The Effect of Income Inequality on Happiness Inequality in India: A Recentered Influence Function Regression Estimation and Life Satisfaction Inequality Decomposition

Abstract

It is now an accepted stylised fact that increase in happiness level in any country is not commensurate with growth in income, a puzzle known as Easterlin Paradox. This paper analyses the income-happiness relationship in India and tries to explain the flat happiness response to income change in terms of rising income inequality. Income growth propels inequality and so also inequality in well-being. Empirically the effects of income inequality, absolute income, relative income, rank position and social capital indices are analysed using World Value Survey data for 12 states of India over 24 years from 1990 to 2014. As the variation in the 10-point scale measure of life satisfaction level is narrow, an recentered influence function (RIF) regression of variance and Gini of life satisfaction are estimated. The life satisfaction inequality is decomposed into composition and coefficients effects using Blinder–Oaxaca (B–O) decomposition method. The estimated RIF coefficients reveal significant effects on life satisfaction inequality of various income measures and social capital indices. The B–O decomposition shows that the functional relationship between material aspirations and life satisfaction, contribute significantly to rising life satisfaction inequality relative to changes over time in happiness influencing factors. Reducing income inequality and improving trust, sociability, health, education and employment over time and space could reduce life satisfaction inequality and improve happiness level in India.

Keywords

Happiness inequality income inequality relative income social capital recentered influence function regression Gini and variance decomposition

Introduction

The relationship between income and happiness is much debated since the seminal contribution of Easterlin (1974). In his path-breaking analysis of US data, Easterlin finds that though income-life satisfaction relationship is positive at the individual level, at an aggregate level, the relationship is weak. At cross-country level, despite vast differences in income across countries, the happiness and life satisfaction levels are pretty much close in developed, developing and transition economies. Moreover, the long-run relationship between aggregate income and average happiness is either nil or flat in almost all economies and there is no systematic relationship between income growth and growth in subjective well-being in any country. The puzzle, known as the Easterlin Paradox, that life satisfaction is not commensurate with income and even tends to fall with income growth in the long-run is initially explained by Easterlin (1974) himself in terms of income adaptation and social comparison. It is argued that people adapt to changes in income and adjust to their relative position in the society. Under income adaptation, as income increases, so also aspirations and hence individuals do not gain extra happiness. Under social comparison, when everyone’s income is growing, the relative position of individuals remains unchanged in the income ladder and hence no rise in aggregate happiness level.

Later cross-country attempts to figure out the income-happiness relationship at an aggregate level and over time show that economic growth-life satisfaction gradient is slightly positive, at least not zero (Deaton, 2008; Stevenson & Wolfers, 2008). Such findings are attributed to improved long series data and better estimation methods. However, Easterlin and his co-authors (Easterlin, 2015; Easterlin & Angelescu, 2010; Easterlin & Sawangfa, 2010; Easterlin et al., 2010) continue to argue that even using much more long-run data could not disprove his earlier findings that there is no direct relationship between economic growth and happiness level, establishing that the paradox is not yet lost, in fact, regained. Some research looks into the social dimensions of life satisfaction suggesting that social capital, such as trust and social networks, could explain the flat relationship between income and happiness (Bartolini & Sarracino, 2014; Becchetti et al., 2012; Bartolini et al., 2013; Bjornskov, 2008; Chang, 2009; Churchill & Mishra, 2016; Graham, 2017; Helliwell et al., 2017; Portela et al., 2013; Ram, 2010). But, empirical studies find a declining or weakening social capital in almost all societies in recent years, and so social capital does not explain life satisfaction directly. Instead, there may be some indirect effect of social capital on happiness through its moderating effects on the individual as well as relative income.

Another plausible reason for the absence of long-run income-happiness relation could be increasing income inequality when income grows (Becchetti et al., 2011; Clark et al., 2014; Kalmijn & Veenhoven, 2005; Schneider, 2012). As such a rise in income inequality enlarges rich-poor gap in other aspects of life and hence majority poor feel unhappy. This inequality in income could be the cause of happiness inequality. Recent empirical evidences suggest that a continuous increase in income inequality may negatively influence mental and physical health, ultimately affecting human happiness (Berg & Veenhoven, 2010; OECD, 2012; Oshio & Kobayashi, 2010; Schneider, 2012; Verme, 2011; Zagorski et al., 2014). Ovask and Takashima (2010) stress the necessity of keeping life satisfaction inequality at a minimum so that everyone can draw benefits from an increase in average life satisfaction. This is because an increase in average life satisfaction does not mean an increase in everyone’s life satisfaction when life satisfaction inequality is increasing (Graafland & Leus, 2018). Ifcher et al. (2018) and Yang et al. (2019) find a positive relationship between income inequality and life satisfaction inequality.

With data based on 93 years of study on income inequality, Chancel and Piketty (2019) show that income inequality is rising sharply in India since the 1980s. Among the Indian population, the income share of top 1 per cent has been 21.3 per cent in 2014–2015, up from 6.2 per cent in 1982–1983. But, the share of middle 40 per cent income groups in national income had fallen to a low level of 29.2 per cent in 2014–2015, down from 40 per cent in 1980. Similarly, the income share of bottom 50 per cent in national income has decreased to 14.9 per cent in 2014–2015 from 20.6 per cent in 2000–2001. Chancel and Piketty also note that the top 0.1 per cent income earners of about 8,00,000 individuals in 2014–2015 alone who capture 28 per cent of total national income. The dramatic increase in income inequality threatens the not only consumption and wealth distribution but also the well-being distribution. In this situation, it is important to understand the causal effect of rising income inequality on life satisfaction and its distribution across space and time in India.

Not only income inequality has been rising, but also there has been a sharp rise in consumption and wealth inequality as well in India. Both consumption and wealth data from National Sample Surveys show an increasing concentration at the top end of the distribution, at the cost of other population. The wealth share of top 10 per cent has increased from 52 per cent in 1992 to 63 per cent in 2012. The consumption share of top 10 per cent has increased from 27 per cent in 1993–1994 to 31 per cent in 2011–2012. The trend in the distribution of income, consumption and wealth distributions show a changing pattern not only over time, but over space. There has been an increasing concentration in urban areas. While population share of rural areas remains stagnant, its share in income, consumption and wealth has been declining, significantly worsening the rural–urban and the rich–poor gaps in lifestyles and living standards.

The inequality in subjective well-being and its relation with income inequality is relatively unexplored areas in happiness literature. It is the focus of this paper to analyse the effect of income inequality on life satisfaction inequality at an individual level over a long period of 24 years between 1990 and 2014. In the Indian context, this paper tries to answer the questions: What is the nature of life satisfaction inequality in India? What are the significant determinants of life satisfaction inequality? What factors affect change in life satisfaction inequality over time? Does income inequality correlate with life satisfaction inequality? Is the effect of income inequality on life satisfaction inequality direct or indirect? If indirect, does it operate through individual income or relative income and/or both? In the empirical analysis, this paper follows a novel estimation method, the recentered influence regression (RIF) method (Firpo et al., 2018; Fortin et al., 2011). Further, a decomposition analysis is also tried to evaluate the causes of an increase in life satisfaction inequality over time in India. The empirical analysis is based on the three waves (1990, 2001, and 2010) data of World Value Survey (WVS) of India.

Income Inequality and Subjective Well-Being Inequality Relationship

A voluminous happiness literature has so far analysed the determinants of happiness and its relationship with income. In contrast, both empirical and theoretical discussion on happiness inequality and its relationship with income inequality is rather scanty. Although there have been few attempts to analyse happiness inequality at a macro level using cross-country data, they do not focus on the relationship between income inequality and happiness inequality. Studies at micro-level either on happiness inequality or on its association with income inequality is virtually nil, especially in developing country contexts. Only recently the issue has received some attention among happiness researchers.

An analysis of life satisfaction inequality is important in different ways. On the one hand, discussion on life satisfaction inequality can significantly influence the debate of the consequences of income inequality on well-being (Fehr & Schmidt, 1999; Ferrer-i-Carbonell, 2005). On the other hand, identifying the drivers of happiness inequality might suggest needful instructions for policy makers regarding strategies on social cohesion and well-being (Becchetti et al., 2011). This is because of the large gap in happiness and life satisfaction within countries is considered as a threat to social cohesion or social and political stability (Guimaraes & Sheedy, 2012).

Life satisfaction inequality can be a better predictor of social harmony than any other economic variable like income because there are theoretical explanations like discontent and expected utility theories that show life satisfaction gap and social unrest are moving in the same direction. According to the former theory, a decrease in life satisfaction has a strong and independent effect on social disruption (Gurr, 1994). According to the latter, rational individuals participate in social protest if only expected gains are higher than the costs involved in their choice (Tullock, 1971). The expected gains are substituted by the satisfaction gap between happy and unhappy people multiplied by the probability of success in riots (Becchetti et al., 2014). In this context, it can be considered that life satisfaction inequality may cause directly envy and social unrest, while income inequality is an indirect cause.

But, Schneider (2012) vouches for the flip side of income inequality. Whether the income gap affects happiness depends on how individuals treat it by considering their personal and social situations. If income inequality is treated as a sign of high potential and social mobility, then poor income people might treat inequality as an opportunity to improve their situation in future, so that income inequality has a positive side. A social group may be economically poorer than another group in a society but still they are happy because of other factors, then economic gap will not create social upheavals (Becchetti et al., 2014). Put in other terms, the income divide does not translate into happiness divide always due to the moderating effect of many other non-economic factors affecting happiness.

Early macroeconomic evidence on happiness inequality has come from a correlational analysis of European Union countries for 21 years (1973–2001) by Veenhoven (2005) which shows a falling trend in happiness inequality in modernised economies. Veenhoven (2005) further derives the same finding more rigorously on a comparative analysis of 53 countries during the 1990s. In addition to this, Veenhoven notes that there is a weak relationship between income inequality and a standard deviation of life satisfaction. A cross-country analysis by Ovaska and Takashima (2010) on the determinants of happiness inequality suggests that income inequality and differences in health status are positively associated with happiness inequality, and poor institutional quality of a country widen the existing situation. Based on a correlational analysis of 78 countries during 1999–2001, Ott (2005) finds that institutional factors like government consumption, transfers, subsidies, social security, etc., contribute positively to happiness and reduce happiness inequality.

As regards the microeconomic analysis of happiness inequality, Stevenson and Wolfers (2008) analyse the relationship between level and dispersion of happiness in the US using General Social Survey (GSS) data over 1972–2006. The study finds that happiness inequality in the US has reduced substantially in the 1970s and 1980s, but then increased, reversing about one-third of the decline in the initial period. The decomposition analysis shows that the main determinants of happiness inequality in the US are the changes in happiness dispersion within gender and racial groups. Further, the observed trend in happiness inequality is different from the observed trend in income growth and income inequality. They also find that non-economic factors strongly affect happiness distribution than economic factors.

Becchetti et al. (2014) find an increasing trend of happiness inequality in Germany over 1992–2007 from German Socio-Economic Panel data (GSOEP). The empirical results indicate that income growth reduces happiness inequality in Germany, but income inequality does not have a significant effect on happiness inequality, a result consistent with the findings of Stevenson and Wolfers (2008) for the US. Further, the study finds that over time changes in education have a reducing effect on happiness inequality and higher unemployment rate increases happiness inequality at a higher rate. Clark et al. (2014), using different data sets over a long period of time (1971–2010), observe that nations with increasing GDP per capita experience decreasing happiness inequality despite the growth of income inequality and constant happiness levels. The regression results support the view that income inequality increases happiness inequality and income growth reduces the same.

Van Praag (2011) argues that relative position of individuals vis-a-vis others influences individual happiness and happiness inequality and suggests that how frequently a person would compare himself with others and the degree of social transparency in society should be taken into consideration in analysing happiness inequality. Becchetti et al. (2014) attempt to test the Van Praag’s relative position concept empirically by defining relative income in terms of whether the respondent is rich or poor compared to the reference group. They find a positive impact for being relatively poor and no impact for being relatively rich on happiness inequality. Niimi (2018) estimates a recentered influence function (RIF) regression and finds that household income has a significant negative impact on happiness inequality and hence relative standing is also important in determining inequality in happiness. Moreover, the fear of losing a job and life after retirement also matters for happiness inequality in Japan.

Data and Empirical Methodology

The empirical analysis of this paper on the relationship between income inequality and life satisfaction inequality in India is based on the WVS data pertaining to 12 states of India for 24 years over the period 1990–2014. The WVS, the largest cross-national survey on life satisfaction and well-being, has been conducted in 6 waves between 1980 and 2014. Since the second wave in 1990–1991, the WVS has been conducted in India. The WVS contains a wealth of information on socioeconomic and demographic characteristics of respondents like income, employment status, educational status, marital status, age, gender, religious faith, social relations, trust, political behaviour, and so on besides subjective well-being indicators in terms of life satisfaction, happiness, and financial satisfaction. This paper uses the sample of individuals aged 18 years and above in the second (1990), fourth (2001) and sixth (2014) waves of WVS in all 12 states of India, consisting of 5309 observations, 2400 in the second wave, 1668 in fourth wave, and 1221 in sixth wave. In the WVS, there are two main direct questions that indicate the self-reported subjective well-being of people. The first measure asks respondents to evaluate their present life in terms of “Taking all things together, would you say you are … very happy, quite happy, rather happy and not at all happy”, and record their own rating of happiness. The second measure asks for an evaluation of whole life: “All things considered, how satisfied are you with your life as a whole are these days?” for which respondents select an answer in a 10-point scale, starting with dissatisfied and ending with satisfied.

The ordinal nature of happiness and life satisfaction measures pose some methodological issues in the measurement of happiness and happiness inequality. An implicit assumption of any standard inequality statistic is that the variable in question is cardinal and continuous in nature. Also, it assumes equal distance between ratings so that interpersonal comparison is possible. The popular surveys that collect information on self-reported happiness report data that are categorical and ordinal in nature and WVS is also not an exception to it. Additionally, the self-reported scale of subjective well-being questions is not uniform across different individuals. Despite some drawbacks and criticisms, growing evidences from happiness literature shows that self-reported happiness and life satisfaction measures are meaningful, valid, reliable, and consistent in many ways, and most importantly, can be used for economic analysis. The self-reported happiness or life satisfaction measures are consistent with other important economic measures of utility. Further, happiness and life satisfaction move in the same direction with many different individual life events such as marriage, employment, etc., and with many macroeconomic indicators like GDP, aggregate unemployment rate and income inequality (Alesina et al., 2004; Di Tella & MacCulloch, 2006).

The survey questions on subjective well-being attempt to measure how individuals assess his overall life rather than his current feelings and as such they involve an evaluative judgment of life circumstances and conditions by the individual himself and self-reported accordingly. Measuring feelings and emotions, with which life evaluations are closely dependent, are very subjective and therefore reliability of those measures mainly depends on how meaningfully the survey respondent understands and answers the survey questions on happiness. It is to be noted here that happiness data is mainly used to identify significant determinants of happiness rather than to make comparison between different levels of happiness of people. Therefore, Stutzer and Frey (2010) argue that it is not necessary to make assumptions that happiness can interpersonally comparable or cardinally measurable.

Empirical analyses consider the reported subjective well-being values as possessing some cardinal properties and hence treat the measures as continuous. Ferrer-i-Carbonell and Frijters (2004) test both cardinal and ordinal assumption of happiness score and find that there is only a small difference in the estimated result of drivers of happiness, which is also supported by Frey and Stutzer (2002). Clark et al. (2014) use an index of ordinal variation as a measure for ordinal variables for robustness check and obtain almost similar results. These results are in parallel with the view of Van Praag (1991) that individuals tend to translate their verbal evaluations regarding their overall quality of life to a numerical scale when they answer to the subjective questions (Niimi, 2018).

As regards the heterogeneity in the scales at which the respondents evaluate their level of happiness, it is argued that such heterogeneity should not affect estimation because it is expected to be random (Frey & Stutzer, 2002; Di Tella & MacCulloch, 2006). Empirical studies also test heterogeneity in individual scale and find that respondents use a different scale when answering their welfare questions and it should not affect the estimated result for it is not an important source of bias in the estimation (Beegle et al., 2012).

Some research argues that the life satisfaction measure used in most surveys is very narrow and restrictive, and hence there is not much variation in well-being levels across samples or countries. Generally, life satisfaction measure, either average or absolute, is in the range of 1–10 and happiness measure is in the interval of 1–4. Moreover, the respondents may be biased to choose the middle values in their reporting, and therefore actual data in effect is only in the range of 5–7 or 2–3. Hence, at the aggregate or average level, there is not much deviation in the reported level of satisfaction between people and countries. Instead, some research suggests that variations, such as standard deviation, variance, or any measure of inequality like Gini, in life satisfaction should be analysed (Clark et al., 2014; Kalmijn & Veenhoven, 2005, 2016).

Kalmijn and Veenhoven (2005) investigate the applicability of various inequality matrices to quantify happiness dispersion. They use nine inequality measures by assuming cardinality across happiness categories. They find four measures of happiness inequality, namely, standard deviation, mean absolute difference, mean pair distance, and interquartile range are efficient measures to quantify happiness dispersion. Since there is no single superior metric to others among these four measures, Kalmijn and Veenhoven (2005) support the use of standard deviation as the measure of happiness inequality. Clark et al. (2014) use standard deviation as the measure of happiness inequality by assuming happiness is a cardinal variable. Following this, empirical studies use the standard deviation of happiness to quantify inequality of happiness. In line with literature, this paper also assumes that life satisfaction data is cardinal and standard deviation as the measure of happiness inequality.

Apart from life satisfaction inequality, this paper also uses individual (absolute) income, relative income, rank income and income inequality variables. As the WVS contains only categorical measure individual income as reported by respondents from 10 income brackets, following (Rousseau, 2009), absolute individual income is computed by assigning the midpoint of the income categories chosen by the survey respondents to each income category. The highest income bracket is computed by adding half of the difference between the top and lower bounds of previous income categories to the lower bound of the highest income category. In this way, the ordinal income variable is converted into a continuous variable and the log of individual income is used.

The relative or comparison income, the level of others’ income with which the respondent compares his income, is computed as the gap between individual income and the average income of a reference group (Yang et al., 2019). Following Becchetti et al. (2014), the reference group is defined as those individuals with the same age, educations and place of region. If individual income is greater than average reference group income, the individual is classified as relatively rich (y–y^r), and on the other hand, if individual income falls short of average income of the comparison group, the respondent is known as relatively poor (y^r–y) (Budria, 2013; Ferrer-i-Carbonell, 2005).

As regards to income inequality, two measures are computed. The first one is individual income below 60 per cent of median income and the second one is individual income above 200 per cent of median income. The rank income, measuring the relative status of an individual in the society, is defined as the rank position of an individual in income distribution within the comparison or reference group.

Social capital, as an indicator of social cohesion, is defined as a composition of two aspects of social behaviour, general trust and social relationship. An index of social capital has been constructed by summing up trust and social relationship dummies. The social capital index is a categorical variable assigning values {0,1,2}, where the highest value 2 means that an individual is trustful and sociable, if the value is 1, the individual is either trustful or sociable, and if the value is 0, then the individual is neither trustful nor sociable (Piekalkiewicz, 2016). The econometric estimation also includes other usual socio-demographic characteristics of respondents such as age, gender, education, employment, marital status, social status, health status, religion and region of residence.

Econometric Specification

Empirically the effects of income inequality on life satisfaction inequality in India is estimated using a distribution regression method, the RIF regression, a la Firpo et al. (2009, 2018), and changes in life satisfaction inequality over time is analysed using the Blinder–Oaxaca decomposition method. Similar methods are used to analyse happiness inequality in Germany by Becchetti et al. (2014), in Japan by Niimi (2018), and in China by Yang et al. (2019). Generally, the influence function (IF) is used to analyse the sensitivity of distributional statistics like variance or Gini to small disturbances in data. When the variability of a variable is too narrow, the RIF estimates the effect of changes in the distribution of independent variables on the unconditional distribution of the dependent variable (Firpo et al., 2009; Fortin et al., 2011). The RIF regression estimate is the partial effect of a small location shift in the distribution of covariates on some distributional statistic of the dependent variable.

Assuming that the dependent variable life satisfaction has a distribution Φ, define v(Φ) as a distributional statistic of LS (such as mean, variance, Gini, quantile, and so on). The distribution regression discusses the influence of explanatory variables x on v(Φ). It has two parts: how does the distributional statistic v(Φ) relate with changes in x, and how much does the relationship between distributional statistic v(Φ) and its influencing factors change over two groups or two time periods? In the decomposition terminology, the former is known as composition effect (changes or differences in characteristics) and the latter is known as a coefficient effect (changes or differences in coefficients) (Yang et al., 2015). When the distributional statistic v(Φ) is mean, then the classical linear regression model can solve the issues. However, when the distributional statistic is other than mean, like Gini, variance or quantiles, then the issue is complicated, and therefore, OLS is not straight forward (Firpo et al., 2009).

The RIF regression method finds solution for the partial effect, the marginal effect of x on life satisfaction inequality, and policy effect, the effect of changes in x on the changes in life satisfaction inequality during the period (2010–2014) compared to the period (1990–2001). Given the outcome variable, life satisfaction y and its cumulative distribution function Φ and probability density function φ, the vector of information required for analysing distributions can be more briefly written as a set of ordered pair:

Φ_{y} = {[y, Φ_{y} (y)] | y} o r φ_{y} = {[y, φ_{y} (y)] | y} .

(1)

Let v(.) be a distributional statistic or functional to be estimated using the information contained in the vector (y, Φ_y, φ_y). The impact of a change in the distribution of y on the distributional statistic v(.), Δv, is generated by comparing the observed and ex post CDFs of y:

Δ ν = ν (ψ_{y}) - ν (Φ_{y})

(2)

where ψ_y and its PDF f_y have the same properties as Φ_y and φ_y. As the magnitude of the change Δv depends on the magnitude of change when moving from Φ_y(.) to ψ_y(.), Δv is standardised by the change of the distribution:

Δ^{s} (ν) = \frac{Δ ν}{Δ (ψ_{y} - Φ_{y})} = \frac{ν (ψ_{y}) - ν (Φ_{y})}{Δ (ψ_{y} - Φ_{y})} .

(3)

As the life satisfaction measure is categorical and ordered, with each stated level of satisfaction has a specified value y^*, the CDF of y can be characterised as:

Ω_{y^{c}} (.) = 0 \forall y 〈 y^{*} a n d ​ ​ ​ Ω_{y^{*}}^{}_{(.)}^{} = 1 \forall y 〉 y^{*} .

(4)

As the mass of the distribution $Ω_{y^{*}}$ (.) is centred at the specified value y^*, the distribution of Ω_y is obtained by combining Φ_y(.) and $Ω_{y^{*}}$ (.) as:

Ω_{y} = (1 - λ) Φ_{y} (.) + λ S_{y}^{}_{*}^{} (.)

(5)

where λ is the change in the distribution when Φ_y(.) → ψ_y(.). Then, the influence function IF is given by (Hampel, 1974):

I F [y^{*}, v (Φ y)] = \frac{\lim}{λ \to 0} \frac{v [(1 - λ) Φ_{y}] - v (Φ_{y *})}{λ} = \frac{\partial v (Φ_{y} \to Ω_{y *})}{\partial λ} .

(6)

Thus, the IF as a derivative shows the change in the distributional statistic v for a small change in the distribution Φ_y that gives more weight to observations with values y^*. The properties of IF are:

\int^{​} I F [y; ν (Φ_{y})] d Φ_{y} = 0

(7)

ν (Φ_{y}) ~ N [ν (Φ_{y}), \frac{σ_{I F}^{2}}{N}]

(8)

σ_{I F}^{2} = \int^{​} I F [{[y; ν (Φ_{y})]}^{2} d Φ_{y} = 0

(9)

RIF can be specified as:

R I F [y_{i}; ν (Φ_{y})] = ν (Φ_{y}) + I F [y_{i}; ν (Φ_{y})] .

(10)

The RIF has an expectation equal to the original function:

\int^{​} R I F [y; ν (Φ_{y})] d Φ_{y} (y) = \int^{​} {ν (Φ_{y}) + I F [y_{i}; ν (Φ_{y})]} d Φ_{y} (y) = ν (Φ_{y}) .

(11)

The RIF can be used directly to estimate the standard errors of the distribution statistic, variance or Gini, and in decompositions.

The unconditional partial effects of small changes in the distribution of life satisfaction on the distributional statistic v[Φ(y)] can be estimated using iterated expectations. The statistic v can be rewritten as:

ν [Φ_{y} (y)] = \int^{​} R I F [y; ν (Φ_{y})] d Φ_{y} (y) = ν [Φ_{y} (y)] + \int^{​} I F [y; ν (Φ_{y})] d Φ_{y} (y) .

(12)

In terms of distribution of independent variables, this function can be rewritten as:

v [Φ_{y} (y)] = v [Φ_{y} (y)] + \iint^{​} I F [y; ν (Φ_{y})] d Φ_{y | x} (y | x = x^{*}) d Φ_{x} (x) .

(13)

Alternatively,

v [Φ_{y} (y)] = \int^{​} E {R I F [y; ν (Φ_{y})] | x = x^{*}} d Φ_{x} (x) .

(14)

The RIF, compared to the standard linear regression model, uses the estimated RIF[y;v(Φ_y)] for each y_i as the dependent variable to relate Δv_y to changes in ΔΦ_x. For the functional variance, the RIF can be specified as:

R I F (y, σ_{y}^{2}) = ν^{ν} [Φ_{y} (y)] = {(y - μ_{y})}^{2} w h e r e σ_{y}^{2} = \int^{​} {(y - μ_{y})}^{2} d Φ_{y} (y) .

(15)

For the distributional statistic Gini, the RIF can be specified as:

R I F (y, G i n i_{y}) = ν^{G} [Φ_{y} (y)] = 1 + \frac{2}{μ_{y}^{2}} Γ_{y} - \frac{2}{μ_{y}} {y [1 - Φ_{y} (y)]} = 1 - 2 μ^{- 1} Γ_{y}

(16)

where $Γ_{y} = \int_{0}^{1} Λ_{y} (p) d p$ with p(y) = Φ_y(y) and $Λ_{y} (p) = \int_{- \infty}^{q_{y} (p)} y d Φ_{y} (y)$ .

Then, the RIF regression can be specified as:

R I F [y_{i}; ν (Φ_{y})] = β x_{i} + ε_{i} .

(17)

Taking expectations:

v (Φ_{y}) = E {R I F [y; ν (Φ_{y})]} = E (β x_{i}) + E (ε_{i}) = β \bar{x} a s E (ε_{i}) = 0.

(18)

The unconditional partial effect is given by:

\frac{\partial ν (Φ_{y})}{\partial {\bar{x}}_{k}} = β .

(19)

The unconditional partial effect, for an average person, is the increase in life satisfaction y for a one unit increase in x.

Next is to calculate the change in life satisfaction inequality from one period (1990–2001) to another period (2010–2014):

Δ ν = (ν_{1} - ν_{0}) = ν (Φ_{y}^{1}) - ν (Φ_{y}^{0})

(20)

Δ ν = ν {\int^{​} Φ_{y | x}^{1} (y | x) d Φ_{x}^{1} (x) -} - ν {\int^{​} Φ_{y | x}^{0} (y | x) d Φ_{x}^{0} (x)} .

(21)

Thus, the differences in the statistic Δv arises due to differences in the distribution of independent variables [ $d Φ_{y}^{1} (x) \neq d Φ_{y}^{0} (x)$ ] and differences in the relationship between y and x [ $Φ_{y | x}^{1} (y | x) \neq Φ_{y | x}^{0} (y | x)$ ], equivalent to the standard Blinder–Oaxaca decomposition method of comparing differences in average characteristics and differences in coefficients (Blinder, 1973; Oaxaca, 1973). The two components of the overall gap in the distributional statistic Δv can then be calculated as:

Δ ν = (Δ ν_{s} + Δ ν_{x}) = (ν_{1} - ν_{c}) + (ν_{c} - ν_{0})

(22)

where Δv_x is the life satisfaction gap due to differences in characteristics and Δv_s is the differences due to the relationships between y and x. The counterfactual v_c is defined as:

ν_{c} = ν (Φ_{y}^{c}) = ν {\int^{​} Φ_{y | x}^{0} (y | x) d Φ_{x}^{1} (x)} .

(23)

To identify the counterfactual statistic, either two separate RIF regressions for each time period can be estimated or the standard Blinder–Oaxaca decomposition be followed. From the former method, $v_{c} = β^{0} {\bar{x}}^{1}$ , and the latter method, $Δ ν_{x} = β^{0} ({\bar{x}}^{1} - {\bar{x}}^{0})$ and $Δ ν_{s} = {\bar{x}}^{1} (β^{1} - β^{0}) .$

The application of decomposition method in happiness literature is rare. Earlier, Ball and Chernova (2008) applied the conventional Blinder–Oaxaca decomposition strategy to identify the relative contribution of relative income and absolute income to subjective well-being.

Empirical Analysis

Figure 1 presents life satisfaction inequality and income inequality in states of India during the 24 years period over 1990 and 2014. Despite initial fluctuations of life satisfaction inequality, there is an overall upward trend with a 5 per cent increase over this period, whereas the standard deviation of lnNSDPpc increased by 3 per cent. The income inequality (SD of log NSDP per capita) lies below the life satisfaction inequality. The log of NSDP per capita increased steadily with an overall increase of 1.94 per cent between 1990 and 2014. Thus, economic growth not only contributes to increasing income inequality, but also to well-being (happiness/life satisfaction) inequality in India over period of fast growth.

Further, Figure 1 shows that life satisfaction inequality increased drastically from 2006 onwards. The standard deviation of life satisfaction has increased during 1990 to 2006 by 4 per cent, but from 2006 to 2014, the overall change is around 14 per cent. Hence, following the framework of Becchetti et al. (2014), Firpo et al. (2018) and Yang et al. (2019), this paper divides the 24 years period (1990 to 2014) into two parts: a period of moderate life satisfaction inequality (1990–2001) and a period of higher life satisfaction inequality (2010–2014) and performs a comparative analysis of the causes for rise of life satisfaction inequality in the latter period.

Figure 2 depicts the relationship between income inequality (Gini coefficient) and life satisfaction inequality (standard deviation) in states of India for the two time periods (1990–2001 and 2010–2014). In both time periods, there is indeed a strong positive relationship between them. In second period, the relationship is steeper than in the first period. Thus, life satisfaction inequality has increased in India over the 24 years as income inequality increases. Figure 2 also reveals higher dispersion of life satisfaction in the latter period than in the first period.

Tables 1 and 2 reveal significant changes in average life satisfaction, life satisfaction inequality and income inequality between the periods (1990–2001) and (2010–2014). The life satisfaction level in India has decreased by 14.02 per cent over time, from 5.99 to 5.16. The life satisfaction inequality has increased over the period 1990–2014, as the variance has increased by 33.51 per cent, from 5.64 to 7.53, and the Gini index increased by 36.94 per cent, from 0.22 to 0.30. At the same time, average household income in India has increased by 18.49 per cent, the average log of income increased from 10.16 to 10.18. But the gap between rich and poor has increased by 14.39 per cent (Gini) and the variance of household income has increased by 72.35 per cent during this 24 years’ time period. These trends in India are somewhat higher compared to other countries. Becchetti et al. (2014) find a slight decrease (2.5 per cent) in average happiness and around 8 per cent increase in variance of happiness and 7 per cent increase in Gini index of happiness in Germany between 1992 and 2007, while Niimi (2018) finds a downward trend in happiness inequality (standard deviation) by 7.2 per cent in Japan between 2003 and 2013. The rank income position has only slightly improved between the periods, from 0.40 to 0.42. The average difference between rich and median income of reference group has reduced from 0.183 in 1990–2001 to 0.174 in 2010–2014 and the average gap between poor and average reference group income has also decreased from 0.469 (1990–2001) to 0.370 (2010–2014).

Figure 1.

Source: Life satisfaction: all five waves of WVS (1990–2014). NSDP per capita: National Accounts of India.

Figure 2.

Source: Authors’ calculation based on the WVS and National Accounts of India.

Table 1.

Change in Life Satisfaction, Income Inequality and Life Satisfaction Inequality in India, 1990–2014

Measure	Life Satisfaction				Household Income
Measure	1990–2001	2010–2014	Change	Percentage Change	1990–2001	2010–2014	Change	Percentage Change
Mean	5.99	5.16	−0.83	−14.02	3.73	4.42	0.69	18.49
Gini	0.222	0.304	0.082	36.94	0.278	0.318	0.04	14.39
Variance	5.64	7.53	1.89	33.51	3.58	6.17	2.59	72.35

Source: Authors’ calculation based on the WVS and National Accounts of India.

Table 2.

Descriptive Statistics of Variables Over Time in India, 1990–2014

Covariate	1990–2001	2010–2014	Covariate	1990–2001	2010–2014
LS	5.99 (2.37)	5.16 (2.74)	LS Gini	0.222 (0.070)	0.304 (0.012)
ln (income)	10.16 (8.93)	10.18 (8.04)	Rank income	0.403 (0.306)	0.417 (0.309)
Rich	0.183 (0.363)	0.174 (0.308)	Poor	0.469 (0.616)	0.370 (0.535)
SC index = 1	0.143 (0.350)	0.097 (0.296)	SC index = 2	0.580 (0.493)	0.612 (0.487)
SC index = 3	0.275 (0.447)	0.289 (0.453)	Female	0.451 (0.497)	0.381 (0.485)
Male	0.548 (0.497)	0.618 (0.485)	Age	37.77 (14.16)	40.35 (14.48)
School education	0.176 (0.381)	0.688 (0.463)	Secondary education	0.445 (0.497)	0.168 (0.374)
College education	0.377 (0.484)	0.142 (0.349)	Employed	0.516 (0.499)	0.524 (0.499)
Unemployed	0.483 (0.499)	0.475 (0.499)	Single	0.241 (0.428)	0.176 (0.381)
Married	0.758 (0.428)	0.823 (0.381)	Health—good	0.632 (0.482)	0.711 (0.453)
Health—fair	0.279 (0.448)	0.203 (0.403)	Health—poor	0.088 (0.283)	0.084 (0.278)
Upper class	0.080 (0.272)	0.152 (0.359)	Middle class	0.554 (0.497)	0.591 (0.491)
Working class	0.178 (0.382)	0.184 (0.387)	Poor class	0.157 (0.364)	0.060 (0.238)
Hindu	0.855 (0.351)	0.688 (0.463)	Muslims	0.088 (0.284)	0.156 (0.363)
Christians	0.030 (0.171)	0.091 (0.289)	Others	0.025 (0.156)	0.063 (0.243)
South India	0.272 (0.445)	0.252 (0.434)	East India	0.235 (0.424)	0.129 (0.335)
North India	0.251 (0.434)	0.235 (0.424)	West India	0.239 (0.426)	0.382 (0.486)

Source: Authors’ calculation based on the WVS.

Note: Standard deviations in parentheses.

The share of people with social capital index = 1 (neither trustful nor sociable) has reduced from 14 per cent to 1 per cent over 24 years, whereas the share of respondents who score social capital index = 2 (either sociable or trustful) has increased from 58 per cent to 61 per cent. Around 28 per cent of total respondents score social capital index = 3 (both sociable and trustful) in 1990–2001 and almost similar per cent of respondents (27.5) are both socially connected and trustful during the first period (1990–2001). The per cent of respondents who qualified with school level of education has increased in the second period (68.8 per cent) from first period (17.6 per cent), whereas the per cent of college educated people has decreased from 38 per cent (1990–2001) to 14 per cent in second period (2010–2014). More people have become rich in the latter period; the per cent of upper class people has doubled from just 8 per cent in the period 1990–2001 to 15 per cent in the period 2010–2014, whereas the per cent of working class people has remained stagnant at about 17 per cent between the periods.

In the econometric analysis, the individual level life satisfaction inequality is estimated by the RIF regression is estimated separately for 1990–2001 and 2010–2014, using two inequality indices, the Gini coefficient, the standard measure of inequality, and the variance of life satisfaction. In the next step, the Blinder–Oaxaca decomposition technique has been applied to decompose the life satisfaction inequality difference between the two periods in order to understand the causes for the sharp increase in life satisfaction inequality after 2006.

The RIF regression estimates of the variance and Gini of life satisfaction presented in Table 3 show that the income variables, both absolute and relative income, significantly influence life satisfaction inequality in both time periods, 1990–2001 and 2010–2014. An increase in absolute income, log of household annual income, significantly reduces the level of life satisfaction inequality on both indices. However, the effect of income on life satisfaction inequality is comparatively higher during the period 1990–2001. A unit increase in absolute income decreases variance of life satisfaction by 1.02 units in period 1990–2001, whereas the decrease in life satisfaction Gini during 1990–2001 is 0.916 units. The income effect on variance of life satisfaction in the second period (2010–2014) is insignificant but the same for life satisfaction Gini is negative and significant. Thus, an increase in individual income significantly reduces life satisfaction inequality in India, a result consistent with absolute income hypothesis that a rise in income is associated with increasing level of happiness of people.

Table 3.

RIF Regression Estimates of Life Satisfaction Inequality in India

Variable	Variance of LS		Gini of LS
Variable	1990–2001	2010–2014	1990–2001	2010–2014
Female	−0.498 (0.315)	−0.015^** (0.007)	0.676 (0.591)	0.018 (0.019)
Age	0.021 (0.014)	0.00 (0.0003)	0.014 (0.015)	0.0007 (0.00)
Secondary education	−0.890^** (0.461)	−0.045^*** (0.012)	0.065 (0.653)	−0.030 (0.020)
College education	−0.191 (0.551)	−0.037^*** (0.012)	0.680 (0.794)	−0.020 (0.018)
Married	−0.125 (0.381)	−0.001 (0.009)	0.391 (0.588)	0.010 (0.023)
Ln (income)	−1.02^*** (0.306)	−0.0003 (0.01)	−0.916^** (0.443)	−0.044^*** (0.013)
Relatively rich	1.34^** (0.627)	0.029^* (0.016)	0.882 (1.07)	0.006 (0.035)
Relatively poor	0.113 (0.220)	0.0001 (0.005)	0.556 (0.428)	0.032^** (0.015)
Rank income	0.608 (0.897)	−0.059^*** (0.021)	−0.210 (0.878)	−0.047^* (0.027)
Living in south India	−1.76^*** (0.385)	−0.034^*** (0.001)	0.438 (0.854)	0.063^*** (0.021)
Living in north India	−0.007 (0.453)	0.003 (0.010)	5.45^*** (0.995)	0.179^*** (0.029)
Living in west India	−1.22^*** (0.415)	−0.022^** (0.009)	2.47^*** (0.870)	0.122^*** (0.021)
Employed	−0.389 (0.301)	−0.017^** (0.007)	0.312 (0.527)	0.037^** (0.016)
Health—food	−2.18^*** (0.772)	−0.100^*** (0.019)	0.832 (0.851)	−0.0008 (0.029)
Health—fair	−1.72^** (0.875)	−0.062^*** (0.019)	1.61 (0.995)	−0.033 (0.033)
Social class—upper	−0.708 (0.711)	−0.026^* (0.015)	1.02 (0.130)	−0.001 (0.046)
Social class—middle	−0.653 (0.511)	0.014 (0.013)	0.827 (1.11)	−0.015 (0.042)
Social class—working	−0.603 (0.601)	−0.018 (0.015)	2.94^** (1.25)	0.036 (0.045)
Religion—Muslim	−0.173 (0.541)	0.010 (0.014)	0.796 (0.643)	0.046^** (0.022)
Religion—Christian	0.169 (0.640)	0.012 (0.015)	1.01(0.814)	0.017 (0.026)
Religion—others	−0.677 (0.892)	−0.019 (0.018)	−0.245 (1.18)	0.054 (0.039)
Social capital index = 1	−0.409 (0.482)	−0.018^* (0.011)	0.875 (0.713)	0.001 (0.025)
Social capital index = 2	−0.870^* (0.509)	−0.036^*** (0.011)	1.18 (0.805)	0.029 (0.027)
Constant	19.30^*** (3.07)	0.412^*** (0.077)	9.80^** (4.41)	0.591^*** (0.142)
R ²	0.0461	0.0934	0.1248	0.1492

Source: Authors’ estimation based on the WVS.

Note: Bootstrap standard errors with 200 replications in parentheses. ***, **, and * indicate significance respectively at 1, 5, and 10 per cent level.

With respect to the effect of relative income, the relatively rich position has a significant impact on life satisfaction inequality in both periods. As the gap between rich and the reference group average income increases, the variance of life satisfaction inequality increases by 1.34 units in the period 1990–2001 and by 0.029 units in the period 2010–2014. The effect of increasing gap between poor and reference group income on life satisfaction Gini has a significant effect by 0.032 units in second period (2010–2014). This result is consistent with the findings of Becchetti et al (2014) and Niimi (2018). Becchetti et al. (2014) find a positive effect for relatively poor on happiness dispersion and negative and significant effect for relatively rich on happiness inequality. Similarly, Niimi (2018) finds that both being rich and poor relative to being average are positively related with happiness inequality. As observed in previous literature and descriptive analysis, income inequality significantly influences life satisfaction inequality. The income inequality or relative income defined by the difference between reference group average income and individual income has an enlarging effect on life satisfaction inequality in India.

With regard to rank position in income distribution, the estimated results show that a higher economic status reduces life satisfaction inequality during 2010–2014 on both indices. Moving to a higher rank position in terms of income reduces the variance of life satisfaction in 2010–2014 by 0.059 units, while the life satisfaction Gini reduces by 0.047 units in 2010–2014. Therefore, with an increase of income inequality, indicated by either relatively poor or relatively rich, people presume that their relative economic status is low and therefore increasing income inequality can raise life satisfaction inequality. The same line of finding has been observed in China by Yang et al. (2019).

Compared to unemployment, employment significantly reduces variance of life satisfaction during 1990–2001, whereas employment effect on life satisfaction Gini is positive. The employment effects are significant only in the second period, 2010–2014. The change in employment effect is in consonance with rising employment opportunities during the 1990s and falling employment opportunities after the financial crisis of 2007–2008. Rising unemployment has slowed the decrease in life satisfaction inequality, and as the estimates indicate even raised the Gini index. Therefore, employment can effectively reduce inequality of life satisfaction in a society. Being a healthy person significantly reduces variance, but not the Gini, of life satisfaction in both time periods and the impact is comparatively higher for people with good health than fair health people. As long as health improves, life satisfaction inequality (variance) significantly reduces. But, such effect declines over time for both good and fair health status.

As for education, having higher secondary level education significantly reduces variance of life satisfaction in both time periods, compared to 10th standard schooling. In the first period (1990–2001), relative to 10th class education, variance of life satisfaction for pre-degree level of education reduces by −0.890 units, but that impact reduces only by half in the second period (2010–2014), by −0.045 units. For college level education, the effect on variance of life satisfaction inequality is insignificant in first period, whereas in second period the effect of higher level of education significantly reduces variance of life satisfaction by 0.037 units. This indicates that higher education is effective in reducing life satisfaction inequality in a society. The effect of education on life satisfaction Gini is insignificant in both periods.

The social capital index also significantly influences variance of life satisfaction in both time periods. Respondents with higher score of social capital index (SC index = 3) have lower level of life satisfaction inequality (variance) compared to people who are neither sociable nor trustful (social capital index = 0). But, the level of significance as well as the size of the coefficients changes over time. The insignificant (−0.409) impact of social capital index = 1 reduces and becomes significant (−0.018) in the second period. Similarly, the effect of social capital index = 2 on variance of life satisfaction in the first period is −0.870, which reduces to −0.036 in the second period. Thus, social capital is also one of the important influencing factors of life satisfaction and its inequality in India.

Living in south and west of India compared to living in eastern parts of India, reduces variance of life satisfaction inequality in both time periods, but the effect becomes positive and significant in both time periods for life satisfaction Gini. With regard to north India, the impact is insignificant for variance, but it becomes positive and significant for Gini, indicating the presence of strong well-being gap in north India compared to east India, and the impact is also high in the case of north India. Therefore, state level characteristics influence the life satisfaction inequality among people and it is subject to change over time. All other demographic and social characteristics such as age, marital status, social status and religion insignificantly influence life satisfaction inequality.

Thus, the RIF regression estimates reveal that socioeconomic variables affect subjective well-being inequality in India and such effects change over time in terms of size, sign and significance. The RIF regression analysis on life satisfaction inequality and its determinants reinforce the view that income measures like absolute income, relative income and income inequality statistically contribute greatly to life satisfaction inequality in India. At individual level, more life satisfaction may come, and inequality in happiness can reduce, from a sustained growth of economic environment that would enable people to have an improved quality of life. Having higher income may offer an individual a sense of protection, recognition in society and acceptability. Further, the level of education, employment, health, place of residence, social capital and economic variables significantly influence life satisfaction inequality in India. Overall, there is some difference between estimated effects of variance of life satisfaction and Gini of life satisfaction, where the effects are generally negative on the former and positive on the latter. Generally, an increase in economic status reduces deviations from average, but at the same time increases life satisfaction gap in India in the latter period. The rising income inequality during post reforms period also increased inequality in life satisfaction, especially post 2010. Once again, the estimated results show that Indians seek happiness and life satisfaction through material processions, and as material aspirations are driven mainly by income, the thrust for more money never ends, and in the end it leads to the treadmill of dissatisfaction in life.

Decomposition of Life Satisfaction Inequality

To identify the causes of increasing life satisfaction inequality during the latter period 2010–2014 in India, the life satisfaction inequality between the two periods 1990–2001 and 2010–2014 is decomposed into composition (changes in influencing factors) and coefficient (change in functional relationship between life satisfaction inequality and its influencing factors) effects, following the Blinder–Oaxaca decomposition. Table 4 presents the Blinder–Oaxaca decomposition estimates of changes in life satisfaction inequality in India during the 24 years from 1990 to 2014.

The decomposition results reveal an increase in variance as well as Gini coefficient of life satisfaction from 1990–2001 to 2010–2014. While both composition effect and coefficient effect have increased life satisfaction inequality indices, in absolute terms the coefficient effect is comparatively higher. Also, the composition effect for variance of life satisfaction is insignificant. Therefore, the increase in life satisfaction inequality in India during 2010–2014 compared to 1990–2001 is attributable much to significant changes in the functional relationship between life satisfaction inequality and its driving factors, rather than changes in covariates between the two periods. This result is different from that observed in Germany (Becchetti et al., 2014), but similar to China (Yang et al., 2019).

Table 4.

Decomposition of Life Satisfaction Inequality as Composition and Coefficient Effects

Variable	Variance of LS		Gini of LS
Variable	Composition Effect	Coefficient Effect	Composition Effect	Coefficient Effect
Female	−0.029 (0.021)	−0.416^* (0.230)	−0.001^* (0.001)	−0.012^* (0.007)
Age	−0.090^* (0.052)	0.294 (0.820)	−0.000 (0.001)	0.028 (0.024)
Medium education	−0.278^** (0.140)	−0.153 (0.132)	−0.014^*** (0.003)	−0.002 (0.004)
Higher education	−0.044 (0.119)	−0.125 (0.129)	−0.008^*** (0.002)	−0.002 (0.003)
Married	0.014 (0.040)	−0.430 (0.611)	0.000 (0.00)	−0.009 (0.018)
ln (income)	0.020 (0.036)	−1.12 (5.43)	0.000 (0.000)	0.451^*** (0.162)
Relatively rich	0.019 (0.020)	0.079 (0.215)	0.000 (0.000)	0.003 (0.006)
Relatively poor	0.009 (0.019)	−0.170 (0.187)	0.000 (0.000)	−0.012^** (0.005)
Rank income	0.006 (0.011)	0.340 (0.486)	−0.001 (0.001)	−0.004 (0.013)
Living in south India	0.012 (0.034)	−0.640^*** (0.285)	0.000 (0.00)	−0.028^*** (0.009)
Living in north India	−0.000 (0.016)	−0.904^*** (0.191)	0.000 (0.00)	−0.029^*** (0.006)
Living in west India	0.215^*** (0.075)	−1.62^*** (0.431)	0.003^** (0.001)	−0.063^*** (0.013)
Employed	−0.003 (0.008)	−0.375 (0.341)	0.005^*** (0.002)	−0.029^*** (0.010)
Health—good	0.127^** (0.054)	−2.16^*** (0.720)	0.006^*** (0.002)	−0.071^*** (0.021)
Health—fair	−0.147^** (0.058)	−0.627^*** (0.215)	−0.005^*** (0.002)	−0.005 (0.006)
Social class—upper	0.021 (0.022)	−0.251 (0.201)	0.001 (0.01)	−0.003 (0.006)
Social class—middle	0.002 (0.013)	−0.913 (0.736)	0.000 (0.000)	0.001 (0.023)
Social class—working	0.020 (0.021)	−0.666^*** (0.249)	0.001 (0.001)	−0.010 (0.007)
Religion—Muslim	0.015 (0.051)	−0.154(0.138)	−0.001 (0.001)	−0.005 (0.004)
Religion—Christian	−0.009 (0.042)	−0.082 (0.110)	−0.001 (0.001)	−0.000 (0.003)
Religion—others	0.023 (0.035)	−0.023 (0.080)	0.001 (0.001)	−0.004^* (0.002)
SC index = 1	0.011 (0.014)	−0.784 (0.578)	0.001 (0.000)	−0.012 (0.018)
SC index = 2	0.025 (0.021)	−0.618^** (0.309)	0.001 (0.001)	−0.019^** (0.009)
Total	−0.057 (0.272)	−2.04^*** (0.387)	−0.017^*** (0.006)	−0.079^*** (0.010)

Source: Authors’ computation based on the WVS.

Note: Bootstrap standard errors with 200 replications in parentheses. ***, **, and * indicate significance respectively at 1, 5, and 10 per cent level.

The composition effect of log of individual income on life satisfaction variance and Gini is not statistically significant, but the coefficient effect is significantly positive during 2010–2014. This implies that the rising life satisfaction inequality is mainly due to increasing income inequality during the period 2010–2014. Similarly, the coefficient effect of relatively poor is negative and statistically significant showing that uplifting the poor can reduce inequality in life satisfaction. There is been no significant composition effect on life satisfaction inequality. Further, the rank income has also no effect on changes in life satisfaction inequality during the period.

The next interesting result comes from the composition effect with respect to health status of respondents. Compared to respondents with poor health status, being a healthy person increases both variance as well as Gini of life satisfaction inequality. Similarly, being with fair health status significantly reduces both indices of life satisfaction inequality. This is because the share of healthy people increased over time. With respect to the coefficient effect of health status, good and fair health status significantly reduces variance of life satisfaction over time, consistent with RIF regression estimated results. The overall coefficient effect mainly comes from the impact of good health and constant coefficient estimates. Therefore, the over time difference in the impact of health status on life satisfaction inequality is more significantly related with the overall coefficient effect.

The same pattern can also be read with the composition effect of higher education on life satisfaction inequality. The decreasing share of highly educated people increases life satisfaction inequality. Compared to the share of school education, shares of both higher secondary and college educated people have declined in the second period. This reduction in share of higher level of educated people significantly reduces the variance as well as Gini of life satisfaction (composition effect of UG/PG for variance is insignificant). The overall negative composition effect mainly comes from the impact of higher secondary level of education: reducing share of higher secondary level of people reduces life satisfaction inequality in India a lot. Precisely, the decomposition estimates show that larger proportion of highly educated people increases life satisfaction inequality. This may be because education affects life satisfaction through higher aspirations for better paid job. If one cannot meet this aspiration, then higher education will negatively affect long-run life satisfaction (Majumdar & Gupta, 2015). In India, unemployment is a structural problem and with educated and industrial unemployment, a large number of educated people remain unabsorbed in the Indian labour market. Even if they get employed, it is largely underpaid and is not adequate to their level of aspirations. Therefore, increasing number of higher educated people and the consequent unemployment may affect positively life satisfaction inequality in India.

Table 4 also shows that composition effect with respect to female is negative and significant at 10 per cent level for life satisfaction Gini, but for variance it is insignificant. This is because the decline in the share of female respondents in the second period as well as the fall in average life satisfaction of females from 5.92 to 5.20 between the periods 1990–2001 and 2010–2014. The overtime decrease in average life satisfaction of female population caused an increase in life satisfaction inequality. This means that increasing the share of female samples can decrease life satisfaction inequality in India. The composition effect with respect to age of respondents significantly reduces variance of life satisfaction, but insignificant for life satisfaction Gini. This is because the mean age of population increased from 38 to 40 years between 1990–2001 and 2010–2014.

Additionally, negative and significant coefficient effect with respect to social capital index shows that life satisfaction inequality can reduce largely with increase of social capital, a result consistent with cross-sectional RIF regression estimates. The higher level of social capital, either sociability or trust in others, reduces variance of life satisfaction by 0.618 and life satisfaction Gini by 0.019 over time. Thus, improving the levels of trust and social connections among people through confidence building measures in the long-run can reduce life satisfaction inequality more strongly than improving income inequality.

The decomposition analysis also reveals that regional differences contribute a lot to happiness inequality in India. Relative to east Indians, the composition effect with respect to people who are living in west India states significantly increases both indices of life satisfaction inequality. The negative coefficient of living in south and west India on variance of life satisfaction in RIF regression results during second period are regained in decomposition estimates of variance coefficient effect. Living either in south, north or west India, compared to life in east India, significantly reduces the variance as well as Gini coefficient of life satisfaction inequality. As is widely known, the eastern India comprises much dense population and poor people, offers less opportunities for education and employment and face frequent disruptions to normal life, compared to other parts of India.

Thus, the increase in life satisfaction inequality in India during 2010–2014 is mainly due to the coefficient effects, changes in functional relationship between life satisfaction inequality and its influencing factors over time, and the overall coefficient effects mainly come from the impact of non-monetary factors like health, residence and social capital. The negative composition effects, the changes in influencing characteristics over time, mainly come from education. Income inequality influences life satisfaction inequality in the short-run, but not in the long-run. Enhancing social capital can mitigate happiness inequality to a significant extent.

Conclusion

The aim of this paper is to explore the relationship between happiness inequality and income inequality in India over time. As changes in life satisfaction measured by ordinal scales is of narrow range, this paper estimated life satisfaction inequality by the recentered influence regression method. The WVS data of 12 Indian states over 24 years from 1990 to 2014 has been used. Life satisfaction inequality is measured by standard deviation, variance and Gini coefficient, and absolute income, relative income, rank income and social capital index measures have been constructed. In the empirical analysis, life satisfaction inequality has been estimated for two periods, 1990–2001 and 2010–2014. After estimating the RIF regression, following the Blinder–Oaxaca method, the life satisfaction inequality in the two periods is decomposed into composition and coefficient effects. The empirical analysis reveals that happiness level in states of India has not increased commensurately with rising income, but income inequality and life satisfaction inequality have been rising over the years. For Indian people, relative position matters a lot as Indians are status conscious and hence their increasing aspirations with rising income causes less satisfaction with life. Further, widening gaps in well-being, in terms of income, consumption and wealth over time and space, worsens the gaps in subjective well-being also.

The estimated results of this paper show that life satisfaction inequality in India can be reduced not only by increasing individual income, but importantly reducing income inequality and relative income gap among people and across regions of India. Enhancing education, promoting employment, enabling better health, strengthening social relationships and providing better civic amenities can significantly reduce life satisfaction inequality in the short-run. The decomposition results show that over time increase of life satisfaction inequality in India is mainly caused by the coefficient effects, that is, the changing functional relationship between life satisfaction inequality and its influencing factors. Among the coefficient effects, inequality in health and social capital and widening dispersion of regional life satisfaction levels play important role individual evaluation of happiness. The overall negative composition effect mainly comes from the reducing share of population of highly educated and better healthy people of India. Further, the level of social capital reduces life satisfaction inequality. Thus, reducing income inequality and enhancing level of trust and social connections over time and space could reduce life satisfaction inequality and help India achieve a better rank in World Happiness Index.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

References

Alesina

, Di Tella

, & MacCulloch

(2004). Inequality and happiness: Are Europeans and Americans different? Journal of Public Economics, 88(9), 2009–2042.

Ball

, & Chernova

(2007). Absolute income, relative income and happiness. Social Indicators Research, 88(3), 497–529.

Bartolini

, Bilancini

& Pugno

(2013). Did the decline in social connections depress Americans’ happiness? Social Indicators Research, 110(3), 1033–1059.

Bartolini

, & Sarracino

(2014). Happy for how long? How social capital and economic growth relate to happiness over time. Ecological Economics, 108(C), 242–256.

Bartolini

, & Sarracino

(2015). The dark side of Chinese growth: Declining social capital and well-being in times of economic boom. World Development, 74(C), 333–351.

Becchetti

, Massari

, & Naticchioni

(2014). The drivers of happiness inequality: Suggestions for promoting social cohesion. Oxford Economic Papers, 66(2), 419–442.

Becchetti

, Ricca

E. G.

, & Pelloni

(2012). The relationship between social leisure and life satisfaction: Causality and policy implications. Social Indicators Research, 108(3), 453–490.

Becchetti

, Trovato

, & Bedoya

D. L.

(2011). Income, relational goods and happiness. Applied Economics, 43(3), 273–290.

Beegle

, Himelein

, & Ravallion

(2012). Frame-of-reference bias in subjective welfare. Journal of Economic Behaviour and Organisation, 81(2), 556–570.

10.

Berg

, & Veenhoven

(2010). Income inequality and happiness in 119 Nations. In Greve

(Eds.), Social policy and happiness in Europe (pp. 174–194). Edward Elgar.

11.

Bjornskov

(2008). Social capital and happiness in the United States. Applied Research in Quality of Life, 3(1), 43–62.

12.

Blinder

A. S.

(1973). Wage discrimination: Reduced form and structural estimates. Journal of Human Resources, 8(4), 436–455.

13.

Budria

(2013). Are relative-income effects constant across the well-being distribution. Journal of Happiness Studies, 14(4), 1379–1408.

14.

Chancel

, & Piketty

(2019). Indian income inequality, 1922-2015: From British Raj to Billionaire Raj? Review of Income and Wealth, 65(S1), 533–562.

15.

Chang

W. C.

(2009). Social capital and subjective happiness in Taiwan. International Journal of Social Economics, 36(8), 844–868.

16.

Churchill

S. A.

, & Mishra

(2016). Trust, social networks and subjective well-being in China. Social Indicators of Research, 132(1), 1–27.

17.

Clark

A. E.

(2018). Four decades of the economics of happiness: Where next? Review of Income and Wealth, 64(2), 245–269.

18.

Clark

A. E.

, Fleche

, & Senik

(2014) The great happiness moderation. In Clark

A. E.

and Senik

(Eds.), Happiness and economic growth: Lessons from developing countries (pp. 32–139). Oxford University Press.

19.

Clark

A. E.

, Fleche

, & Senik

(2016). Economic growth evens-out happiness: Evidence from six surveys. Review of Income and Wealth, 62(3), 405–419.

20.

Deaton

(2008) Income, health, and well-being around the World: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2), 53–72.

21.

Di Tella

, & MacCulloch

(2006). Some uses of happiness data in economics. Journal of Economic Perspectives, 20(1), 25–46.

22.

Easterlin

R. A.

(1974). Does economic growth improve the human lot? Some empirical evidence. In David

P. A.

and Reder

M. W.

(Eds.), Nations and households in economic growth (pp. 56–81). Academic Press.

23.

Easterlin

R. A.

, & Angelescu

(2010). Happiness and growth the world over: Time series evidence on the happiness–income paradox. In Hinte

& Zimmermann

K. F.

(Eds.), Happiness, growth, and the life cycle (pp. 56–81). Oxford University Press.

24.

Easterlin

R. A.

, McVey

L. A.

, Switek

, Sawangfa

, & Zweig

J. S.

(2010). The happiness-income paradox revisited, Proceedings of the National Academy of Sciences USA, 107(52), 22463–22468.

25.

Easterlin

R. A.

, & Sawangfa

(2010) Happiness and growth: Does the cross section predict time trends? Evidence from developing countries. In Diener

E. D.

., Helliwell

, & Kahneman

(Eds.), International differences in well-being (pp. 162–212). Oxford University Press.

26.

Easterlin

R. A.

(2015). Happiness and economic growth: The evidence. In Glatzer

, Camfield

, Moller

, & Rojas

(Eds.), Global handbook of quality of life: Exploration of well-being of nations and continents (pp. 283–299). Springer.

27.

Easterlin

R. A.

(2017). Paradox lost? Review of Behavioral Economics, 4(4), 311–339.

28.

Fehr

& Schmidt

K. M.

(1999). A theory of fairness, competition and co-operation, Quarterly Journal of Economics, 114(3), 817–868.

29.

Ferrer-i-Carbonell

(2005). Income and well-being: An empirical analysis of the comparison income effect. Journal of Public Economics, 89(5), 997–1019.

30.

Ferrer-i-Carbonell

, & Frijters

(2004). How important is methodology for the estimates of the determinants of happiness? Economic Journal, 114(497), 641–659.

31.

Firpo

, Fortin

N. M.

, & Lemieux

(2018). Decomposing wage distributions using recentered influence function regressions. Econometrics, 6(28), 68–107.

32.

Firpo

, Fortin

N. M.

, & Lemieux

(2009). Unconditional quantile regressions. Econometrica, 77(3), 953–973.

33.

Fortin

, Lemieux

, & Firpo

(2011). Decomposition methods in economics. In Ashenfelter

& Card

(Eds.), Handbook of labor economics, volume 4, part A (pp. 1–102). North-Holland-Elsevier.

34.

Frey

B. S.

, & Stutzer

(2002). What can economists learn from happiness research? Journal of Economic Literature, 40(2), 402–435.

35.

Graafland

, & Lous

(2018). economic freedom, income inequality and life satisfaction in OECD Countries. Journal of Happiness Studies, 19(7), 2071–2093.

36.

Graham

(2017) Happiness for all? Unequal hopes and lives in pursuit of the American Dream. Princeton University Press.

37.

Guimaraes

, & Sheedy

K. D.

(2012). A model of equilibrium institutions (Centre for Economic Performance Discussion Paper No. 1123). London School of Economics.

38.

Gurr

T. R.

(1994) Peoples against states: Ethnopolitical conflict and the changing World System, International Studies Quarterly, 38(3), 347–377.

39.

Hampel

F. R.

(1974). The influence curve and its role in Robust estimation, Journal of the American Statistical Association, 69(346), 383–393.

40.

Helliwell

J. F.

, Huang

, & Wang

(2017). Social foundations of world happiness. In Helliwell

J. F.

, Layard

, & Sachs

(Eds.), World happiness report 2017 (pp. 8–47). Sustainable Development Solutions Network.

41.

Ifcher

, Zarghamee

, & Graham

(2018). Local neighbours as positives, regional neighbours as negatives: Competing channels in the relationship between others’ income, health, and happiness, Journal of Health Economics, 57, 263–276.

42.

Kalmijn

, & Veenhoven

(2005). Measuring inequality of happiness in nations: In Search for proper statistics, Journal of Happiness Studies, 6(4), 357–396.

43.

Lakshmanasamy

(2010). Are you satisfied with your income? The economics of happiness in India, Journal of Quantitative Economics, 8(2), 115–141.

44.

Majumdar

, & Gupta

(2015). Don’t worry, be happy. A survey of the economics of happiness. Economic and Political Weekly, 50(40), 50–62.

45.

Niimi

(2018). What affects happiness inequality? Evidence from Japan. Journal of Happiness Studies, 19(2), 521–543.

46.

Oaxaca

(1973). Male-female wage differentials in urban labor markets, International Economic Review, 14(3), 693–709.

47.

OECD (2012). Reducing income inequality while boosting economic growth: Can it be done? Economic policy reforms 2012: Going for growth. Organisation for Economic Cooperation and Development.

48.

Oshio

, & Kobayashi

(2010). income inequality, perceived happiness, and self-rated health: Evidence from nationwide surveys in Japan. Social Science and Medicine, 70(9), 1358–1366.

49.

Ott

(2005). Level and inequality of happiness in Nations: Does greater happiness of a greater number imply greater inequality in happiness? Journal of Happiness Studies, 6(4), 397–420.

50.

Ovaska

, & Takashima

(2010). Does a rising tide life all the boats? Explaining the national inequality of happiness, Journal of Economic Issues, 44(1), 205–223.

51.

Piekalkiewicz

(2016). Money, social capital and materialism: Evidence from happiness data. In Notebooks of the Department of Political Economy and Statistics 731. Department of Economics, University of Siena.

52.

Portela

, Neira

& Salinas-Jimenez

M. D. M.

(2013). Social capital and subjective well-being in Europe: A new approach on social capital. Social Indicators Research, 114(2), 493–511.

53.

Ram

(2010). Social capital and happiness: Additional cross-country evidence. Journal of Happiness Studies, 11(4), 409–418.

54.

G Rousseau

J.-B.

(2009). Essays on the economics of happiness (Unpublished Ph.D Dissertation). University of Michigan.

55.

Schneider

(2012). Income inequality and its consequences for life satisfaction: What role do social cognitions play? Social Indicators Research, 106(3), 419–438.

56.

Stevenson

, & Wolfers

(2008). Economic growth and happiness: Reassessing the Easterlin paradox. Brookings Papers on Economic Activity, Spring, 1–87.

57.

Stutzer

, & Frey

B. S.

(2010) Recent advances in the economics of individual subjective well-being. Social Research: An International Quarterly, 77(2), 679–714.

58.

Tullock

(1971). The paradox of revolution. Public Choice, 11(1), 89–99.

59.

Van Praag

B. M. S.

(1991). Ordinal and cardinal utility: An integration of the two dimensions of the welfare concept. Journal of Econometrics, 50(1–2), 69–89.

60.

Veenhoven

(2005). Return of inequality in modern society? Test by dispersion of life satisfaction across time and nations. Journal of Happiness Studies, 6(4), 457–487.

61.

Verme

(2011). Life satisfaction and income inequality. Review of Income and Wealth, 57(1), 111–137.

62.

Yang

, Liu

, & Zhang

(2019). Happiness inequality in China. Journal of Happiness Studies, 20(8), 2747–2771.

63.

Zagorski

, Evans

, Kelley

, & Piotrowska

(2014). Does national income inequality affect individuals’ quality of life in Europe? Inequality, happiness, finances, and health. Social Indicators Research, 117(3), 1089–1110.