1 Introduction

The history of ‘wellbeing’ dates back to the eighteenth century; Jeremy Bentham, the philosopher and social reformer, first pointed out that the government should maximize the wellbeing of its citizens (Bentham, 1789; Brown et al., 2012). Diener (1984) suggested that wellbeing is considered the strongest motivator for human action. Therefore, human wellbeing is increasingly acknowledged in many world development initiatives (D’Acci, 2011), such as the United Nations’ Sustainable Development Goals (SDGs), SDG3: ensuring healthy lives and promoting wellbeing for all (United Nations, 2015). Thus, measurement and determinants of wellbeing are of utmost importance in exploring ways to enhance it.

Wellbeing has historically been measured using a macroeconomic indicator, the gross domestic product (GDP) or gross national product (GNP) (Cummins et al., 2003; Fischer, 2009). However, GDP as a measure of wellbeing is highly criticized (Fischer, 2009). Therefore, parallel discussions have highlighted the need to move away from GDP-based measures towards measures that better capture levels of overall human wellbeing (Stiglitz et al., 2009).

Recently, ‘subjective wellbeing’ has been gaining popularity and interest among policymakers and researchers as a measure of wellbeing (Diener et al., 2013; Jayasinghe et al., 2021a). Subjective wellbeing captures an individual’s perception of the current state of life using a response to a common question, “How do you feel about your life as a whole?” with a 10-point scale ranging from 1 (least satisfied) to 10 (extremely satisfied). This approach is also criticized on the grounds that it focuses on feelings (Gasper, 2005), which might lead to biased responses due to the wording or placement of the survey questions (Helliwell and Barrington-Leigh, 2010). Alternative to the subjective approach, there has been a notable development in the literature on constructing indices to capture a wide range of human development aspects in measuring wellbeing (Castells-Quintana et al., 2019).

While the research on wellbeing is growing, until recently it has primarily concentrated on Western developed nations (Itaba, 2022). In recent years, there has been substantial growth in wellbeing research in the context of developing countries. However, a review of the literature suggests that the analysis of wellbeing still remains an under-researched area in the context of developing countries (Itaba, 2022) and is an important area of research to achieve SDGs (Costanza et al., 2016).

To this end, this study provides new evidence on wellbeing in Bangladesh using a multidimensional wellbeing index (MDWI). The specific objectives are to (1) investigate the level and intensity of wellbeing in Bangladesh by estimating an MDWI, and; (2) identify the socioeconomic and demographic determinants of wellbeing in Bangladesh. To provide a further nuanced understanding, this study also includes insights into how the level of wellbeing has changed in Bangladesh during the period 2002–2018. The empirical analysis is mainly based on Wave-7 (2018) of the World Values Survey (WVS). To identify the intertemporal changes in wellbeing, we also use Wave-4 (2002) data. The WVS also provides data on the popular subjective wellbeing measure, self-rated life satisfaction, enabling us to test the robustness of the estimation results based on MDWI and those that are based on self-rated life satisfaction. To the best of our knowledge, no previous studies on wellbeing have undertaken such a comparative analysis.

The selection of Bangladesh as a case study is motivated by the fact that Bangladesh is an emerging economy in South Asia, and the country recently achieved the status of a middle-income country, moving from a low-income country status (World Bank, 2021). This is a result of the notable progress in reducing poverty and boosting shared prosperity over the past five decades (World Bank, 2021). Nevertheless, little is known about the impact of improved standard of living on wellbeing. Osberg and Sharpe (2002) suggested that accurately measuring and tracking wellbeing is essential as the availability of accurate information on wellbeing informs the design of policy, regulatory, and financial strategies to promote wellbeing.

Several wellbeing related research studies have focused on Bangladesh as detailed in Sect. 2.3. To the best of our knowledge, there is only one study, Camfield and Ruta (2007), that used an index, the global person generated index (GPGI), to examine the quality of life in Bangladesh. However, the GPGI was constructed to undertake a qualitative assessment of methodological validity and contained only a sample of twenty-two respondents from Bangladesh in a multi-country study (Camfield & Ruta, 2007).

Therefore, our study is substantially different and extends the wellbeing literature on Bangladesh and in general on developing countries in four ways: (1) unlike previous studies that used self-rated life satisfaction as the measurement of wellbeing, this study constructs a wellbeing index, the MDWI, which incorporates seven domains and twenty-five indicators that broadly affect human wellbeing; (2) this study is the first to investigate the prevalence, extent, and determinants of multidimensional wellbeing at the national level by constructing the MDWI in Bangladesh using a nationally representative sample of 1200 respondents; (3) this study provides an understanding of how wellbeing in Bangladesh changed over time; and (4) the use of principal component analysis-based (PCA) weights to construct the index contributes to the advancement of methodological approaches to wellbeing analysis. The estimation results indicate that, on average, Bangladeshi people enjoy a moderate level of wellbeing (the value of MDWI is 0.565), where the health domain is the most significant contributor.

The remainder of this paper is organized as follows. Section 2 presents a review of the literature on wellbeing theories and measurement. Section 3 discusses the WVS data and methodology used for the empirical analysis of this study. Section 4 presents estimated results and discussion, followed by conclusions and policy implications in Sect. 5.

2 Review of Literature

In this section, we present a short review of the literature on wellbeing theories, wellbeing indices, and wellbeing research in Bangladesh.

2.1 Wellbeing Theories

Commonly used measures of wellbeing, such as self-rated life satisfaction or scales of happiness (Diener et al., 2013; Veenhoven, 2011) to capture satisfaction with life are subjective and subsequently suffer from reliability issues (International Wellbeing Group, 2013). The estimation of wellbeing is a rather complex task due to its multidimensional and multifaceted nature (Murias et al., 2006). Over time, multi-item scales such as life domain scales have evolved (Diener, 1984; International Wellbeing Group, 2013) to measure wellbeing using a more objective approach.

The life domain scale approach adopts a domain-level representation of overall life satisfaction. The individual items refer to specific life domains and the scores are averaged to produce a measure of wellbeing (Cummins & Weinberg, 2013). The domain scale approach is further divided into two main theories, (1) the ‘top-down’ and (2) the ‘bottom-up’ theories of wellbeing (Loewe et al., 2014; Stones & Kozma, 1985). The top-down theory implies that overall life satisfaction determines the satisfaction of the specific life domains (Stones & Kozma, 1985). In contrast, the bottom-up theory implies that overall life satisfaction is an aggregate of domains’ satisfaction (Cummins, 2003). When considering the domains, it is worthwhile to note that the list of domains that affect wellbeing is inconclusive. In particular, evidence suggests that people from different cultures and in various stages of their lives weigh life domains differently while judging wellbeing (Loewe et al., 2014). As there is no consensus on the accuracy and usage of the two theories, this study followed the widely used bottom-up approach. Figure 1 presents the conceptual framework of the same with the most commonly used domains in the literature.

Fig. 1
figure 1

Source: Authors’ compilation based on Headey et al. (1991) and Cummins (1996)

The conceptual framework of the bottom-up approach to wellbeing.

Full size image

As can be seen in Fig. 1, the existing literature on multidimensional wellbeing used material wellbeing, health, finance, culture, safety, governance, and environment (Cummins et al., 2003; OECD, 2011; Ura et al., 2012) as important domains of wellbeing. There is a considerable body of literature that supports the role of these domains in enhancing human wellbeing. For example, Higgs (2007) noted that satisfaction with basic health, both mental and physical, is an essential element of an individual’s wellbeing. Similarly, Brown et al. (2012) and Sacks et al. (2010) noted that employment and income lead to financial satisfaction, which is also essential for overall wellbeing. Ura et al. (2012) suggested that the distinctive culture of a country represents the sovereignty of the country, the identity, and the sense of belonging of the people, leading to human wellbeing. Similar to culture, Kortt et al. (2015) documented that people’s involvement in religious activities, particularly in some cultures, is substantially linked with wellbeing. Botha (2016) suggested that good governance is essential for a stable democracy where individuals have the freedom of speech and choice to vote, which is also expected to significantly contribute to overall human wellbeing. Botha (2016) further noted that the wellbeing of the citizens is enhanced when they feel safe from personal violence. Koç and Turan (2021) highlighted that, as technological advances represent emerging opportunities and challenges in various aspects of individuals’ daily lives, science and technology is increasingly becoming an integral part of human wellbeing. Ura et al. (2012) also suggested that people’s ability to live in an environment free of all forms of ecological degradation plays an important role in enhancing human wellbeing.

2.2 Major Wellbeing Indices

Researchers have developed a number of composite wellbeing indices in various countries predominantly based on the bottom-up approach (Senasu et al., 2019). For example, Johnston (1988) proposed a comprehensive quality-of-life (QoL) index for the United States. Cummins et al. (2003) developed the Australian unity wellbeing index for Australia. Subsequently, the personal wellbeing index (PWI), the national wellbeing index (NWI), and the international wellbeing index (IWI) (the IWI is a combination of PWI and NWI) were also developed (International Wellbeing Group, 2013) and tested in Algeria (Tiliouine et al., 2006) and Austria (Renn et al., 2009). The Organisation for Economic Cooperation and Development (OECD) launched the OECD better life index for thirty-eight OECD member countries (OECD, 2011). Ura et al. (2012) developed the gross national happiness (GNH) index for Bhutan. Collomb et al. (2012) developed a multidimensional index for Namibia, while Haq and Zia (2013) developed a multidimensional index for Pakistan. Similar to Bhutan’s GNH, Pratt (2016) constructed the GNH index for the Solomon Islands and Tonga. Senasu et al. (2019) constructed the Thai happiness index. Additionally, Botha (2016) developed the good African society index (GASI) for forty-five African countries.

Significant variations were observed in the selection of domains, the number of indicators, and the methodology of constructing the indices. The most commonly used domains include material wellbeing, health, finance, safety, culture, religion, environment, and governance (Cummins, 1996; Cummins et al., 2003). The list of domains is inconclusive, and there are numerous others (for a detailed description of domains, see, Cummins, 1996).

Decancq and Lugo (2013) stated that a crucial step in constructing a multidimensional index is the selection of the relative weights for the different dimensions. Most wellbeing indices are unweighted in nature. Bhutan’s GNH and the Thai Happiness Index used normative or arbitrary weights and were constructed based on the Alkire and Foster (2011) multidimensional poverty index approach (Senasu et al., 2019; Ura et al., 2012). Several researchers criticized the use of normative weights (equal or arbitrary) when constructing multidimensional indices (Jayasinghe et al., 2021b). Nussbaumer et al. (2012) also emphasized that arbitrary weights should only be used for demonstration purposes.

2.3 Wellbeing Research in Bangladesh

An in-depth review of the literature revealed that only a limited number of research has focused on the analysis of wellbeing in Bangladesh.

Camfield and Ruta (2007) employed a global person generated index (GPGI) to assess the individual quality of life in Bangladesh, Thailand, and Ethiopia. The GPGI was constructed as a measure of the quality of life using a survey with five questions to assess the qualitative methodological validity of the construct and to assess the correlation between the GPGI and health and material wellbeing. The study found a strong correlation between GPGI and health and a moderate correlation between material wellbeing and GPGI. However, the study was limited to only twenty-two respondents from Bangladesh. Using quantitative and qualitative data, Camfield et al. (2009) showed the relationship between happiness and income, and personal and social relational values. The study found that irrespective of the level of income, people in Bangladesh consider personal (relation with the spouse) and social (relation with the community) relational values to be important in achieving happiness. Camfield et al. (2010) using the wellbeing in developing (WeD) database found that subjective wellbeing is significantly influenced by the lack of basic needs such as health and autonomy, otherwise referred to as objective need deprivation.

Asadullah and Chaudhury (2012) analyzed the correlation between subjective wellbeing and absolute income (own income) and relative income (own income compared to the neighbors’ income) using multi-purpose household survey data of 2400 households from Bangladesh. The results revealed there is a strong positive association between absolute income and life satisfaction and a weak positive relationship between relative income and life satisfaction in Bangladesh. Joarder et al. (2017) explored the influences of remittances on happiness using matched samples of Bangladeshi migrants living in the United Kingdom (UK) and Malaysia. Results revealed that remittances significantly stimulate migrants’ happiness. Bhuiyan and Ivlevs (2019) examined the role of microcredit borrowing in promoting livelihoods and reducing poverty and its impact on people’s subjective wellbeing in Bangladesh. Despite having no direct effects, the study found that microcredit borrowing has an indirect adverse impact on overall life satisfaction through increased worry.

Devine et al. (2019) examined the influence of religion on people’s wellbeing and life satisfaction in Bangladesh using a sample of 1500 respondents in WeD data between 2002 and 2007. Regression analysis identified differences in reported happiness between the country’s two largest religious populations, Muslims and Hindus. Hossain et al. (2019) revealed the relationship between life satisfaction and empowerment in rural Bangladesh using data generated from the Bangladesh integrated households survey (BIHS) 2012 and a sample of 3860 households. The empowerment index captures agricultural and community life, leisure, and decision-making autonomy, and found a positive relationship between empowerment and life satisfaction. Tauseef (2022) empirically analyzed the importance of income, relative income, monetary, and nonmonetary poverty on individual wellbeing in rural Bangladesh utilizing a sample of 6503 households from the BIHS. A linear panel model with individual random effects revealed a significant positive relationship between wellbeing and income and a negative relationship between all forms of poverty and happiness.

Overall, the majority of wellbeing research in Bangladesh is based on self-rated life satisfaction. The only existing composite index of wellbeing, the GPGI is limited by only capturing health and material wellbeing aspects. The lack of adequate scientific research on wellbeing using a measure of multidimensional construct in Bangladesh necessitates comprehensive research to identify the incidence, intensity, and drivers of wellbeing in Bangladesh to formulate appropriate policy guidelines in achieving SDGs. In doing so, this study will address the gap in the current literature and contribute to the existing literature using a multidimensional wellbeing index (MDWI). This study also aims to contribute to the methodological advancement of wellbeing measurements by incorporating PCA-based weights, acknowledging the importance of this data-driven approach which is free from explicit value judgment (Decancq & Lugo, 2013).

3 Data and Methodology

This section briefly explains the sources of data, the selection of domains and indicators, variables of interest, sufficiency cut-offs, and the econometric methods used in this study.

3.1 Data: the World Values Survey (WVS) for Bangladesh

This study utilizes the latest wave (Wave-7) of 2018 WVS (Haerpfer et al., 2022) data for Bangladesh, which covers a wide variety of information including self-reported life satisfaction, political, economic, cultural, social, and religious beliefs (Ergin & Mandiracioglu, 2015). The survey covers a nationally representative sample of 1200 respondents across all eight administrative divisions of Bangladesh, including urban and rural regions. To compare the changes in MDWI over time, this study also considers Wave-4 (2002) data. The estimates in this study are derived using sampling weights given in the dataset to obtain nationally representative results. Table 1 provides the summary statistics of the sociodemographic and economic variables. Overall, the distribution of the respondents based on gender, level of education, region, and administrative divisions is consistent with the national-level distribution of the population in Bangladesh (BBS, 2023).

Table 1 Summary statistics of socioeconomic and demographic characteristics, WVS-2018 (Wave-7).
Full size table

3.2 Selection of domains and related indicators

In this study, we use twenty-five indicators belonging to seven domains: (1) health, (2) finance, (3) culture, (4) safety, (5) governance, (6) religion, and (7) science and technology. The selection of domains and indicators was informed by the existing literature discussed in Sect. 2.1 and presented in Table 2.

Table 2 Domains and indicators of wellbeing based on WVS-2018 (Wave-7) dataset, Bangladesh.
Full size table

3.3 Methodology

This section presents the description of the methodologies used for the empirical analysis of this study.

3.3.1 Steps in Constructing the MDWI

While developing the MDWI for Bangladesh, this study adopts the methodological approach used by Alkire and Foster (2011), Nussbaumer et al. (2013), and Jayasinghe et al. (2021b).

Step 1—identifying the domains and indicators: let us consider \(d\) indicators for a set of \(n \times d\) matrices where \(y\in {{\varvec{M}}}^{n,d}\) is the matrix of endowments of \(n\) respondents in \(d\) different indicators. For each respondent \(i\)(= 1, 2,…, n) and indicator \(j\)(= 1, 2, …, d), the entry \({y}_{ij}\) of \(y\) is respondent \(i\)’s endowment in indicator \(j\). The \(1 \times d\) row vector \({Y}_{i}=[{y}_{i1, }{y}_{i2, }\dots ,{y}_{id }]\) is the sufficiency vector of respondent \(i\) across the \(d\) indicators such that \(Y=[{Y}_{i}]\) and the \(n \times 1\) column vector \({Y}_{j}={[{y}_{1j, }{y}_{2j, }\dots {y}_{nj }]}{\prime}\) is the distribution of sufficiency in the variable \(j\) across \(n\) respondents such that\(Y=[{Y}_{j}]\).

Step 2—assigning weights: weights are assigned for each indicator \({w}_{j}\) (with \(0<{w}_{j}<1\)) and the sum of the weights adds to 1. That is,

$$\sum_{j=1}^{d}{w}_{j}=1 \;or\; w\tau =1$$
(1)

where \(w=({w}_{1},{w}_{2},\dots , {w}_{d})\) is a row vector of weights, and τ is a unit column vector.

Step 3—determining sufficiency matrix: based on previous literature, a threshold is developed for indicator sufficiency as

$${g}_{ij}=\left\{\begin{array}{c}{w}_{j} \quad if {y}_{ij}\ge {Z}_{j}\left(respondent \;i\; is\; sufficient\; in\; indicator\; j\right) \\ 0 \quad if {y}_{ij}<{Z}_{j}(respondent\; i\; is\; not\; sufficient\; in\; indicator\; j)\end{array}\right.$$
(2)

where \({Z}_{j}\) is the sufficiency (cut-off) threshold in indicator \(j\) and defines the threshold value \({g}_{ij}\) for respondent \(i\). Therefore, the \(n\times d\) matrix \(G={[g}_{ij}]\) is known as the sufficiency matrix. Thus, the sum of each row across the columns would provide the total sufficiency of each respondent given by \(n\times 1\) column vector C, which is \(G\tau\), where τ is a unit column vector such that

$$C=\left[{c}_{i}\right]=G\tau$$
(3)

where \({c}_{i}=\sum_{j=1}^{d}{g}_{ij}\) represents the sum of total sufficiency endowed by an individual \(i (i=1, 2, \dots , n)\).

Step 4—calculating the proportion of people meeting the indicator sufficiency threshold: an individual is classified as meeting the indicator sufficiency threshold if the total indicator sufficiency of individual \(i\), \({c}_{i}\) is greater than or equal to a defined cut-off value \(k\). That is, the ith element of the indicator sufficiency column vector \(M=[{m}_{i}]\) and corresponding head-count column vector \(Q=[{q}_{i}]\) are defined as

$${m}_{i}=\left\{\begin{array}{c}{c}_{i} \,if\, {c}_{i}\ge k\\ 0\, if\, {c}_{i}<k\end{array}\right.$$
(4)
$${q}_{i}=\left\{\begin{array}{c}1\, if\, {c}_{i}\ge k\\ 0 \,if \,{c}_{i}<k\end{array}\right.$$
(5)

Thus, the number of indicator-sufficient individuals (where \({c}_{i}\ge k\)) in the population size \(n\) is given by

$$q={\tau }{\prime}Q$$
(6)

where \(\tau\) is a \(n\times 1\) unit column vector. Therefore, the percent of indicator-sufficient individuals (H) in a population size \(n\) (also known as head-count-ratio) is given by

$$H=\frac{{\tau }{\prime}Q}{n}=\frac{q}{n}$$
(7)

The head-count ratio (H) also represents the incidence of multidimensional indicator sufficiency.

Step 5—calculating the intensity of wellbeing: the intensity of wellbeing is calculated as the average of total indicator sufficiency in the population, which is the sum of all elements in the multidimensional indicator-specific sufficiency vector \(M\) divided by the number of indicator-specific sufficient individuals \(q\) as follows

$$A=\frac{{\tau }{\prime}M}{{\tau }{\prime}Q}=\frac{\sum_{i=1}^{n}{m}_{i}}{q}$$
(8)

Step 6—calculating the multidimensional wellbeing index (MDWI): the MDWI is calculated as the product of H and A. That is,

$$MDWI=H\times A$$
(9)

The MDWI given by Eq. (9) captures the information on both the incidence and the intensity of wellbeing. The value of the MDWI ranges between 0 and 1; the higher the value, the greater the level of wellbeing.

3.3.2 Calculation of the MDWI for Bangladesh

The seven domains considered for this study comprise twenty-five indicators (I1 to I25), as detailed in Column (1) of Table 3. These indicators incorporate individual wellbeing status and capture the sufficiency of the status. Once the indicators are finalized, they are converted to percentages using the formula as \(\frac{Self \, reported \, scale \, score}{Max \, value \, score}\times 100\) to assign the sufficiency (cut-off) threshold. For example, self-reported financial satisfaction (how satisfied are you with the financial situation of your household on a scale of 1 = completely dissatisfied to 10 = completely satisfied), if someone’s response was seven, it is converted to 70% (\(\frac{7}{10}\times 10=70 \%)\). Moreover, following Ura et al. (2012), additivity is also employed, for example, the frequency of various crimes in the safety domain comprised five similar variables, each with a 4-point Likert scale. These responses are added and converted to percentages using the above formula and are reported in Column (2) of Table 3.

Table 3 Domains, respective indicators with sufficiency cut-offs and weights, WVS-2018 (Wave-7).
Full size table

Based on previous literature, this study selected the sufficiency threshold (cut-off points) as reported in Table 3. For example, Cummins et al. (2003), Renn et al. (2009), and Tiliouine et al. (2006) showed that sufficiency is usually achieved within a 60–70 percent score maximum (%SM) for developing countries.

Following this approach, we defined the upper limit of 70%SM as the cut-off for the sufficiency threshold for most of the indicators. As shown in Table 2, the measurement scale of different indicators varies. For indicators that are measured on a 10-point Likert scale where 1-least satisfied to 10-very satisfied, the sufficiency threshold for 70%SM is 7/10. Hence, individuals with at least 7/10 satisfaction with a particular indicator are considered as having met the sufficiency threshold. Similarly, under the 70%SM cut-off, for indicators measured at a 5-point Likert scale where 1-least satisfied to 5-highly satisfied, an individual with at least 4/5 satisfaction is considered as having met the sufficiency threshold while at least 3/4 was the cut-off for questions with 4-point Likert scale responses.

Based on Ura et al. (2012), a slightly different approach was used for indicators twenty-one (I21) (attendance at religious services) and twenty-two (I22) (how often do you pray). This is because Ura et al. (2012) suggested that in some religions, attending religious activities and praying is a more prominent part of people’s lives than in other religions. Hence, engaging in these activities at least once a week is considered the most reasonable sufficiency cut-off. Based on the seven-point Likert scale for indicator twenty-one and eight-point Likert scale for indicator twenty-tow, six and seven, respectively, are the cut-off points. This means that individuals who attend religious activities and pray once a week or more than once a week are considered to have met the sufficiency threshold. We also observed that some indicators have been measured as dichotomous responses (yes or no). In this case, for indicators five (I5), six (I6), and seven (I7), those who responded ‘yes’ are considered to have met the sufficiency threshold, while for indicator sixteen (I16), those who responded ‘no’ are considered to have met the sufficiency threshold. Column (2) of Table 3 summarizes the cut-off for each indicator.

In this study, different weights, such as equal weights and weights generated from Principal Component Analysis (PCA) are utilized for comparison purposes. The equal weights are further divided into two categories; (1) equal domain weights but unequal indicator weights obtained using PCA, and (2) equal indicator weights. Columns (2)–(3) of Table 3 present the sufficiency threshold, and the share of respondents above the sufficiency threshold while Column (4) presents the weights assigned under an equal domain (1/7 = 0.141) but distributed unequally within the domain calculated based on PCA. Column (5) presents equal weights for each indicator (1/25 = 0.04), and Column (6) presents the unequal indicator weights calculated based on PCA. Based on the literature (Alkire et al., 2017; Nussbaumer et al., 2012), this study also defines that individuals are considered to enjoy a reasonable level of wellbeing if the MDWI > k (= 0.33). The estimated values of H, A, and the MDWI are presented and discussed in detail in Sect. 4.1.

To provide an understanding of how wellbeing in Bangladesh has changed during the last two decades, we also estimated MDWI using WVS-2002 (Wave-4) data. However, it is worthwhile noting that there are only eleven relevant indicators associated with five domains available in Wave-4 data while with Wave-7, there are seven domains and twenty-five indicators available. As a result, for the comparative analysis, MDWIs are calculated using the eleven indicators that are common to both Wave-4 and Wave-7.Footnote 1 A decomposition analysis of the MDWI is also conducted by socioeconomic and demographic characteristics. Following Pacifico and Poege (2017) and using Stata 17 software, the results in this paper have been generated by user-written commands for multidimensional poverty index estimation.

3.3.3 Determinants of MDWI in Bangladesh

Previous studies identified several socioeconomic and demographic factors that determine wellbeing. These include gender (Dolan et al., 2008); age (Donovan & Halpern, 2002); level of education (Kristoffersen, 2018); income (Sacks et al., 2010); marital status (Pagán, 2015); and political party affiliation (Appleton & Song, 2008). This study also considers the above determinants and estimates their effects on the MDWI. Since income data is not available in the WVS data, this study included the social class group as a proxy for income.

To assess the impacts of the determinants on MDWI, the following model is estimated.

$$\begin{aligned} MDWI =\; & \beta_{0} + \beta_{1} \left( {gender} \right) + \mathop \sum \limits_{j = 1}^{3} \beta_{2j } (age) + \mathop \sum \limits_{j = 1}^{4} \beta_{3j } (education\;level) \\ & \; + \mathop \sum \limits_{j = 1}^{2} \beta_{4j } (marital\;status) + \mathop \sum \limits_{j = 1}^{4} \beta_{5j } (social\;class) + \beta_{6} \left( {political\;affiliation} \right) \\ & \; + \beta_{7} \left( {region} \right) + \mathop \sum \limits_{j = 1}^{7} \beta_{8j } (division) + \varepsilon \\ \end{aligned}$$
(10)

where gender is a dummy variable (1 = male and 0 = female); age = age group of the respondent \(j\) (j = 1 for 18–34 years and 0 for others; j = 2 for 35–49 years and 0 for others; and j = 3 for 50–64 years and 0 for others); education level = level of education of the respondent \(j\) (where j = 1 for illiterate and 0 for others; j = 2 for primary and 0 for others; j = 3 for secondary and 0 for others; and j = 4 for higher secondary and 0 for others); marital status = marital status of the respondent \(j\) (where j = 1 for single/divorced/separated and 0 for others; and j = 2 for widowed and 0 for others); social class = belonging to the social class group \(j\) (where j = 1 for lower and 0 for others; j = 2 for working and 0 for others; j = 3 for lower-middle and 0 for others; and j = 4 for upper-middle and 0 for others); political affiliation is a dummy variable to indicate whether the respondent is affiliated with a political party (1 = affiliated and 0 = not affiliated); region is a dummy variable to present the locality of the respondents’ households (1 = urban and 0 = rural); and division = household’s location to the administrative divisions of the respondent \(j\) (where j = 1 for Dhaka and 0 for others; j = 2 for Chittagong and 0 for others; j = 3 for Rajshahi and 0 for others; j = 4 for Khulna and 0 for others; j = 5 for Sylhet and 0 for others; j = 6 for Barisal and 0 for others; and j = 7 for Rangpur and 0 for others). The justification for employing the quantile regression model introduced by Koenker and Bassett (1978) at the 25th, 50th, and 75th quantiles to estimate Eq. (10) is threefold. First, it enables a greater understanding of data by changing a specified part of the distribution of MDWI, allowing for different percentiles. Second, it is considered more robust than the ordinary least squares (OLS) models. Third, quantile regression models are suitable for heteroscedastic data (Cameron & Trivedi, 2010; Jayasinghe et al., 2021b). However, this study also presents the OLS estimates to validate robustness.

To ensure that the MDWI constructed in this study broadly aligns with the widely-used wellbeing measure, the self-rated life satisfaction, and to enable the comparison of our results with those of the existing studies, we also conduct a robustness analysis of the determinants of wellbeing using the data on self-rated life satisfaction. In doing so, we estimate an ordered logit model taking the self-rated life satisfaction, which ranges from 1–10 (where 1 = extremely dissatisfied and 10 = extremely satisfied) as the dependent variable with the same set of independent variables in Eq. (10). The estimation results of Eq. (10) are presented in Sect. 4.3. Figure 2 below shows the distribution of self-rated life satisfaction and the MDWI constructed in this study.

Fig. 2
figure 2

Histograms of self-rated life satisfaction against the calculated MDWI

Full size image

Using the multidimensional international wellbeing index (IWI), Cummins et al. (2003) showed that for developed nations (Australia and Austria), average life satisfaction ranges from 70 to 80 percent score maximum (%SM) while it ranges from 60 to 70 for developing nations such as Algeria (Tiliouine et al., 2006). As can be seen in Fig. 2, the distribution of self-rated life satisfaction in Bangladesh is skewed to the left with a high average of 70–90% SM, showing a substantial deviation from that of developing countries. Such distribution, however, is in line with Camfield and Ruta (2007) and Camfield et al. (2009), which found that self-rated life satisfaction is much higher in Bangladesh than in the United Kingdom (UK) and other South Asian Countries. However, the MDWI we constructed in this study is normally distributed with an average of 55–70%SM and is broadly in line with Cummins et al. (2003) for developing countries.

4 Results and Discussion

This section first presents the MDWI and the decomposition analysis, followed by the estimation results of Eq. (10).

4.1 Multidimensional Wellbeing Status in Bangladesh

Columns (2)–(4) of Table 4 present the MDWI based on Wave-7 under three different weighting schemes. As can be seen, Part A presents the H, A, and the MDWI at the national level, using the cut-off \((k)\) as 0.33. Part B presents the proportional contribution of each indicator to the overall MDWI.

Table 4 Multidimensional wellbeing (MDWI), national level, WVS-2018 (Wave-7).
Full size table

As explained in Sect. 3.3, the estimated MDWI in Columns (2)-(4) reveals a moderate level of wellbeing MDWI = 0.668 (equal domain weights but unequal for indicators within domain derived using PCA) and MDWI = 0.626 (equal indicator weights) compared to MDWI = 0.565 (unequal weight for each indicator and derived using PCA).Footnote 2 As can be seen from the Column (4) estimates, the estimated value of H, 0.964, implies that on average about 96% of the respondents enjoy better wellbeing outcomes in Bangladesh. The estimated value of A, 0.586, indicates, on average, Bangladeshi people enjoy about 59% sufficiency among all the indicators. The health domain has the highest contribution (26.6%) to overall MDWI, followed by the financial satisfaction (18.4%), culture (15.8%), safety (15.2%), governance (12.2%), religion (6.0%), and science and technology (5.8%). This observation, however, appears to be sensitive to the weigting scheme used for estimation of the index.

Tables 5 and 6 present decomposition analyses of incidence and intensity of wellbeing by socioeconomic and demographic characteristics such as gender, age group, social class group, region (rural–urban), and administrative divisions of the households based on Wave-7 data. In accordance with the national level results, the disaggregated analyses also revealed that all groups experience a moderate level of wellbeing, on average, and the health domain is the largest contributor to the MDWI, followed by finance, culture, safety, governance, religion, and science and technology. Irrespective of gender, age group, and region, in general, Bangladeshi people enjoy a somewhat similar level of wellbeing except for the age group 65 and above (Table 5 and 6). This might be due to old-age health problems, financial insecurity of retired people, and seperation from their children (Sarker, 2021).

Table 5 Decomposition of multidimensional wellbeing by gender, age groups, and social class, WVS-2018 (Wave-7).
Full size table
Table 6 Decomposition of multidimensional wellbeing by region and administrative divisions, WVS-2018 (Wave-7).
Full size table

Notable differences in wellbeing are observed among the social class groups. It is evident that the upper-middle social class group enjoys the highest level of wellbeing while the lower class group reports the lowest level of wellbeing (Table 5). Causes of this finding might include the inability of the lower class group to meet the day-to-day basic needs of food, housing, education, and medicine due to low income whereas higher income leads to a higher level of satisfaction (Brown et al., 2012). As can be seen in Columns (4)–(11) of Table 6, among the administrative divisions, results indicate that respondents from the Khulna and Rajshahi divisions enjoy marginally higher levels of wellbeing compared to other divisions.

4.2 Intertemporal Changes in Wellbeing in Bangladesh

Columns (2) and (3) of Table 7 present a comparison between MDWIs of WVS 2002 (Wave-4) and 2018 (Wave-7) based on eleven commonly available indicators for both Waves. As can be seen, over time, the head-count ratio has increased from 0.843 in 2002 (Wave-4) to 0.910 in 2018 (Wave-7), indicating an increase in the proportion of individuals with better wellbing. However, the intensity remained more or less similar between the two time periods. The MDWI at the national level has seen a marginal increase in 2018 (MDWI = 0.539) compared to 2002 (MDWI = 0.468).

Table 7 Intertemporal comparison of wellbeing in Bangladesh WVS-2002 (Wave-4) and 2018 (Wave-7).
Full size table

The results suggest some intertemporal changes in the percentage contribution of each indicator to the MDWI. For example, in 2002, the highest contributing domain was culture, followed by finance, health, and governance. Religion did not appear to contribute to wellbeing in 2002. However, the 2018 results which are analogous to 2002 [Column (3) of Table 7] reveal that the finance domain is the most important contributor to the MDWI, followed by culture, religion, health, and governance. In contrast, the estimation results with all twenty-five indicators and seven domains [Column (4) of Table 4] show that health is the largest MDWI contributor. Overall, the results reveal that the life domains which contribute to people’s wellbeing have changed over time.

4.3 Determinants of MDWI in Bangladesh

The estimates of OLS, quantile and ordered logit regressions of Eq. (10) of Sect. 3.3.3 are presented in Table 8. Consistency in the sign of the regression coefficients is observed in all model estimation results. The estimation results presented in Columns (3)–(5) of Table 8 reveal, in Bangladesh, males enjoy higher wellbeing at the 50th and 75th quantile compared to females. This study also finds that older people are unhappier than the young at the 25th quantile, which implies that with an increase in age wellbeing declines among those who are in the lowest wellbeing quantile. Compared to the illiterate education group all other groups enjoy higher wellbeing in all quantiles. This means, an increase in education level leads to increased wellbeing. All the social class groups, compared to the lower class, enjoy higher wellbeing at all quantiles, which implies an improvement in the social class group leads to improved wellbeing. Political party affiliation is negatively associated with wellbeing at 50th and 75th quantiles, which implies that political party affiliation reduces wellbeing. Mixed results can be seen for administrative divisions. The Rajshahi and Khulna divisions are positively associated with wellbeing while the Barisal and Rangpur divisions are negatively associated with wellbeing. These estimates are somewhat consistent with the available previous studies in Bangladesh (see, for example, Asadullah & Chaudhury, 2012; Joarder et al., 2017; Bhuiyan & Ivlevs, 2019; Tauseef, 2022).

Table 8 Determinants of multidimensional wellbeing in Bangladesh, WVS-2018 (Wave-7).
Full size table

The estimated results presented in Column (6) of Table 8 for the ordered logit model incorporating self-rated life satisfaction as the dependent variable reveal consistent results with those of OLS and quantile regressions models, however, with a much smaller number of statistically significant variables.

5 Conclusions and Policy Implications

Considering the bottom-up theory of wellbeing and using WVS data, this study, by constructing a multidimensional wellbeing index (MDWI), examined the incidence, intensity, and determinants of wellbeing in Bangladesh. While constructing the MDWI, this study considered seven important domains namely (1) health, (2) finance, (3) culture, (4) safety, (5) governance, (6) religion, and (7) science and technology, and twenty-five indicators associated with the seven domains. A quantile regression analysis is also used to examine the effects of socioeconomic and demographic determinants of multidimensional wellbeing in Bangladesh.

The results on the multidimensional wellbeing reveal that over time, the wellbeing of Bangladeshi people has increased (MDWI = 0.468 in 2002, and MDWI = 0.539 in 2018), and currently, people are enjoying a moderate level of wellbeing (MDWI = 0.565). More importantly, this study found health domain is the largest contributor to wellbeing followed by finance, culture, safety, governance, religion, and science and technology. The disaggregated analyses reveal notable differences in wellbeing among the social class groups and administrative divisions. The Quantile regression analysis reveals that gender, education, social class groups, and some of the administrative divisions appeared to be significantly and positively associated with wellbeing in Bangladesh.

Ensuring wellbeing for all is paramount to achieving the sustainable development goal 3 (SDG 3) (United Nations, 2015). Since there is a sense of urgency for all countries and governments across the world to achieve the SDGs, the findings from our study provides timely evidence and policy direction towards this SDG agenda. To this end, a number of broad policy implications can be drawn from the results of the study. First, the constructed MDWI shows that the health domain and its related indicators constitutes more than a quarter (26.6%) of the wellbeing index. This implies that when people feel satisfied with their health, their wellbeing enhances. The World Bank (2023) highlighted that even though Bangladesh has made significant improvements in healthcare facilities and reduced the rural–urban disparities, it needs to do more for the poor to avoid expensive private healthcare. Since our results suggest that health is the most important contributor to the wellbeing of Bangladeshi people, this study calls for increased attention by policymakers toward providing cheaper and readily available public healthcare facilities for underprivileged people. There is an urgent need for the government to invest in quality and accessible healthcare, in particular, by employing qualified doctors, nurses, technicians, and other staff members, providing new technological equipment to hospitals, and undertaking appropriate regulatory measures to minimize the cost of private healthcare facilities.

Second, the MDWI results show that the finance domain is the second largest (18.4%) contributor to wellbeing. Also the disaggregated MDWI reveals that the upper and upper-middle class groups enjoy greater wellbeing than all other social class groups. The quantile regression results also reveal that, compared to the lower class group, all other groups enjoy a higher level of wellbeing. Klement and Terlau (2022) argued that people’s financial satisfaction and social class group can be improved by generating more income earning opportunities and providing quality education, which are also essential for eradicating poverty and enhancing wellbeing. Thus, we highlight the need for the Bangladesh government to attract more national and foreign investments and public and private partnerships to generate more employment opportunities and quality education. Strenthening the existing social security system to extend its service to the most underprivileged people in the society is also required.

The importance of strengthening healthcare services and financial satisfaction was further informed by the findings of the quantile regression analysis, which showed that people older than 65 have a negative association with the MDWI, meaning that as people get older, their level of wellbeing declines. This might be because of older people suffer from various health issues and face financial issues after retirement. The United Nations (2023) noted that 8% of the total population in Bangladesh is above sixty years old, which will increase to 22% by 2050 (HelpAge, 2023). Therefore, addressing the concerns of senior citizens in the country needs urgent attention from the policymakers.

Third, MDWI results show that the level of wellbeing differs across the administrative divisions in Bangladesh. For example, the Rangpur division shows the lowest level of wellbeing, which is also confirmed by a negative and significant association of the MDWI and the Rangpur division by quantile regression. Therefore, the development of policies should include a focus on targeting balanced regional development. Regional areas where people may suffer from lack of economic and education opportunities, lack of health and education infrastructure and services, high crime rates, and governance-related concerns need the government’s attention and action to effectively address these concerns.

Fourth, this study found that the level of education positively and significantly affects MDWI in Bangladesh. Therefore, it is necessary for policymakers to initiate appropriate measures for universal and quality education by establishing more educational institutions and employing qualified professionals.

By focusing on critical aspects of wellbeing such as health, employment, education, safety, and governance, Bangladesh will be able to speed up its trajectory to achieve the global SDG agenda while ensuring the enhanced wellbeing of the Bangladeshi people. Furthermore, these findings may provide useful policy insights to enhance the wellbeing of people in other countries with similar socioeconomic backgrounds.