “Please imagine a ladder, with steps numbered from 0 at the bottom to 10 at the top. The top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. On which step of the ladder would you say you personally feel you stand at this time?” The response is recorded on an 11-step Cantril ladder.19

Factors affecting well-being

The variables included in our empirical analysis adhere to established conceptual frameworks,21, 22 as well as empirical research.9, 10, 19, 23 GWP allows incorporating numerous individual- and country-level variables, including demographic characteristics, marital status, educational attainment, health status, income and work status, social relationships, and well-being- and income-inequality at the country level.24, 25 The supplementary material presents details of all variables and Supplementary Table S1 presents descriptive statistics.

Empirical methods

Our empirical strategy is drawn from conceptual frameworks of multidimensional well-being.18, 21, 22 We estimate the conditional mean and variance of individual-level well-being using a two-step approach.26, 27 In the first stage, we estimate the conditional mean of well-being for individual i in country j in year t:
Step 1 : SWB i , j , t = α + n = 1 N β n X n , i , j , t + θ j + θ t + ε i , j , t , (1)
where n is an indicator for the control variables. In the second stage, we exploit the mean-zero property of the residuals from the first step. Therefore, the conditional variance of SWB, Var ( SWB i , j , t X i , j , t , equals the conditional expectation of the squared residuals from regression (1),
Var ( SWB i , j , t X i , j , t = E ε i , j , t E ε i , j , t 2 = E ε i , j , t 2 X i , j , t . (2)
Then, the second step regresses the squared residuals on the same covariates:
Step 2 : ε i , j , t 2 = φ + n = 1 N ω n X n , i , j , t + θ j + θ t + ϵ i , j , t . (3)

Our measure of inequality, the conditional variance (2), reflects the dispersion of SWB among otherwise identical respondents i in a country j at time t, and, hence, is a direct measure of inequality.15, 27

In both steps, we include a rich set of control variables, X i , j , t , and country and year fixed effects, θ j , θ t , which capture stable differences across countries, such as cultural and policy differences, economic cycles, and survey context effects. All regressions are estimated using Stata/SE18 using a linear probability model and cluster-robust standard errors at all separate levels of country and year.18 We choose a linear probability model based on the findings by reference 28 showing that the results from this model are almost identical to an ordered logit model (see Supplementary Table S2).

We lose about 33% of observations from the original sample due to missing data. These observations were dropped from the sample because respondents failed to provide valid information on one or more of the questions. This is mainly due to the following variables: religiosity, native-born, social capital, feeling safe, and health problems. Although we reject Little's test of missing-completely-at-random, we find that the differences in mean missingness among our variables are often small in magnitude and likely only significant due to the large sample size. We used multiple imputation and full information maximum likelihood estimation and find no differences (see Supplementary Table S3).

Lastly, the descriptive statistics use survey weights, but our regression analysis does not, as the survey weights were constructed from the same control variables (robustness checks using sample-weighted regressions reveal no differences).29-31 All p-values for this study were two-sided and β-values present the regression coefficient with both variables in standard deviations.

FINDINGS

Figure 1 illustrates the country-level trends in mean and inequality in SWB among older people by World Bank development groups (Supplementary Figure S1 shows the histogram of SWB). We find the expected mean differences across development groups: older adults in high-income countries have the highest well-being. In addition, there is no discernible trend in mean SWB over time, nor is there any convergence in SWB levels across development groups.

Details are in the caption following the image
Evolution of mean and inequality in well-being. Left panel shows country-level means of SWB and right panel shows the country-level inequality in SWB (computed using the measure by reference 32 appropriate for bounded, categorical variables). Values computed from individual-level, survey weighted data from the 2009–2017 waves of the Gallup World Poll for adults 60 years and older. Colors and line patterns represent World Bank economic development groups.

In contrast, we observe substantial increases (or divergence) in SWB inequality in low-, lower-middle-, and upper-middle-countries. The greatest increase in well-being inequality occurs in low-income countries (~38%). Only older people in countries with high incomes have not experienced this increasing trend in well-being inequality. This suggests that older persons are, on average, as happy in 2017 as they were in 2009, but that over time well-being is distributed more unequally among them.

Mean well-being

Table 1 presents our results. We find that age is positively associated with mean well-being among older persons (p < 0.01, β = 0.03). Women have higher well-being than men (p < 0.01, β = 0.03) and married respondents are happier than those that are divorced, widowed, or separated (p < 0.01, β = 0.02). Children (p < 0.01, β = −0.01) reduce well-being, as does health problems (p < 0.01, β = −0.12) and living in a rural area (p < 0.01, β = −0.01). Higher well-being is correlated with social relationships, as measured by confidence in friends and family (p < 0.01, β = 0.12), feeling safe (p < 0.01, β = 0.05), and being religious (p < 0.01, β = 0.02). More educated respondents have higher well-being (p < 0.01, β = 0.09) and income has a U-shaped relationship with well-being.

TABLE 1. Regression results—Well-being.
≥60 years of age <60 years of age
(1) Conditional mean (2) Conditional variance (3) Conditional mean (4) Conditional variance
Age 0.010*** 0.020*** −0.010*** 0.002**
(0.001) (0.003) (0.0005) (0.001)
Adults −0.007 −0.009 −0.007*** 0.014*
(0.005) (0.014) (0.003) (0.007)
Children −0.083*** 0.070 −0.093*** 0.037*
(0.017) (0.056) (0.008) (0.020)
Female 0.163*** 0.103*** 0.145*** −0.051***
(0.012) (0.035) (0.008) (0.018)
Single 0.009 0.278*** 0.216*** 0.036
(0.023) (0.075) (0.012) (0.036)
Married/partner 0.112*** −0.206*** 0.231*** −0.173***
(0.014) (0.042) (0.010) (0.032)
Education
Secondary education 0.335*** −0.551*** 0.274*** −0.698***
(0.015) (0.053) (0.011) (0.045)
Post-secondary education 0.608*** −0.828*** 0.537*** −1.045***
(0.021) (0.067) (0.014) (0.057)
Log household income −0.275*** −0.014 −0.290*** 0.006
(0.017) (0.041) (0.012) (0.029)
Log household income 0.038*** −0.017*** 0.040*** −0.019***
(0.001) (0.003) (0.001) (0.002)
Health problems −0.593*** 0.168*** −0.350*** 0.518***
(0.013) (0.038) (0.013) (0.029)
Native-born 0.048* −0.352*** 0.078*** −0.337***
(0.024) (0.066) (0.017) (0.040)
Rural −0.049*** −0.047 −0.110*** −0.059*
(0.014) (0.040) (0.013) (0.031)
Religiosity 0.102*** 0.152*** 0.060*** 0.189***
(0.013) (0.040) (0.010) (0.023)
Confidence friends and family 0.756*** −0.482*** 0.650*** −0.459***
(0.017) (0.049) (0.014) (0.033)
Safety 0.272*** −0.161*** 0.215*** −0.104***
(0.014) (0.037) (0.010) (0.025)
Employment
Out-of-workforce 0.094*** −0.476*** 0.174*** −0.128***
(0.036) (0.105) (0.014) (0.038)
Self-employed 0.091** −0.355*** 0.100*** −0.308***
(0.037) (0.118) (0.015) (0.040)
Employed 0.113*** −0.380*** 0.081*** −0.257***
(0.035) (0.111) (0.012) (0.034)
National well-being inequality 0.224 0.931***
(0.192) (0.251)
National income inequality 0.008 −0.033
(0.024) (0.027)
Observations 179,075 179,075 820,049 820,049
Adjusted R2 0.32 0.07 0.28 0.06
  • Note: All regressions (ordinary least squares) include country- and year-fixed effects. Clustered standard errors at the country-year level in parenthesis.
  • * Significance level: p < 0.10;
  • ** Significance level: p < 0.05;
  • *** Significance level: p < 0.01.

The results for younger adults are comparable to those found in the literature.18, 23, 33 There are, however, some intriguing differences between younger and older adults. Age has a negative association with well-being in younger adults (p < 0.01, β = −0.05). More adults in the household also are associated with lower well-being (p < 0.01, β = −0.01). Being single (p < 0.01, β = 0.04) and being native-born are positively related to well-being (p < 0.01, β = 0.01). Surprisingly, well-being inequality at the national level is positively correlated with well-being (p < 0.01, β = 0.04). Also unexpected is the non-significance of national income inequality, despite evidence from previous studies indicating its significance.24, 25 However, these studies do not account for inequality in well-being, which may explain the difference.

Well-being inequality

Inequality in well-being is positively related to age (p < 0.01, β = 0.02), being female (p < 0.01, β = 0.01), having health problems (p < 0.01, β = −0.03), and being religious (p < 0.01, β = 0.01). Individuals who are married (p < 0.01, β = −0.02), have a higher education (p < 0.01, β = −0.04), are native-born (p < 0.01, β = −0.01), and have strong social relationships (confidence in friends and family p < 0.01, β = −0.03; feeling safe; p < 0.01, β = −0.01) have lower well-being inequality. Furthermore, inequality is lower among those with higher incomes (p < 0.01, β = −0.004). There are two differences in well-being inequality between older and younger adults: inequality is lower among younger women, and it is greater among older singles.

These regression results conceal considerable development group variation. Figure 2 illustrates the estimates by development group (classified using gross national income per capita: low-income, ≤$995, middle-income, >$996 and <$12,055, and high-income, >$12,056).

Details are in the caption following the image
Coefficient estimates by economic development group. Coefficient estimates for mean (left panel) and inequality (right panel) in well-being across World Bank development groups with 95% confidence intervals.

Most coefficients for mean well-being differ only marginally across development groups. The only noteworthy differences are income, which has a smaller effect in low-income countries than in middle- and high-income countries, and gender, which is significant in middle- and high-income countries but not in low-income countries. Furthermore, in middle- and high-income countries, health matters more than in low-income countries. In high-income countries, employment has a positive correlation with well-being, whereas in low-income countries, the correlation is negative. Lastly, only in low-income countries is there a positive association between national well-being inequality and well-being.

In contrast, we find substantial differences for the inequality in well-being across development groups for almost all coefficients. This holds for the sign of the relationship (for example income has a positive relationship in high-income countries but a negative one in low-income countries) and the magnitude of the relationship (especially for education). The only variables that have a similar association are age, the squared income term, and the number of adults within a household.

DISCUSSION

We present novel evidence regarding the global trends and characteristics associated with mean and inequality in older persons' well-being. Our findings indicate that the average level of well-being in older adults varies little over time. This holds for all economic development groups. In contrast, we show that inequality in well-being increases for all development groups except those with the highest incomes.

The importance of income, social relationships, and health is highlighted by examining the factors that determine the mean and inequality in well-being. We also document gender gaps and find that on average, women experience a greater sense of well-being than men. This contradicts evidence showing American men are happier than American women.34 This study, however, contains a limited number of control factors that could explain this finding and we leave this for future research. We find interesting differences between younger and older adults. Older adults' well-being is diminished by children, which may be attributable to worries over their children.35 Well-being inequality is lower among married, educated, native-born, high-income, and socially connected older adults. There is greater inequality among older, female, religious, and unhealthy respondents. These results are similar to younger adults with some differences, for example, women and non-married singles.

Finally, whereas the differences in inequality-causing factors are negligible when comparing older persons to younger ones, they become large when comparing across economic development groups. Almost every factor under consideration varies in magnitude, as well as in significance or direction.

Limitations

Several limitations should be highlighted. First, our regression results likely suffer from omitted variable bias and possible reverse causality, therefore they should be viewed as correlations. Using panel data would considerably improve our capacity to estimate causal effects and capture any dynamic effects. Additionally, the GWP survey does not use a complex survey design, which may limit the data's ability to capture historically underserved populations. Finally, we statistically reject the missing at completely at random assumption, but multiple imputation and full information maximum likelihood estimations do not reveal differences.

Policy implications

If well-being inequality is a clearer indicator of life's circumstances than mean well-being and GDP, and if the objective of policymakers is to improve the lives of all individuals, then more evidence is needed to guide the development of policies that can address it. To achieve the sustainable development goals for all people, a greater understanding of inequities for specific groups of people is essential.

Our findings suggest that policymakers should shift their focus from policies aimed at the mean level of well-being to those aimed at the full distribution of well-being. This includes a particular emphasis on health, social relationships, and education. Moreover, our results show that policy proposals should contain statements of the distributional effects on various populations.

AUTHOR CONTRIBUTIONS

Michael D. Smith: Conceptualization, methodology, writing—original draft, writing—review & editing. Dennis Wesselbaum: Conceptualization, methodology, formal analysis, writing—original draft, writing—review & editing.

ACKNOWLEDGMENT

Open access publishing facilitated by University of Otago, as part of the Wiley - University of Otago agreement via the Council of Australian University Librarians.

    CONFLICT OF INTEREST STATEMENT

    The authors declare no conflicts of interest.

    SPONSOR'S ROLE

    No specific funding was received for this work.

    FINANCIAL DISCLOSURE

    No specific funding was received for this work.