Vol.:(0123456789)
Social Indicators Research (2020) 148:1–46https://doi.org/10.1007/s11205-019-02187-9
1 3
Approaches and Alternatives to the Wealth Index to Measure Socioeconomic Status Using Survey Data: A Critical Interpretive Synthesis
Mathieu J. P. Poirier1,2 · Karen A. Grépin3 · Michel Grignon4
Accepted: 3 September 2019 / Published online: 12 September 2019 © The Author(s) 2019
AbstractMonitoring progress towards the Sustainable Development Goals by 2030 requires the global community to disaggregate targets along socio-economic lines, but little has been published critically analyzing the appropriateness of wealth indices to measure socioeco-nomic status in low- and middle-income countries. This critical interpretive synthesis ana-lyzes the appropriateness of wealth indices for measuring social health inequalities and provides an overview of alternative methods to calculate wealth indices using data cap-tured in standardized household surveys. Our aggregation of all published associations of wealth indices indicates a mean Spearman’s rho of 0.42 and 0.55 with income and con-sumption, respectively. Context-specific factors such as country development level may affect the concordance of health and educational outcomes with wealth indices and urban–rural disparities can be more pronounced using wealth indices compared to income or con-sumption. Synthesis of potential future uses of wealth indices suggests that it is possible to quantify wealth inequality using household assets, that the index can be used to study SES across national boundaries, and that technological innovations may soon change how asset wealth is measured. Finally, a review of alternative approaches to constructing household asset indices suggests lack of evidence of superiority for count measures, item response theory, and Mokken scale analysis, but points to evidence-based advantages for multiple correspondence analysis, polychoric PCA and predicted income. In sum, wealth indices are an equally valid, but distinct measure of household SES from income and consumption measures, and more research is needed into their potential applications for international health inequality measurement.
Keywords Wealth index · Principal components analysis · Demographic and health surveys · Socioeconomic status · Critical interpretive synthesis · Low- and middle-income countries
* Mathieu J. P. Poirier [email protected]
Extended author information available on the last page of the article
2 M. J. P. Poirier et al.
1 3
1 Introduction
To evaluate global progress in achieving the Sustainable Development Goals (SDG) by 2030, there is a need to disaggregate key indicators according to the socioeconomic status (SES) of households. Goals of ending poverty in all its forms everywhere and of reducing income inequality within and among countries take aim at SES directly, while several goals targeting health and education outcomes now aim to reduce socioeconomic inequalities (United Nations 2015). Nevertheless, in many countries, and especially among neglected populations and low- and middle-income countries (LMICs), reliable and timely data on income and consumption1 are not always available. In addition to missing data, there are challenges in using income or consumption measures in many LMICs, since income can be highly variable from month to month or difficult to accurately measure (Bollen et al. 2002). Alternatively, consumption data, such as that measured by the Living Standards and Measurement Studies, can be extremely time consuming and expensive to collect (Sahn and Stifel 2003).
Given the challenges in measuring SES with income and consumption, proxy indicators have been developed. In global health, the key proxy measure is the wealth index. Wealth indices use information about household durable assets, such as housing materials, toilet or latrine access, phone ownership, or agricultural land and livestock, which are regularly collected in most household surveys to create an index of household wealth. Their use has become widespread in large part because of the pre-existing availability of data measuring household durable assets in key standardized household surveys which span decades and cover nearly all LMICs of the world, such as the Demographic and Health Survey (DHS) and Multiple Indicator Cluster Surveys (MICS). Despite the near ubiquity of use of the wealth index in global health research, debates over which calculation method results in the best proxy for income or consumption, and even whether wealth indices should be consid-ered as SES measures that are fundamentally distinct from income or consumption remain open questions (Howe et al. 2009; Sahn and Stifel 2003).
Hundreds of manuscripts have used the wealth index to examine topics ranging from malnutrition (Mohsena et al. 2010; Sahn and Stifel 2003), educational attainment (Booy-sen et al. 2008; Nwaru et al. 2012), malaria transmission (Chuma and Molyneux 2009; Rohner et al. 2012), and poverty (Harttgen and Vollmer 2013; Zeller et al. 2006). For fif-teen years, the overwhelming majority of researchers creating these indices have followed the method developed by Filmer and Pritchett (2001) that summarizes multi-dimensional information on ownership of various household assets using principal components analysis (PCA) (Filmer and Scott 2012). This innovative application of PCA to the measurement of household wealth using DHS surveys allowed researchers to convert a series of ownership variables, many of which were binary (yes/no) or categorical (roof material, e.g.), into a continuous SES gradient (Rutstein 2008).
The PCA approach provides a way to go beyond simple sums of asset ownership by orthogonally layering linear combinations of the variables with maximum variation. More precisely stated, the covariance matrix underlying the structure of the data is used to solve for coefficient vectors for each independent variable such that each layer (or principal com-ponent) produces the direction of greatest variance. Other applications of PCA, such as factor reduction techniques, make use of several of these layered combinations ordered by
1 Also referred to as household expenditure or consumption expenditure.
3Approaches and Alternatives to the Wealth Index to Measure…
1 3
the degree of underlying data variance (i.e. eigenvalues), sometimes visually inspecting a scree plot for changes in slope to decide how many components to keep. In calculat-ing asset wealth, however, only the first principal component (which extracts the largest amount of information from the underlying asset data) is typically used as a measure of the “size” of the underlying structure of SES and ordinal data is often recoded as several binary dummy variables (Kolenikov and Angeles 2009).
Since the publication of Filmer and Pritchett’s (2001) foundational study, many researchers have focused on proving the utility and improving the process of this origi-nal method, while others have proposed alternative methods of wealth index construction. The only systematic review yet published on the topic of whether wealth indices function as effective proxies for household consumption found only weak to moderate association between the two measures (Howe et al. 2009). Other studies have compared different meth-ods of calculating wealth indices in isolation (Filmer and Scott 2012; Kolenikov and Ange-les 2009), but have not extended findings of strength of association between different SES measures to the theoretical questions of what exactly wealth indices are measuring and under what conditions they are appropriate measures. In sum, there has been no compre-hensive synthesis of the evidence and debates surrounding the method which continues to be the standard for constructing a proxy for household SES in lieu of consumption or income data.
This study systematically collected and synthesized information from the diverse bodies of literature examining wealth indices to evaluate two primary research questions. First, under what conditions is the use of wealth indices appropriate when measuring health ine-qualities using household surveys in LMICs? Second, what alternative methods of calcu-lating wealth indices are available and how do they compare to the most commonly used wealth index calculation method? This study does not aim to rank the various methods used to measure SES or select a method that dominates the others under all circumstances, but does aim to map these tools to normative choices and values. The findings of this study should be of particular interest to global health researchers, who should be aware that there is no gold standard for measuring household SES and that the choices they make regarding how to measure this latent and disputed concept have significant implications for the research they conduct, the policies they inform, and ultimately, the SDGs we aim to achieve.
2 Methods
This critical interpretive synthesis (CIS) integrates the diverse literatures informing the theoretical foundations of the wealth index, the appropriateness of its use in the field, and alternative methods of wealth index calculation. Since many of the constructs underpin-ning this research have yet to achieve universal definitions and the relevant literature is dis-persed throughout field-specific journals of economics, demography, epidemiology, global health, and sociology; a systematic review is neither ideal or appropriate (Gough et al. 2017; Higgins and Green 2011). This is because asset wealth is defined and calculated in a multitude of ways, the “gold standard” it is evaluated against is highly field-dependent, and even when the same methodology and comparator are used, methods used to evaluate performance can be incomparable from study to study. In other words, what is needed is an interpretive synthesis rather than an aggregative synthesis. Because of these challenges, CIS—a method created to assemble findings from a complex body of evidence to inform
4 M. J. P. Poirier et al.
1 3
policy in a theoretically grounded manner (Dixon-Woods et al. 2005, 2006)—was used fol-lowing established norms within the health policy literature (Ako-Arrey et al. 2016; Boyko et al. 2012; Ellen et al. 2018; Moat et al. 2013).
The compass questions guiding the initial search and article evaluation were whether the standard DHS wealth index should retain its status as the primary method for estimat-ing asset wealth in LMICs and whether the different contexts in which it is used affect its concordance with alternative SES measures. Constant reflexivity in the search and evalu-ation process resulted in the incorporation of several emerging themes, including delving into the ways in which wealth indices differ from income and consumption measures, a specific focus on how the urban–rural divide affects the choice of SES measure, and the possibilities, challenges, and advances in the effort to extend the use of wealth indices to the study of international health inequalities.
Guided by these compass questions, an initial search strategy broadly targeted articles comparing different methodologies for constructing wealth indices—especially as they related to the DHS wealth index. Specifically, initial searches of EconLit, Database of Abstracts of Reviews of Effects, PubMed, and Google Scholar in September 2015 focused on terms of “wealth index” “asset index” “principal components analysis”, “survey”, and “wealth” restricting searches to years following the publication of Filmer and Pritchett’s foundational article in 2001. Articles focused on the use of PCA in clinical research, imag-ing research, and any other unrelated applications were excluded from the review. In addi-tion, applied studies that use wealth indices without comparing results with at least one other measure of SES were excluded. After evaluating titles and abstracts for relevance, bibliographies were combed for any studies that were not identified through database searches.
This stage of literature search was followed by a first stage of synthesis, workshopping of initial findings at the McMaster University Centre for Health Economics and Policy Analysis (CHEPA), and consulting with content experts. Following this stage of article evaluation, a second systematic search was conducted in September 2018 following the search strategy outlined in Fig. 1 and the same inclusion and exclusion criteria as the first search. A comprehensive screen of titles and abstracts was possible for each database except for Google Scholar, which was screened until saturation was reached and article titles were no longer relevant. This resulted in a total of 53 articles included for synthe-sis, of which 11 articles could be used for the quantitative comparison of wealth indices,
Fig. 1 Flow chart of article inclusion process
5Approaches and Alternatives to the Wealth Index to Measure…
1 3
income, and consumption. Detailed information from each article, including SES measures investigated, countries of study, academic discipline, key themes, and evidenced used for synthesis were extracted into a table presented in Appendix Table 3.
2.1 Analysis
The data gathered through the systematic search strategy were analyzed through a multi-stage process. Specifically, measures of SES, discipline, study design, countries of study, key ideas, and specific contributions to CIS were first extracted into Appendix Table 3, and then organized and synthesized according to emergent themes. These emergent themes led to the division of results into four sections—the utility of the wealth index as a proxy for income and consumption, the performance of the wealth index as a measure of social welfare, the appropriateness of use of wealth indices in the field, and alternative methods of constructing wealth indices. Within each of these sections, the extracted information is presented according to established CIS practices in a format that was originally adapted from meta-ethnographic review (Ako-Arrey et al. 2016; Boyko et al. 2012; Dixon-Woods et al. 2005, 2006; Ellen et al. 2018; Moat et al. 2013). Key themes and concepts for each subsection are presented (reciprocal translational analysis), then contradictions between studies are examined (refutational synthesis), and finally, a general interpretation of find-ings grounded in the literature is proposed (lines-of-argument synthesis). In practice, this means synthesizing key introductory information, presenting qualitative data supporting and opposing the concept under study, and proposing an overall interpretation of the state of published research for each subsection. The content of the study data extracted was also continually evaluated against the credibility of each study, as determined by the strength of supporting data, methods used to generate results, and appropriateness of conclusions with regards to the results.
In order to synthesize data on alternatives to the standard PCA approach of calculat-ing asset indices, the merits of alternative asset indices were evaluated for their statistical validity, ease of calculation, and validity of results; all of which had to be supported by empirical research in a diversity of settings. Statistical validity examined issues such as the statistical assumptions underlying each method and issues that categorical, ordinal, and interval variables could have on the calculation of the index. Ease of calculation evaluated how much training was necessary to begin using the method, how dependent the method was on human judgement, and whether the method was supported by statistical packages. Validity of results do not rely on any one gold standard, but rather synthesize informa-tion from alternative SES measures, health outcomes, and contextual social factors. This resulted in an evaluation framework that could be consistently applied to asset index calcu-lation methods, despite some methods having more published evaluations than others.
The results of the CIS are presented according to the themes that emerged from this analysis. The complex relation of wealth indices with income and consumption measures is discussed first, including the most complete compilation of quantitative comparisons of these three SES measures yet assembled. Once wealth indices’ relation to these traditional measures of household SES is established, a synthesis of studies evaluating the perfor-mance of the wealth index as a measure of social welfare is presented. This is followed by a discussion of the appropriateness of use of wealth indices with a special focus on urban–rural issues (i.e. are wealth indices applicable across the urban–rural divide?), alter-ations to the standard approach, extension to the study of multiple countries, and emerg-ing trends and opportunities for future research. The final section then evaluates all major
6 M. J. P. Poirier et al.
1 3
alternatives to the DHS wealth index, with a critical interpretation of the merits and weak-nesses of each method.
3 Results
3.1 Income, Consumption, and Wealth Indices
Wealth indices are generally viewed as a measure of long-term wealth or SES, but not of short-term poverty, income, or consumption (Filmer and Pritchett 2001; Howe et al. 2012) because the household assets on which they are based are accumulated gradually over time and are unlikely to change rapidly even in periods of shifting income or consumption pat-terns. This confers the advantage of an index which is far more stable than income and consumption, but may also obscure real improvements (or declines) in household living standards over the short and medium term (Booysen et al. 2008). The inclusion of spend-ing on household durables in the calculation of consumption measures means that there is overlap what is being measured, but there is a notable difference between the amount a household is willing to pay for an asset, and the implicit utility derived from the ownership of that asset.
Despite these distinctions, wealth indices are frequently compared to consumption data based on the argument that it is the most accessible and closely related comparator with which to measure their performance (Aryeetey et al. 2010; Howe et al. 2012). Even though many authors default to household consumption as a gold standard measure of SES, report-ing errors are known to affect even the most carefully planned and executed surveys due to recall error, exclusion of some expenses, choice of deflator, and currency exchange fluc-tuations (Bollen et al. 2002; Kolenikov and Angeles 2009; Moser and Felton 2007; Sahn and Stifel 2003). In addition to being compared to income and consumption, wealth index data has even been used as a counterpoint to national accounts data. A heated debate over whether an African growth “miracle” occurred in the 1990s was sparked due to compari-sons of well-being based on asset indices, which had grown considerably, with well-being based on national accounts, which had not. Further analysis of this mystery revealed that factors such as new cheap imports of household durables from Asia and the tendency of household asset prices to drop over time were driving this discrepancy, but to this day there are many dissenting opinions and uncertainty over whether welfare has truly improved (Johnston and Abreu 2016).
A major difference between the wealth index and other measures of SES is that the for-mer is based on household assets and cannot be expressed in per-capita units. Other meas-ures of SES are not necessarily superior in this regard, since intrafamilial distribution of income is often highly unequal and consumption is usually inexactly divided into house-hold equivalents using one of several methods (Aaberge and Melby 1998). This means that the wealth index is more closely related to household economies of scale models than per-capita consumption models, reinforcing the idea that it is tracking a separate, but equally valid construction of SES (Filmer and Scott 2012). Nevertheless, there is evidence that some conditions improve the concordance of the two measures. There is evidence, for example, that consumption data tracks wealth indices more closely in middle income coun-tries, and especially if a greater variety of assets are included (Howe et al. 2009). Similar findings suggest that asset indices and consumption expenditure are more closely related when a higher percentage of consumption is captured by assets included in DHS surveys
7Approaches and Alternatives to the Wealth Index to Measure…
1 3
and that they are more highly correlated in countries where the average share of non-food expenditures is high (Filmer and Scott 2012). Despite the paucity of knowledge about the degree of concordance between income, consumption, and wealth indices, few studies have quantified this relationship in a systematic manner.
In general, there are two approaches to comparing income and consumption rakings with wealth index rankings—comparison of ordered subgroups such as quintiles2 or ter-ciles, or comparison of entire distributions. Studies opting for the first approach may use consumption or income as a “gold standard” and quantify the percentage of households missing from the poorest subdivision as errors of exclusion (Aryeetey et al. 2010). As one example, this was done in Turkey, finding that a wealth index was moderately associated with consumption and income, with 54.1% being in the lowest quintile for both wealth index and consumption, and 47.1% in the lowest quintile for both wealth index and income (Ucar 2015). For the second approach, the most common method of measuring distribu-tional associations in this literature is Spearman rank correlation, which is a nonparametric measure of association varying between −1 and +1 for two variables across a ranked dis-tribution (Spearman 1904). The second distributional approach was chosen as the measure of interest for this CIS because rank correlations take the entire distribution into account rather than losing granularity of data through grouping. Additionally, direct comparison of population groupings would not have been possible because the construction of subgroups varied too much for systematic comparison. Therefore, only the studies explicitly reporting Spearman correlation coefficients were included in the meta-analytic tables (Tables 1, 2).
The results of the pooled Spearman rank correlation coefficients (Table 1) indicate wide variability and overall moderate agreement of wealth indices with consumption or income. Spearman’s rho values ranged from 0.34 to 0.84, with a sample-size weighted average of 0.55 for consumption data and 0.42 for income data. Since wealth index constructions can differ even when the same data source and method are used because of decisions such as asset inclusion, and income and consumption comparators can also vary according to the calculation methods used, a range of correlation coefficient magnitudes is not unexpected. It is notable that besides Ferguson et al.’s (2003) two country comparison, no one has yet examined the relationship between all three SES measures in more than one country. There appears to be a moderate association between wealth indices and both income and con-sumption, allowing us to move on to understanding more about how the index relates to health and social welfare outcomes, when it is appropriate to use, and the alterations to the index that are possible for researchers.
3.2 The Wealth Index and Social Welfare
Rather than using consumption or income comparators, many researchers appraise the per-formance of wealth indices by examining their relationship with health or educational out-comes. Decades of research spanning nearly every country of the world have documented inequalities in health and educational outcomes associated with SES. Some have claimed that wealth indices may be more directly associated with these outcomes than household consumption or income because health and education outcomes are more significantly affected by long-run household SES than by monetary highs or lows (Mohanty 2009). This
2 All sampled households divided into fifths in order of the raw wealth index score. Wealth index quintiles are included in all DHS survey datasets and are commonly used as the primary measure of household SES.
8 M. J. P. Poirier et al.
1 3
Table 1 Spearman rank correlation coefficients for wealth indices with household consumption and income
Authors Country Comparator Data source Sample size Spearman’s rho
Ferguson et al. (2003)
Peru Consumption LSMS (2000) 4000 0.73
Pakistan Consumption PIHS (1991) 4752 0.34Filmer and Pritch-
ett 2001Nepal Consumption NLSS (1996) 3372 0.64
Indonesia Consumption DHS (1994) 16,242 0.56Pakistan Consumption PIHS (1991) 1192 0.43
Filmer and Scott (2012)
Albania Consumption ALSMS (2002) 3598 0.47
Brazil Consumption BPPV (1997) 4940 0.72Ghana Consumption GLSS (1992) 4522 0.43Nepal Consumption NLSS (1996) 3373 0.48Nicaragua Consumption EMNV (2001) 4191 0.71Panama Consumption PENV (1997) 4945 0.70Papua New
GuineaConsumption PNGHS (1996) 1144 0.47
South Africa Consumption SAIHS (1993) 8791 0.67Uganda Consumption UNHS (2000) 10,696 0.55Vietnam Consumption VLSS (1993) 4800 0.61Zambia Consumption LCMS (2004) 19,247 0.39
Lindelow (2006) Mozambique Consumption DHS (1997) 8250 0.37McKenzie (2005) Mexico Consumption ENIGH (1998) 10,777 0.84Opuni et al. (2011) Tanzania (men) Consumption KHDS (2004) 691 0.61
Tanzania (women) Consumption KHDS (2004) 833 0.59Sahn and Stifel
(2003)Ivory Coast Consumption CILSS (1988) 2169 0.51
Ghana Consumption GLSS2 (1988) 3192 0.43Ghana Consumption GLSS3 (1992) 4552 0.42Jamaica Consumption JSLC (1998) 7375 0.39Madagascar Consumption EPM (1994) 4800 0.50Nepal Consumption NILSS (1995) 3388 0.55Pakistan Consumption PIHS (1991) 4794 0.42Papua New
GuineaConsumption PNGHS (1996) 1396 0.47
Peru Consumption ENNIV (1994) 3623 0.71South Africa Consumption SAIHS (1994) 8848 0.71Vietnam Consumption VNLSS (1993) 4800 0.55Vietnam Consumption VNLSS (1998) 5999 0.67
Simple mean 5461 0.55Weighted mean 0.55Ferguson et al.
(2003)Peru Income LSMS (2000) 4000 0.72
Pakistan Income PIHS (1991) 4752 0.16Nkonki
et al. (2011)South Africa Income Good Start (2008) 133 0.42
Balen et al. (2010) China Income Wuyi (2006) 258 0.27
9Approaches and Alternatives to the Wealth Index to Measure…
1 3
reasoning is often applied to outcomes such as childhood stunting which take many years to develop, but also applies to other social welfare outcomes (Filmer and Pritchett 2001; Sahn and Stifel 2003).
One of the first multi-country comparative studies suggested that the use of wealth indi-ces resulted in smoother declines in stunting by wealth quintiles when compared to house-hold consumption in 10 LMICs (Sahn and Stifel 2003). This evidence is supported by the correlation of wealth index quintiles with low birthweight, education level, and occupation in the Vietnamese context, indicating that the index was capturing both a measure of social class and health outcomes (Vu et al. 2011). Another evaluation used Bayesian informa-tion criterion to predict fertility rates with several SES proxies, finding that a wealth index performed better than all other measures, including consumption measures (which pre-dicted almost no variation in fertility) (Bollen et al. 2002). There is weaker evidence from a wealth index of a Chinese community, finding only low to moderate correlation with maternal and child health indicators; although both an occupational index and educational index found equally weak associations (Nwaru et al. 2012).
In general, there is some evidence that asset measures may increase the magnitude of social health inequalities. Pro-rich inequalities in immunizations, maternity care, institu-tional deliveries, and hospital visits were greater when measured with a wealth index than consumption data in Mozambique (Lindelow 2006). In Tanzania, the use of a wealth index instead of household expenditure resulted in a statistically significant change in concentra-tion index for AIDS mortality in men, but the effect was small and made no difference for women (Opuni et al. 2011). A focused review of SES ranking specifically for tuberculosis
Table 1 (continued)
Authors Country Comparator Data source Sample size Spearman’s rho
China Income Laogang (2006) 246 0.27Simple mean 1878 0.37Weighted mean 0.42
Table 2 Spearman rank correlation coefficients for polychoric PCA with household consumption and income
Authors Country Comparator Data source Sample size Spearman’s rho
Reidpath and Ahmadi (2014)
Vietnam Consumption WHS (2003) 4154 0.57
SIMPLE MEAN 4154 0.57WEIGHTED MEAN 0.57Ward (2014) China Income CHNS (1989) 4400 0.35
China Income CHNS (1991) 4400 0.40China Income CHNS (1993) 4400 0.41China Income CHNS (1997) 4400 0.33China Income CHNS (2000) 4400 0.42China Income CHNS (2004) 4400 0.43China Income CHNS (2006) 4400 0.44
Simple mean 4400 0.40Weighted mean 0.40
10 M. J. P. Poirier et al.
1 3
surveys concluded that wealth indices more consistently identified inequities in health than income or consumption surveys (Van Leth et al. 2011). Another team investigating insec-ticide-treated net ownership in Kenya found a mixed picture of larger inequalities in urban areas using a wealth index compared to consumption, but smaller inequalities in rural areas (possibly due to free net distributions in rural areas), concluding that neither the wealth index or consumption index approach is superior for health research in LMICs (Chuma and Molyneux 2009).
The evidence supporting larger social health inequality magnitudes when using asset measures did not translate to the outcome of seeking medical care, with wealth indices and consumption levels generating almost identical results in a large cross-sectional coun-try comparison (Filmer and Scott 2012). There was greater health-seeking behavior found among the relatively poor in these countries, although this is hypothesized to be a product of the poorest quintile’s disproportionate share of illness. This theory is supported by the finding that the highest levels of child mortality are not uniformly found in the poorest quintiles of the consumption model, but are always found in the poorest quintiles of wealth index models—a statistically significant difference-in-difference (Filmer and Scott 2012). In general, the tendency of publicly-provided services tending to be of more importance in the lower end of the SES gradient and private goods tending to be more important for the upper end (Booysen et al. 2008) may have some impact on inequalities in health and healthcare seeking behaviour.
The outcome of educational attainment has similarly mixed results. One study using wealth index and consumption data to rank households in a multi-country data exercise found a statistically significant educational inequality in 7 of 11 countries included, with the DHS wealth index most often resulting in larger inequalities (Filmer and Scott 2012). In Ghana, however, the wealth index was modestly correlated to parental education levels (maternal r = 0.32, paternal r = 0.36), explaining only 14% of parental education and occu-pation variance (Doku et al. 2010). Using education as a ranking variable rather than an outcome yields similarly mixed results. A multi-country study found that wealth indices are not statistically different than maternal education as a ranking variable for quantify-ing inequalities in vaccination coverage, although wealth index inequalities were slightly smaller,3 and some countries had much larger inequalities using one or the other4 (Arse-nault et al. 2017).
Considering these results as a whole, we can conclude that the widespread practice of comparing wealth indices to income or consumption in studies of social inequalities of health and educational outcomes produces some contradictory outcomes, but generally points in the same direction as income and consumption research. That is, poor health and educational attainment is found among lower SES populations regardless of how SES is measured. However, there is an undeniably tautological reasoning underlying many com-parisons. Even if wealth indices are an equally valid, but separate measure of household SES than income and consumption; then verifying both the validity of wealth indices through the presence of health and educational inequalities and confirming the presence of health and educational inequalities using wealth indices risks being dismissed as circu-lar and baseless evidence. Larger inequalities in health outcomes such as child mortality
3 Haiti had larger inequalities using education [SII = 0.34 95% CI = 0.20, 0.48] than the wealth index [SII = 0.10 95% CI = 0.04, 0.24].4 Mozambique had larger inequalities using wealth index [SII = 0.30 95% CI = 0.22, 0.37] than maternal education [SII = 0.16, 95% CI = 0.09, 0.24].
11Approaches and Alternatives to the Wealth Index to Measure…
1 3
clearly lend face validity to a measure of SES, but the many causal pathways that may lead one context to have larger inequalities using wealth indices, consumption, or income should be explored in all their complexity rather than relying on the unidimensional logic of a “true” effect size being the largest.
3.3 Challenges and Opportunities
3.3.1 Urban–Rural Considerations
Since the DHS wealth index was first developed, there have been concerns over compa-rability of results between urban and rural areas (Filmer and Pritchett 2001). Many have expressed concerns that since urban households are more likely to own many assets and are more likely to benefit from publicly provided assets such as piped water, they will be inappropriately classified as wealthier than comparable rural households (Booysen et al. 2008). Others counter that this is not misclassification, but an accurate representation of the relative affluence of urban households (Vyas and Kumaranayake 2006). Adding to this complexity, there are indications that unmet healthcare need can be underestimated in rural areas and overestimated in urban areas (Mohanty 2009). Beyond misclassification errors, the issue is complicated by the fact that assets like chickens or bicycles are an indicator of relative wealth in rural areas, while also being an indicator of relative poverty in urban areas (Chuma and Molyneux 2009).
Regardless of whether it is an accurate representation of household SES or not, urban–rural disparities appear to be larger when SES is measured using a wealth index than income or consumption measures. The difference in urbanization between the poor-est and richest quintiles can be as large as 75% in a wealth index compared to 22% in an expenditure model in Albania, with several other countries also having large discrepancies in SES ranking due to urban status (Filmer and Scott 2012). Perhaps the most dramatic example of these vast differences was demonstrated in Kenya, where a wealth index placed no rural households in the richest quintile and only one rural household in the second rich-est (Chuma and Molyneux 2009).
The mechanism for this urban divide can largely be attributed to a combination of rural households having fewer assets, more commonly owned assets, and agricultural assets often being assigned negative factor loadings. As one illustrative example, there is a vil-lage in Guinea-Bissau where (unlike the rest of the country) portable gas stoves are highly desired, and therefore behave as a normal good,5 but because that village is relatively poor compared to other villages, the wealth index scoring of gas stoves is negative (Johnston and Abreu 2016). Similarly, owning a common asset will usually imply a negative scor-ing, which could perversely rank a household as poorer than one lacking the asset at all (Wittenberg and Leibbrandt 2017). Even obtaining reliable information on rural assets is complicated by survey respondents often having difficulty answering questions about the number of hectares of agricultural land owned or even whether they live in an urban or rural area (Chakraborty et al. 2016).6 In response to these concerns, a variety of strategies to identify and address urban–rural issues with wealth indices have been proposed.
5 i.e. a good for which demand increases as when SES increases.6 This can be resolved by having survey teams classify urban and rural areas rather than eliciting the infor-mation from survey respondents.
12 M. J. P. Poirier et al.
1 3
One common strategy to identify relative affluence in both urban and rural areas is simply to split the sample into two groups and calculate a rural wealth index and an urban wealth index. In fact, the standard DHS approach to dealing with urban–rural issues is to regress both an urban-only sample and a rural-only sample against the com-plete sample to obtain a modified index influenced by all three factor loadings (Rutstein 2008). This strategy can lead to agricultural assets having positive weights for rural households and negative weights for urban households (Ward 2014), but the magnitude of effect appears to depend heavily on the setting. One study comparing a rural-only sample wealth index to one calculated for the full sample in Zimbabwe found a Spear-man rank correlation coefficient of 0.862 (95% CI 0.854–0.869) between the two indices, indicating a fairly high association (Chasekwa et al. 2018). A Ghanaian study using con-sumption as a comparator found a weaker concordance: the wealth index “misclassified” 63% of consumption-poor households in urban settings compared to 46% in semi-urban settings, and 53% in rural settings (Aryeetey et al. 2010). Another Chinese study using income as a comparator found the same relatively weak association to household income (Spearman’s rho 0.27) in both a rural and peri-urban village, and a similar proportion of variation captured by the wealth index (27.8% vs. 24.3%) for both villages (Balen et al. 2010).
The different strategies used to address urban–rural biases also appear to have a mod-erate effect. One small Zambian study found that dropping all assets which are more likely to be found in an urban household did not significantly affect the overall variance explained by the wealth index (Boccia et al. 2013). Another large multiyear pooled analysis in China that found negative factor weights for all agricultural assets suggested that sec-ondary principal components weights (which were positive) could be used in these cases, although dropping all agricultural assets appeared to make little difference, with 90% of households being classified in the same quintile (Ward 2014). A study designed to evalu-ate this approach in India found 39% of households to be classified in the same quintile, 50% to have moved to an adjacent quintile, and 10% to have moved to the farthest quin-tile (Mohanty 2009). Alternatively Ngo and Christiaensen (2018) have proposed adding a small number of binary consumption variables such as food and clothing purchases, find-ing that it increased identification of consumption-poor households in rural settings by 9%, but made no difference in urban settings. In sum, urban–rural differences should always be monitored and can be addressed through a number of approaches, but do not present an insurmountable obstacle to the use of wealth indices.
3.3.2 Robustness to Changes in the Asset Mix
Another common criticism of wealth indices relates to their reliance on assets which have direct impact on health, such as water and sanitation quality or food availability in research on the associations of SES and health (Homenauth et al. 2017). There is some evidence for this effect, with one study finding that dropping household construction variables from a wealth index in Uganda resulted in a significant association with mosquito human bit-ing rate becoming insignificant, even though the two indices are highly correlated7 (Tust-ing et al. 2016). Another set of researchers in Zambia built an alternative index without food-related variables (which may have affected tuberculosis outcomes of interest directly)
7 Spearman’s rho = 0.93.
13Approaches and Alternatives to the Wealth Index to Measure…
1 3
and found no significant difference with the wealth index using all variables (Boccia et al. 2013). A low-to-moderate effect is supported by a 10-country World Bank comparison of three alternative indices that exclude direct determinants of health and factors provided at the community-level, in which only 18% of households were categorized in a different wealth quintile with most of these shifting to an adjacent quintile (Houweling et al. 2003); as well as the use of a simplified asset list dropping various country-specific, urban–rural specific, and agricultural questions with 16 surveys finding inter-quintile agreement rang-ing from 75 to 83% (Chakraborty et al. 2016).
Despite relatively strong concordance of indices based on difference assets, some measures of social health inequality may be sensitive to these changes. One study found up to a 60% change in the relative index of inequality for five health outcomes with alternative wealth indices, although the direction of change appeared to be random and was not significant in some countries (Houweling et al. 2003). Dropping assets can also be motivated by time and resource savings for survey collection teams. In one example comparing two simplified asset indices to the full index, there was almost perfect agree-ment (kappa value greater than 0.61) after reducing the number of variables from 111 to 24 variables in Honduras and from 111 to 21 variables in Senegal using an iterative ranking of factor loadings (Ergo et al. 2016). The limited effect of dropping variables is mirrored by an expanded set of assets collected to measure progress in the Millen-nium Villages Project failing to predict income poverty more effectively than the stand-ard DHS asset mix (Michelson 2013). Nevertheless, the variables that matter most vary according to the country context and reducing the number of accepted answers might not reduce the amount of time needed to survey a household because each variable may still require a separate question. In sum, changing the asset mix included in surveys may have a smaller effect than many anticipate, meaning that avoiding appearance of endo-geneity with health-related variables or simplifying a survey instrument can be done with appropriate care.
3.3.3 Future Applications
Research on the use of wealth indices is not limited to refining existing applications. One emerging area of research concerns extending the wealth index to the study of eco-nomic inequality research. Care must be taken before applying inequality measures to asset indices, however, because Gini coefficients can only be applied to the absence or presence of real assets, due to the inherent lack of scale for categorical variables (Witten-berg and Leibbrandt 2017). One of the earliest investigations into whether wealth index inequality8 was correlated to expenditure-based inequality in 31 Mexican states found a Spearman’s rho of 0.566 (about the same strength of association as food expenditure), and slightly stronger association than either a housing-based wealth index or a utility-based wealth index (McKenzie 2005). This method was recreated in China and evaluated ecologically against known consumption inequality, appearing to track the same pattern of rising inequality through the 1990s until a peak was reached around 2000, suggest-ing broadly shared growth and an eventual decline in urban and rural wealth inequality (Ward 2014). A more recent application in South Africa found wealth index inequality
8 Wealth index inequality is calculated as the proportion of variation of wealth explained by the first eigen-value.
14 M. J. P. Poirier et al.
1 3
fell from a Gini coefficient of 0.47 to 0.29 from 1993 to 2008, but cautions that the use of a negatively loaded first eigenvalue in the calculation of wealth index inequality could lead to this method performing poorly (Wittenberg and Leibbrandt 2017). Although there is clearly more research to be done on the limitations of wealth index inequality, this is an area of research which could grow rapidly given the increasing public interest on this topic.
Another commonly cited limitation of the wealth index is the perceived inability to make comparisons in wealth across countries. Since the wealth index in any given coun-try is a relative measure, comparisons across countries may neglect important differences in cultural and social values associated with household assets. Much of the reasoning behind this skepticism lies on claims that the assets contained in standardized household surveys cannot be relied upon in countries that traditionally value assets differently than others. Despite this assertion, a “traditional wealth index” constructed to represent Ken-yan cultural constructions of wealth was nearly identical to standard PCA of household assets (Opuni et al. 2011). Further support comes from a finding that a wealth index is more strongly correlated with locally identified factors indicating poverty (female-head-edness of household, dependency ratios, and household food insecurity) than household income (Michelson 2013). Another effort to construct a wealth index applicable to 21 Latin American and Caribbean countries using telephone survey data found generally encouraging results. Pooling wealth indices resulted in broadly applicable SES rankings from poorer countries like Peru to richer countries like Costa Rica, and the resulting rel-ative wealth quintiles were strongly correlated with years of schooling and self-reported income (Córdova 2008).
The largest effort to construct an asset index of worldwide comparability with wealth indices, however, comes from a team that overcame the difficulty of incomparability of many survey items by grouping accessories into cheap and expensive utensil categories (Smits and Steendijk 2013).9 This approach results in a wealth index applicable to 165 household surveys across 97 LMIC and is robust to removal of any region from analy-sis (Pearson correlation coefficient ≥ 0.996), removal of any time period (Pearson cor-relation coefficient ≥ 0.997), and removal of any one asset (Pearson correlation coeffi-cient ≥ 0.986).10 Furthermore, there is good agreement between the international wealth index and country-specific wealth indices, country-specific poverty levels, life expectancy, and most strong agreement with the Human Development Index. Finally, the authors assert that reasonable estimates of purchasing power parity (PPP) poverty levels can be placed at 30th percentile of the index equivalent to PPP$1.25 a day and 50th percentile at PPP$2.00 a day (Smits and Steendijk 2013). This international poverty line can be coupled with the finding that transitions out of poverty occur at the same rate using asset indices and house-hold income, with approximately 12–20% of the lowest quartile households transitioning to the highest quartile households after two years (Michelson 2013). These studies are break-ing new ground, but it appears that international poverty studies using wealth indices are becoming increasingly possible.
10 Although these results were obtained with ordinary PCA, sensitivity checks with MCA, factor analysis, and categorical PCA did not change the results.
9 Incidentally, this approach has also been used to compare assets over time for variables like landlines and cell phones, which can be combined into one "phone" asset in response to criticisms that the social signifi-cance of certain assets such as landlines, radios, and bicycles changes significantly over time (Wittenberg and Leibbrandt 2017; Harttgen and Vollmer 2013).
15Approaches and Alternatives to the Wealth Index to Measure…
1 3
Lastly, there are several research teams attempting to proxy wealth indices using new technologies. One team has developed a machine learning algorithm that can be used to roughly approximate wealth indices using phone usage characteristics in Rwanda and Afghanistan, although the models must be developed separately for each country and can-not be applied naively across borders (Blumenstock 2018). Another team has created a convolutional neural network trained on ground imagery that is able to predict 37–55% of variation in consumption and 55–75% of variation in asset wealth if trained separately for each country, although this drops to 19–52% and 24–71% if applied to other countries (Jean et al. 2016). This ground imagery method is slightly more predictive than phone-use estimation models, and the difference in estimation is likely due to the area’s wealth itself rather than directly identifying household features such as roofing materials directly. The fact that ground imagery is more highly correlated with a wealth index than with consump-tion also provides further evidence that a separate, but equally valid construct of SES is being captured by the method. In sum, ground-breaking research is being conducted into new ways to apply wealth indices to measuring SES inequality, to constructing high-qual-ity cross-country pooled sample analysis, and to using new technologies to measure house-hold SES.
3.4 Alternative Approaches
The standard wealth index constructed using PCA is not the only method used to measure SES using information of assets collected by household survey data. What follows is a short summary of the intersection of alternative approaches and a DHS-style wealth index, evaluating statistical validity, ease of calculation, and consistency of results supported by empirical research in a diversity of settings.
3.4.1 Count Measures
The most basic asset indices used for household survey data are simple asset counts. One Albanian team found that a wealth index was more highly correlated with consumption than a count measure consisting of water and sanitation provision, adequate housing provi-sion, less crowded dwellings, and minimum education of household head (Azzarri et al. 2005). Similarly, an early comparison of the DHS wealth index, consumption measures, and count measures as predictors of fertility rate found the simple count measures to have the second-best fit (after the wealth index) using Bayesian Information Criterion (Bollen et al. 2002). These outcomes are contradicted by a team in Bangladesh, asserting that using a basic count measure outperforms the DHS wealth index in discriminating households more at risk for stunting, wasting, and underweight (Mohsena et al. 2010). Their use of a simple count of radio, television, bicycle, motorcycle, telephone, and electricity to con-struct wealth quintiles resulted in 49.1% of households in the lowest SES quintile and only 4.2% of the highest SES quintile having all three outcomes of interest, while the wealth index produced equivalent percentages of 28.6% and 11.4%, respectively. However, this study was strongly disputed by another team using the same indices in Cote d’Ivoire with rigorous biometric measures of nutritional status while accounting for the effect of malar-ial infection, age, and residency; where the wealth index resulted in larger socioeconomic inequalities in anemia, stunting, and wasting in children and women of reproductive age
16 M. J. P. Poirier et al.
1 3
than the count score (Rohner et al. 2012). In sum, count measures may present an easily constructed and persuasive SES measure, but results are highly dependent on the judge-ment of the analyst of which household assets to include and may not be transferable to other contexts.
3.4.2 Multiple Correspondence Analysis
Many researchers have pointed out a fundamental flaw in the application of PCA to the types of variables needed to construct a wealth index; namely, that the technique is not meant to be applied to binary and categorical variables (Howe et al. 2012; Kolenikov and Angeles 2009). Since an average of 60% of household survey questions used to construct asset indices are binary, this is no minor limitation (Kolenikov and Angeles 2009). This limitation is commonly skirted by applying a qualitative judgement of supe-riority by an analyst recoding variables, possibly introducing bias and most likely affect-ing the fidelity of the final data. A long-standing alternative to PCA which does not have these inherent weaknesses is Multiple Correspondence Analysis (MCA), which has a similar approach of using a correlation matrix to determine “principal inertias” of the assets included for analysis and can be calculated using modules for most statistical packages (Booysen et al. 2008).
A large seven-country analysis of DHS data opted to use MCA rather than PCA because of these limitations, but found that despite some differences in variable weight orders, there was no significant difference between both indices (r = 0.953, p < 0.01) and the few households that were classified into different quintiles were restricted to one level higher or lower (Booysen et al. 2008). Another application of MCA in Kenya found it to be highly correlated to the DHS wealth index (r = 0.997, p < 0.01) with 93% of households placed in the same quintiles, although it explained the highest total variation of variables (47.3%) (Amek et al. 2015). Yet another com-parison of MCA and the DHS wealth index found that they were not significantly dif-ferent in year over year change and were both more strongly autocorrelated to them-selves than to household income in several sub-Saharan countries (Michelson 2013). In this case we conclude that although MCA has yet to significantly differentiate itself empirically from the DHS wealth index when applied in the field, its theoretical supe-riority in handling a diverse set of variables makes MCA a valid alternative measure of household SES.
3.4.3 Item Response Theory/Latent Trait Modeling
Seizing on the controversial application of PCA to non-continuous data, other research-ers have advocated the adoption of Item Response Theory (IRT), which is also referred to as Latent Trait Modeling (LTM). At a basic level, observed assets (whether they are dichotomous, polytomous, nominal, or ordinal) which demonstrate the most discrimi-nation according to a latent trait (SES) are given larger weights, and are then assessed for reliability with a non-parametric bootstrap (Vandemoortele 2014). Despite claims of differentiation, an independent 11 country comparison found rank correlations for the DHS wealth index and IRT between 0.95 and 1.00—the most highly correlated alternative measure in the study (Filmer and Scott 2012). Another empirical evalua-tion of this technique by a strong advocate of IRT on Malawian DHS data also found
17Approaches and Alternatives to the Wealth Index to Measure…
1 3
high correlation with a PCA index (Spearman rank correlation = 0.88) (Vandemoortele 2014). Furthermore, the two key assumptions of normal distribution of data and inde-pendence of variables offer no improvement to the existing PCA approach, and its cal-culation is acknowledged to be more time consuming (Vandemoortele 2014). Given these disadvantages and the lack of significant difference in the field, the DHS wealth index remains the more viable option until evidence of superiority can be presented.
3.4.4 Mokken Scale Analysis
Mokken Scale Analysis (MSA) is a nonparametric technique which relies on Guttman scales of items which are statistically determined to be increasingly “harder” to answer. Using a combination of positive ownership of assets with the difficulty of eliciting a posi-tive response, MSA is able to rank households along a latent SES gradient (Reidpath and Ahmadi 2014). Key assumptions include unidimensionality of SES, local independence of variables, monotonicity of responses, and invariant item ordering. An empirical applica-tion of the technique found very high Pearson product moment correlation with a poly-choric PCA index (r = 0.96) and a lower correlation to household expenditure (r = 0.59) (Reidpath and Ahmadi 2014), a result which the authors concluded was similar to the pat-tern observed for the DHS wealth index. The real or perceived downside of complexity of the technique with only marginal statistical effect may limit the widespread adoption of MSA, so the DHS wealth index also remains the more viable option of the two options at this time.
3.4.5 Polychoric PCA
As a response to the primary statistical vulnerability levelled against the DHS wealth index—its inappropriate application to non-continuous variables—an improved poly-choric PCA was proposed by Kolenikov and Angeles (2009). Criticizing Filmer and Pritchett’s technique for creating spurious correlation through the introduction of dummy variables and for losing directionality of ordinal data, Kolenikov and Angeles propose the use of a slightly amended multivariate technique, originally derived by the same statisti-cian as ordinary PCA. Not only is there greater statistical fidelity, but the status of not owning an asset is also taken into account. This additional information can be important in cases like indoor plumbing, which may only be missing from a small percentage of the poorest households of a population (Moser and Felton 2007). The key findings of the proof-of-concept study were that polychoric PCA demonstrated lower misclassification rates compared to consumption, explained a higher proportion of variance in asset owner-ship, was more robust to the number of categories used, and was more robust to changes in variable coding scheme than the Filmer and Pritchett PCA procedure (Kolenikov and Angeles 2009).
Interestingly, the standard PCA and polychoric PCA methods demonstrate divergent classifications at the lower end of the SES spectrum with increasing agreement of classifi-cation on the upper end of the SES spectrum (Kolenikov and Angeles 2009). An independ-ent comparison of the DHS wealth index with polychoric PCA using Bangladeshi DHS data also concluded that the DHS index lacks the ability to discriminate at the lower end of the spectrum due to its under-emphasis of common assets (Benini 2007). Despite this lower-end discrepancy, agreement remains very high. A Kenyan study found polychoric
18 M. J. P. Poirier et al.
1 3
PCA to be closely correlated with standard PCA (r = 0.991, p < 0.01) and to even more closely mirror MCA (r = 0.991, p < 0.01), while placing 87% of households in the same quintiles as standard PCA and 91% in the same quintiles as MCA (Amek et al. 2015). Another comparison conducted using Zimbabwean data found a Spearman rank coeffi-cient of 0.910 (95% CI 0.904–0.915) and 94% agreement between wealth quintiles between the DHS wealth index and polychoric PCA (Chasekwa et al. 2018). Similarly, Filmer and Scott’s 11 country comparison found both indices to be generally comparable (2012). There may be evidence of lack of robustness to variable loss, however, with an attempt to reduce a 17 item asset index to 11 items using polychoric PCA in Vietnam resulting in much lower concordance with both expenditure (r = 0.57 vs. r = 0.41) and an MSA-derived asset index (r = 0.98 vs. r = 0.68) (Reidpath and Ahmadi 2014).
As the only other method systematically compared to income and consumption by several studies, all available Spearman correlation coefficients between polychoric PCA wealth indices and either household consumption or income encountered in the litera-ture search are presented in Table 2. The results are not as robust as those presented in Table 1 due to both income and consumption comparisons being based on one study,11 but polychoric PCA appears to have an almost identical association as the DHS wealth index for both consumption (0.57) and income (0.40). Given that polychoric PCA over-comes the challenges relating to variable types, overcomes issues of “clumping” through greater discriminatory power at the lower end of the SES spectrum, and is integrated into several statistical packages, there is a strong case to be made for the superiority of this approach.
3.4.6 Predicted Income
A newly emerging technique overcomes the limits imposed by the ordinal nature of wealth indices by linking a country and year-specific predicted income to households accord-ing to their relative standing, as determined by a wealth index. An early application of a similar method using regressed prediction of consumption based on household assets found that it resulted in inequality levels in between those predicted by a wealth index approach and actual consumption, and that rankings of Mexican states by inequality were more similar to consumption than using a wealth index (McKenzie 2005). Since this early application, Harttgen and Vollmer (2013) have proposed a streamlined method, in which any wealth index is used to rank households into centiles or quintiles, and the resulting ordering is linked to an open access dataset estimating household income for 88 LMICs from 1993 to 2014.
The strength of this method is supported by studies finding more variation in stunt-ing prevalence using the predicted income approach (38%) compared to wealth quintiles (20%) (Fink 2016), and predicted income better predicting skilled birth delivery in a large 100-country study, with log-normalized predicted income explaining 51.6% of variation, wealth quintiles predicting 22.0%, and the raw wealth index predicting 12.8% (Joseph et al. 2018). It is also possible to compare health outcomes taking predicted income inequality into account using tools such as equiplots with this approach, revealing countries which have similar outcomes at any given income level, and others that are performing poorly at a given income level (Fink 2016). Furthermore, comparisons of all countries over time
11 Even so, the income comparisons include seven separate survey comparisons.
19Approaches and Alternatives to the Wealth Index to Measure…
1 3
reveals important trends such as countries that have succeeded in increasing skilled birth attendance in spite of stalled income growth, and those that have not improved outcomes even in times of sustained economic growth (Joseph et al. 2018). More study of this emerg-ing method is clearly needed including whether the predicted income is more closely asso-ciated to actual household income or the wealth index on which it is based, but the new avenues of research made possible by the approach warrants its inclusion in future studies.
4 Conclusion
The construction of a wealth index using household survey data must be conducted with an awareness that the methodology chosen to quantify SES using assets contained in the survey data has a significant effect on the results. More straightforward alterna-tives to constructing asset indices like count measures offer simplicity but may overly depend on context and analyst expertise. While more complex methods of MCA, IRT, MSA, polychoric PCA, and predicted income offer varying degrees of improvement of statistical validity, they may do so at the expense of simplicity with only marginal improvement in outcomes compared to the standard DHS wealth index. Taking all pub-lished alternatives and evidence into account, analysts striving for an alternative to con-structing a wealth index from household survey data can consider polychoric PCA as a method which meets the standards of statistical validity, ease of calculation, and valid-ity of results, with MCA as another valid alternative. If wealth rankings in a meaningful scale are needed, the predicted income approach based on either the DHS wealth index or any comparable alternative offers great promise but must also be investigated in a greater diversity of settings and applications.
Evidence gathered in this review lends support to the idea that wealth indices repre-sents a related, but distinct measure of latent SES from consumption or income measures. There is robust evidence linking the wealth index to health and educational outcomes at least as strongly as household consumption and income throughout the world. However, interpreting wealth indices as having a causal effect on health and educational outcomes cannot be taken as a given; especially with the knowledge that wealth indices, income, and consumption measures take aim at entirely separate models of SES. Long-known vulner-abilities to urban–rural distortions or changes in the asset mix included in surveys should always be considered, but with proper care, these vulnerabilities can be seen as ultimately informative rather than confounding. Future applications to inequality research, large-scale international studies, and the use of new technologies are promising prospects for which the groundwork has yet to be fully laid.
The main limitations of these conclusions stem from the paucity of research designed to answer these methodological issues specifically, rather than as a secondary research ques-tion dispersed throughout many fields. We are further limited by highly variable and some-times inconsistent definitions of key concepts, which in many cases such as asset wealth, even lack a commonly agreed-upon name. These limitations can only be overcome with greater research intensity and debate. Because of these limitations, a critical interpretive synthesis was the most appropriate choice to present the debates surrounding this method-ology in all its complexity. This presentation of key concepts, exploration of contradictions in the literature, and proposal of lines-of-argument synthesis aims to promote a shared
20 M. J. P. Poirier et al.
1 3
understanding of an emerging field of study across the multitude of disciplines that are involved in its development. Further strengths of the study include our inclusion and syn-thesis of more studies than any prior work on wealth indices, and the first systematic search and compilation of Spearman correlation coefficients between wealth indices and both con-sumption and income.
The implications of these findings to measuring progress in achieving the SDGs cannot be understated. Developing countries and neglected populations which lack consumption and income data will necessarily be studied using wealth indices as a proxy for SES. If we are to adequately measure progress in achieving equity-focused SDGs around the world for these populations, we must acknowledge the challenges in developing reproducible, rigor-ous, and easily implemented methodologies for constructing asset indices using household surveys. However, we can also look to the many strengths of the method, not the least of which is the increasingly real possibility of worldwide comparability of SES among all populations of the world. Further study of this possibility must account for the many poten-tial pitfalls in conducting research across national boundaries. Finally, it is remarkable that with the hundreds of studies using the wealth indices to measure health and social welfare outcomes, no study has yet systematically examined whether inequalities in health or social outcomes are larger in magnitude than would be measured using income or consumption in more than one country. Wealth indices have become the dominant method to measure SES in LMICs in the field of global health. Researchers using the method to develop sur-veys, analyze data, or interpret data for policymakers must understand its strengths, its lim-itations, the normative choices associated with the tool, and the potential to improve and extend the method to new areas of research.
Acknowledgements I gratefully acknowledge Dr. Emmanuel Guindon for helpful comments in the for-mulation and review of this research, members of the Centre for Health Economics and Policy Analysis (CHEPA) at McMaster University for contributions to the design of the study, and Dr. Michelle Dion for her insightful revisions.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 Interna-tional License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Appendix
See Table 3.
21Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
Dat
a ex
tract
ion
tabl
e fo
r crit
ical
inte
rpre
tive
synt
hesi
s
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Am
ek e
t al.
(201
5)PC
A; P
olyc
horic
PC
A; M
CAG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Ken
yac
MCA
is h
ighl
y co
rrel
ated
to P
CA (r
= 0.
997,
p <
0.01
), 93
% o
f hou
seho
lds p
lace
d in
the
sam
e qu
intil
es, a
nd
MCA
exp
lain
ed th
e hi
ghes
t tot
al v
aria
tion
of v
aria
bles
(4
7.3%
)Po
lych
oric
PCA
is c
lose
ly c
orre
late
d w
ith st
anda
rd P
CA
(r =
0.99
1, p
< 0.
01) a
nd M
CA (r
= 0.
991,
p <
0.01
). Pl
aces
87%
of h
ouse
hold
s in
the
sam
e qu
intil
es a
s st
anda
rd P
CA a
nd 9
1% in
the
sam
e qu
intil
es a
s MCA
Ars
enau
lt et
al.
(201
7)PC
A; M
ultid
imen
-si
onal
Pov
erty
In
dex
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
oA
rmen
ia; A
zerb
aija
n;
Ban
glad
esh;
Ben
in;
Bol
ivia
; Bur
undi
; B
urki
na F
aso;
Cam
-bo
dia;
Cam
eroo
n;
Com
oros
; Con
go;
Côt
e d’
Ivoi
re; D
emo-
crat
ic R
epul
ic o
f the
C
ongo
; Eth
iopi
a;
Gam
bia;
Gha
na;
Gui
nea;
Guy
ana;
H
aïti;
Hon
dura
s;
Indi
a; In
done
sia;
K
enya
; Kyr
gyzs
tan;
Le
soth
o; L
iber
ia;
Mad
agas
car;
Mal
awi;
Mal
i; M
oldo
va;
Moz
ambi
que;
Nep
al;
Nig
er; N
iger
ia; P
aki-
stan
; Rw
anda
; São
To
mé
and
Prín
cipe
; Se
nega
l; Si
erra
Le
one;
Taj
ikist
an;
Tanz
ania
; Tim
or
Leste
; Uga
nda;
Zam
-bi
a; Z
imba
bwe
c, d
, e, i
PCA
der
ived
wea
lth in
dex
is n
ot st
atist
ical
ly d
iffer
ent
than
mat
erna
l edu
catio
n or
mul
tidim
ensi
onal
pov
erty
in
dex,
wea
lth in
dex
ineq
ualit
ies s
light
ly sm
alle
r, an
d so
me
coun
tries
hav
e m
uch
larg
er in
equa
litie
s usi
ng o
ne
or th
e ot
her.
Hai
ti ha
d la
rger
usi
ng e
duca
tion
(SII
= 0.
34
95%
CI =
0.20
, 0.4
8) th
an th
e w
ealth
inde
x (S
II =
0.10
95
% C
I = 0.
04, 0
.24)
; Moz
ambi
que
larg
er u
sing
wea
lth
inde
x (S
II =
0.30
95%
CI =
0.22
, 0.3
7) th
an m
ater
nal
educ
atio
n (S
II =
0.16
, 95%
CI =
0.09
, 0.2
4)
22 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Ary
eete
y et
al.
(201
0)PC
A; P
artic
ipat
ory
Wea
lth R
anki
ng,
Con
sum
ptio
n
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
oG
hana
b, f
Com
pare
d to
con
sum
ptio
n, b
oth
PCA
and
par
ticip
ator
y w
ealth
rank
ing
had
high
incl
usio
n an
d ex
clus
ion
erro
rs
in G
hana
usi
ng c
onsu
mpt
ion
as th
e go
ld st
anda
rdIn
the
urba
n se
tting
, PW
R e
xclu
des f
ewer
poo
r hou
se-
hold
s (50
%) t
han
PMT
(63%
), bu
t als
o in
clud
es m
ore
non-
poor
hou
seho
lds (
50%
) tha
n PM
T (3
6%).
In th
e ru
ral s
ettin
g, P
WR
exc
lude
s mor
e po
or h
ouse
hold
s (7
3%) t
han
PMT
(53%
), bu
t als
o in
clud
es fe
wer
non
-po
or h
ouse
hold
s (17
%) t
han
PMT
(21%
). In
the
sem
i-ur
ban
setti
ng, P
WR
exc
lude
s few
er p
oor h
ouse
hold
s (3
%) t
han
PMT
(46%
), bu
t als
o in
clud
es m
ore
non-
poor
ho
useh
olds
(60%
) tha
n PM
T (2
7%)
Azz
arri
et a
l. (2
005)
PCA
, Con
sum
ptio
nEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyN
oA
lban
iab,
c, f
PCA
-der
ived
ass
et in
dex
was
mor
e hi
ghly
cor
rela
ted
with
co
nsum
ptio
n th
an a
cou
nt m
easu
re c
onsi
sting
of w
ater
an
d sa
nita
tion
prov
isio
n, a
dequ
ate
hous
ing
prov
isio
n,
less
cro
wde
d dw
ellin
gs, a
nd m
inim
um e
duca
tion
of
hous
ehol
d he
adB
alen
et a
l. (2
010)
PCA
, Inc
ome,
PA
FEp
idem
iolo
gyC
ross
-sec
tiona
l su
rvey
No
Chi
naa,
b, c
, fPC
A a
nd P
AF
perfo
rmed
ver
y si
mila
rly in
two
Chi
nese
vi
llage
s with
bot
h ru
ral a
nd p
eri-u
rban
are
as h
avin
g th
e sa
me
asso
ciat
ion
rela
tivel
y w
eak
asso
ciat
ion
(0.2
7) to
ho
useh
old
inco
me
The
prox
y w
ealth
mod
els e
xpla
ined
a h
ighe
r pro
porti
on
of d
ata
in th
e pe
ri-ur
ban
setti
ng th
an th
e ru
ral s
ettin
g (2
7.8%
vs.
24.3
% fo
r PCA
), w
hich
may
add
stre
ngth
to
the
conc
ern
that
an
asse
t-bas
ed in
dex
is a
mor
e ‘a
ppro
-pr
iate
’ mea
sure
of w
ealth
in u
rban
are
as c
ompa
red
with
ru
ral a
reas
Ben
ini (
2007
)PC
A, P
olyc
horic
PC
AEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyN
oB
angl
ades
hc
Com
paris
on o
f PCA
and
pol
ycho
ric P
CA u
sing
Ban
gla-
desh
i DH
S da
ta c
oncl
udes
that
ord
inar
y PC
A la
cks t
he
abili
ty to
dis
crim
inat
e at
the
low
er e
nd o
f the
spec
trum
du
e to
its u
nder
-em
phas
is o
f com
mon
ass
ets
23Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Blu
men
stock
(201
8)PC
AEc
onom
ics
Mac
hine
lear
n-in
gN
oR
wan
dac,
iA
mac
hine
lear
ning
alg
orith
m c
an b
e us
ed to
roug
hly
appr
oxim
ate
PCA
wea
lth in
dex
usin
g ph
one
usag
e ch
ar-
acte
ristic
s in
Rw
anda
and
Afg
hani
stan
, the
mod
els m
ust
be d
evel
oped
sepa
rate
ly fo
r eac
h co
untry
and
can
not b
e ap
plie
d na
ivel
y ac
ross
bor
ders
Boc
cia
et a
l. (2
013)
PCA
, Con
sum
ptio
nG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Zam
bia
c, e
, f, g
Dro
ppin
g al
l var
iabl
es w
hich
are
mor
e lik
ely
to b
e fo
und
in a
n ur
ban
hous
ehol
d di
d no
t sig
nific
antly
affe
ct th
e ov
eral
l var
ianc
e ex
plai
ned
by th
e PC
A m
odel
In a
dditi
on, a
smal
ler Z
ambi
an st
udy
(N =
318)
foun
d th
at
a re
gres
sion
of c
onsu
mpt
ion
mea
sure
s had
onl
y m
ild
agre
emen
t with
PCA
-bas
ed a
sset
inde
x, w
ith 4
6% o
f ho
useh
olds
cla
ssifi
ed in
the
sam
e te
rcile
An
alte
rnat
ive
PCA
-der
ived
inde
x w
ithou
t any
food
-re
late
d va
riabl
es (w
hich
may
hav
e aff
ecte
d tu
berc
ulos
is
outc
omes
of i
nter
est d
irect
ly) w
as n
ot si
gnifi
cant
ly
diffe
rent
than
the
PCA
-der
ived
inde
x us
ing
all v
aria
bles
(3
2.1%
vs.
34.5
%)
A re
gres
sion
of c
onsu
mpt
ion
mea
sure
s had
onl
y m
ild
agre
emen
t with
PCA
-bas
ed a
sset
inde
x, w
ith 4
6% o
f ho
useh
olds
cla
ssifi
ed in
the
sam
e te
rcile
24 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Bol
len
et a
l. (2
002)
PCA
, Sim
ple
Sum
In
dex,
Con
sum
p-tio
n
Dem
ogra
phy
Cro
ss-s
ectio
nal
surv
eyN
oG
hana
; Per
ub,
c, e
, gIn
man
y lo
w a
nd m
iddl
e in
com
e co
untri
es (L
MIC
), in
com
e ca
n be
hig
hly
varia
ble
or d
ifficu
lt to
acc
urat
ely
mea
sure
and
ass
et in
dice
s can
rely
on
asse
ts w
hich
hav
e di
rect
impa
ct o
n he
alth
, suc
h as
wat
er a
nd sa
nita
tion
qual
ity o
r foo
d av
aila
bilit
yA
n al
tern
ativ
e PC
A-d
eriv
ed in
dex
with
out a
ny fo
od-
rela
ted
varia
bles
(whi
ch m
ay h
ave
affec
ted
tube
rcul
osis
ou
tcom
es o
f int
eres
t dire
ctly
) was
bui
lt an
d fo
und
no
sign
ifica
nt d
iffer
ence
with
the
PCA
-der
ived
inde
x us
ing
all v
aria
bles
(32.
1% v
s. 34
.5%
)A
PCA
-der
ived
ass
et in
dex
usin
g D
HS
data
per
form
ed
bette
r tha
n al
l oth
er m
easu
res,
incl
udin
g co
nsum
ptio
n m
easu
res (
whi
ch p
redi
cted
alm
ost n
o va
riatio
n in
ferti
l-ity
) bas
ed o
n B
ICTh
e si
mpl
e co
unt m
easu
res t
o ha
s the
seco
nd-b
est fi
t (a
fter P
CA-d
eriv
ed a
sset
inde
x) a
ccor
ding
to B
ayes
ian
Info
rmat
ion
Crit
erio
n
25Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Boo
ysen
et a
l. (2
008)
MCA
, Con
sum
p-tio
nEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyN
oG
hana
; Ken
ya; M
ali;
Sene
gal;
Tanz
ania
; Za
mbi
a; Z
imba
bwe
b, c
, fU
rban
hou
seho
lds a
re m
ore
likel
y to
ow
n m
any
asse
ts an
d ar
e m
ore
likel
y to
ben
efit f
rom
pub
licly
pro
vide
d as
sets
such
as p
iped
wat
er, s
o th
ey w
ill b
e in
appr
opria
tely
cla
s-sifi
ed a
s wea
lthie
r tha
n co
mpa
rabl
e ru
ral h
ouse
hold
sO
pted
to u
se M
CA ra
ther
than
PCA
bec
ause
of s
tatis
tical
lim
itatio
ns, b
ut fo
und
that
des
pite
som
e di
ffere
nces
in
varia
ble
wei
ght o
rder
s, th
ere
was
no
sign
ifica
nt d
if-fe
renc
e be
twee
n bo
th in
dice
s (r =
0.95
3, p
< 0.
01) a
nd
the
few
hou
seho
lds t
hat w
ere
clas
sifie
d in
to d
iffer
ent
quin
tiles
wer
e re
stric
ted
to o
ne le
vel h
ighe
r or l
ower
The
gene
ral t
rend
of p
ublic
ly-p
rovi
ded
serv
ices
tend
ing
to b
e of
mor
e im
porta
nce
in th
e lo
wer
end
of t
he
soci
oeco
nom
ic g
radi
ent a
nd p
rivat
e go
ods t
endi
ng to
be
mor
e im
porta
nt in
the
uppe
r end
whe
n us
ing
PCA
may
ha
ve so
me
impa
ct o
n he
alth
ineq
ualit
ies
Ass
et in
dice
s are
slow
to c
hang
e ov
er ti
me—
even
whe
n sig
nific
ant h
ouse
hold
impr
ovem
ents
can
be m
easu
red
usin
g co
nsum
ptio
nC
hakr
abor
ty e
t al.
(201
6)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Ban
glad
esh;
Ben
in;
Cam
bodi
a; C
am-
eroo
n; E
thio
pia;
M
alaw
i; M
ozam
-bi
que;
Nep
al;
Nig
eria
; Pak
istan
; Ph
ilipp
ines
; Rw
anda
; Se
nega
l;Tan
zani
a;
Uga
nda;
Zim
babw
e
f, g,
iSu
rvey
resp
onde
nts m
ay h
ave
diffi
culty
ans
wer
ing
ques
-tio
ns a
bout
the
num
ber o
f hec
tare
s of a
gric
ultu
ral l
and
owne
d, w
hich
type
of t
oile
t con
struc
tion
is us
ed in
thei
r ho
me,
or e
ven
whe
ther
they
live
in a
n ur
ban
or ru
ral a
rea
An
atte
mpt
to c
reat
e a
sim
plifi
ed a
sset
list
for P
CA-
deriv
ed in
dex
usin
g da
ta fr
om 1
6 D
HS
surv
yes f
ound
in
ter-q
uint
ile a
gree
men
t ran
ging
from
75%
to 8
3% b
y dr
oppi
ng v
ario
us c
ount
ry-s
peci
fic, u
rban
–rur
al sp
ecifi
c,
and
agric
ultu
ral q
uesti
ons
26 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Cha
sekw
a et
al.
(201
8)PC
A, P
olyc
horic
PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Zim
babw
ec,
fA
Zim
babw
ean
surv
ey fi
nds a
Spe
arm
an ra
nk c
oeffi
cien
t of
0.9
10 (9
5% C
I: 0.
904–
0.91
5) a
nd 9
4% a
gree
men
t be
twee
n w
ealth
qui
ntile
s bet
wee
n PC
A a
nd P
olyc
horic
PC
ASe
para
tely
, a ru
ral-o
nly
sam
ple
usin
g on
ly P
CA a
chie
ved
Spea
rman
rank
cor
rela
tion
coeffi
cien
t of 0
.862
[95%
CI
0.85
4–0.
869]
with
the
stan
dard
DH
S w
ealth
inde
xC
hum
a an
d M
olyn
eux
(200
9)PC
A, C
onsu
mpt
ion
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
oK
enya
b, e
, fIn
vesti
gatio
n of
inse
ctic
ide-
treat
ed n
et o
wne
rshi
p in
K
enya
foun
d a
mix
ed p
ictu
re o
f lar
ger i
nequ
aliti
es in
ur
ban
area
s usin
g a
PCA
-der
ived
ass
et in
dex
com
pare
d to
expe
nditu
re-c
lass
ifica
tion,
but
smal
ler i
nequ
aliti
es
in ru
ral a
reas
. The
y sp
ecul
ate
that
this
may
be
due
to
free
net d
istrib
utio
ns in
rura
l are
as, a
nd c
oncl
ude
that
ne
ither
the
asse
t ind
ex o
r con
sum
ptio
n in
dex
appr
oach
is
supe
rior f
or h
ealth
rese
arch
in L
MIC
A tr
aditi
onal
PCA
-der
ived
inde
x pl
aced
no
rura
l hou
seho
lds
in th
e ric
hest
quin
tile
and
only
one
rura
l hou
seho
ld in
the
seco
nd ri
ches
t. Bo
th in
dice
s bor
e lit
tle re
sem
blan
ce to
the
cons
umpt
ion
data
, with
onl
y 30
.5%
of r
ural
expe
nditu
re-
poor
hou
seho
lds a
lso c
lass
ified
ass
et p
oor a
nd 4
3.4%
of
urba
n ex
pend
iture
-poo
r hou
seho
lds a
lso b
eing
cla
ssifi
ed
as a
sset
-poo
rW
hen
cons
truct
ing
sepa
rate
indi
ces f
or u
rban
and
rura
l ar
eas i
n th
e sa
me
coun
try, a
sset
s lik
e ch
icke
ns o
r bic
ycle
s ca
n be
an
indi
cato
r of r
elat
ive
wea
lth in
rura
l are
as, w
hile
be
ing
an in
dica
tor o
f rel
ativ
e po
verty
in u
rban
are
as
27Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Cór
dova
(200
8)PC
A, I
ncom
eEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyN
oA
rgen
tina;
Bol
ivia
; B
razi
l; C
hile
; Col
om-
bia;
Cos
ta R
ica;
D
omin
ican
Rep
ublic
; Ec
uado
r; El
Sal
-va
dor;
Gua
tem
ala;
H
aiti;
Hon
dura
s;
Jam
aica
; Mex
ico;
N
icar
agua
; Pan
ama;
Pe
ru; T
rinid
ad a
nd
Toba
go; U
rugu
ay
a, d
, iC
onstr
ucte
d a
PCA
inde
x ap
plic
able
to 2
1 La
tin A
mer
i-ca
n an
d C
arib
bean
cou
ntrie
s usi
ng te
leph
one
surv
ey
data
foun
d ge
nera
lly e
ncou
ragi
ng re
sults
. Poo
led
prin
-ci
pal c
ompo
nent
s wer
e br
oadl
y ap
plic
able
from
poo
rer
coun
tries
like
Per
u to
rich
er c
ount
ries l
ike
Cos
ta R
ica,
an
d th
e re
sulti
ng re
lativ
e w
ealth
qui
ntile
s wer
e str
ongl
y co
rrel
ated
with
yea
rs o
f sch
oolin
g an
d se
lf-re
porte
d in
com
e sc
ale
Dok
u et
al.
(201
0)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Gha
nac,
dTh
e as
soci
atio
n of
PCA
-der
ived
“m
ater
ial a
fflue
nce”
of
adol
esce
nts w
as o
nly
mod
estly
cor
rela
ted
to p
aren
tal
educ
atio
n le
vels
(mat
erna
l r =
0.32
, pat
erna
l r =
0.36
), ex
plai
ning
onl
y 14
% o
f par
enta
l edu
catio
n an
d oc
cupa
-tio
n va
rianc
eEr
go e
t al.
(201
6)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Hon
dura
s; S
eneg
alc,
g, i
Com
parin
g tw
o si
mpl
ified
ass
et in
dice
s to
the
gold
st
anda
rd o
f the
full
DH
S in
dex
in H
ondu
ras,
ther
e w
as
alm
ost p
erfe
ct a
gree
men
t (ka
ppa
valu
e gr
eate
r tha
n 0.
61) a
fter r
educ
ing
the
num
ber o
f var
iabl
es fr
om 1
11
to 2
4 va
riabl
es u
sing
an
itera
tive
rank
ing
of fa
ctor
load
-in
gs, a
nd fr
om 1
11 to
21
varia
bles
in S
eneg
al. H
owev
er,
the
varia
bles
that
mat
tere
d m
ost v
arie
d ac
cord
ing
to th
e co
untry
con
text
, and
sing
le v
aria
bles
may
still
requ
ire
thei
r ow
n qu
estio
ns
28 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Ferg
uson
et a
l. (2
003)
PCA
, Reg
ress
ion,
In
com
e, C
on-
sum
ptio
n
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyYe
sG
reec
e; P
eru;
Pak
istan
a, b
, c, i
A re
gres
sion
-bas
ed a
ppro
ach
used
a v
aria
nt o
f a h
iera
rchi
-ca
l ord
ered
pro
bit m
odel
(DIH
OPI
T) to
link
hou
seho
ld
asse
ts to
bot
h in
com
e an
d ex
pend
iture
with
var
ying
su
cces
s. In
Gre
ece
and
Peru
, whi
ch a
chie
ved
low
leve
ls
of m
easu
rem
ent e
rror
, the
regr
essi
ons a
ll re
ache
d 0.
60
Spea
rman
’s rh
o or
gre
ater
with
bot
h ho
useh
old
inco
me
and
cons
umpt
ion;
and
impo
rtant
ly, f
ound
resu
lts n
early
id
entic
al to
a st
anda
rd P
CA in
dex
Even
in P
akist
an, w
hich
foun
d ve
ry lo
w c
orre
latio
n to
in
com
e an
d ex
pend
iture
due
to m
easu
rem
ent e
rror
, the
PC
A a
nd D
IHO
PIT
mod
els g
ave
near
ly id
entic
al re
sults
Film
er a
nd P
ritch
ett
(200
1)PC
A, C
onsu
mpt
ion
Dem
ogra
phy
Cro
ss-s
ectio
nal
surv
eyYe
sIn
dia
b, c
, d, f
The
PCA
app
roac
h pr
ovid
ed a
way
to sy
stem
atic
ally
re
duce
the
varia
bles
of i
nter
est t
o a
min
imum
whi
le
redu
cing
relia
nce
on th
e ju
dgm
ent o
f an
anal
yst.
Ther
e is
a h
igh
Spea
rman
rank
cor
rela
tion
of “
asse
t pov
erty
” in
Indi
an st
ates
with
nat
iona
l pov
erty
stat
istic
s of 0
.794
(p
< 0.
001,
N =
16) a
nd a
ccep
tabl
e pe
rform
ance
with
da
ta fr
om N
epal
, Ind
ones
ia, a
nd P
akist
an; a
llow
ing
the
auth
ors t
o co
nclu
de th
at “
Prin
cipa
l-com
pone
nts
anal
ysis
pro
vide
s pla
usib
le a
nd d
efen
sibl
e w
eigh
ts fo
r an
inde
x of
ass
ets t
o se
rve
as a
pro
xy fo
r wea
lth”
Ass
et in
dice
s are
gen
eral
ly v
iew
ed a
s mea
sure
s of l
ong-
term
wea
lth o
r SES
, but
not
of s
hort-
term
pov
erty
, in
com
e, o
r con
sum
ptio
nTh
ere
are
conc
erns
ove
r com
para
bilit
y of
resu
lts b
etw
een
urba
n an
d ru
ral a
reas
29Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Film
er a
nd S
cott
(201
2)PC
A, C
onsu
mpt
ion,
IR
T, C
ount
inde
xD
emog
raph
yC
ross
-sec
tiona
l su
rvey
Yes
Alb
ania
; Bra
zil;
Gha
na;
Nep
al; N
icar
agua
; Pa
nam
a; P
apua
N
ew G
uine
a; S
outh
A
fric
a; U
gand
a; V
iet-
nam
; Zam
bia
b, c
, d, e
, f,
iSi
nce
asse
t ind
ices
are
bas
ed o
n ho
useh
old
asse
ts a
nd
cann
ot b
e di
vide
d on
a p
er-c
apita
bas
is, i
nter
pret
atio
n of
resu
lts m
ust b
ear t
his i
n m
ind.
Thi
s mea
ns th
at a
sset
in
dice
s are
mor
e cl
osel
y re
late
d to
hou
seho
ld e
cono
-m
ies o
f sca
le m
odel
s tha
n pu
re p
er-c
apita
con
sum
ptio
n m
odel
s, re
info
rcin
g th
e id
ea th
at th
e in
dice
s are
trac
k-in
g a
sepa
rate
, but
equ
ally
val
id c
onst
ruct
ion
of S
ESA
sset
indi
ces a
nd c
onsu
mpt
ion
mod
els a
re m
ore
clos
ely
rela
ted
whe
n a
high
er p
erce
ntag
e of
con
sum
ptio
n is
ca
ptur
ed b
y as
sets
incl
uded
in th
e ho
useh
old
surv
eys,
and
they
are
mor
e hi
ghly
cor
rela
ted
in c
ount
ries w
here
th
e av
erag
e sh
are
of n
on-fo
od e
xpen
ditu
res i
s hig
hPC
A a
nd c
onsu
mpt
ion
indi
ces g
ener
ate
alm
ost i
dent
ical
re
sults
for i
nequ
aliti
es in
car
e se
ekin
g be
havi
or. T
here
w
as g
reat
er h
ealth
-see
king
beh
avio
r fou
nd a
mon
g th
e re
lativ
ely
poor
, alth
ough
this
is h
ypot
hesi
zed
to b
e a
prod
uct o
f the
poo
rest
qui
ntile
’s d
ispr
opor
tiona
te sh
are
of il
lnes
sTh
e hi
ghes
t lev
els o
f chi
ld m
orta
lity
are
not u
nifo
rmly
fo
und
in th
e po
ores
t qui
ntile
s of t
he e
xpen
ditu
re
mod
el, b
ut a
re a
lway
s fou
nd in
the
poor
est q
uint
iles o
f PC
A m
odel
s—a
stat
istic
ally
sign
ifica
nt d
iffer
ence
-in-
diffe
renc
eSt
atis
tical
ly si
gnifi
cant
edu
catio
nal i
nequ
ality
in 7
of 1
1 co
untri
es in
clud
ed, w
ith th
e PC
A a
ppro
ach
mos
t ofte
n re
sulti
ng in
larg
er in
equa
litie
sTh
e di
ffere
nce
in c
lass
ifica
tion
of u
rban
izat
ion
betw
een
the
poor
est a
nd ri
ches
t qui
ntile
s can
be
as la
rge
as 7
5%
in a
PCA
mod
el a
nd 2
2% in
an
expe
nditu
re m
odel
in
Alb
ania
, with
seve
ral o
ther
cou
ntrie
s als
o ha
ving
larg
e di
scre
panc
ies i
n SE
S ra
nkin
g du
e to
urb
an st
atus
Ran
k co
rrel
atio
ns fo
r PCA
and
IRT
betw
een
0.95
and
1.
00—
the
mos
t hig
hly
corr
elat
ed a
ltern
ativ
e m
easu
re
to P
CA
30 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Fink
et a
l. (2
017)
PCA
; Pre
dict
ed
Inco
me
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
o88
Low
and
Mid
dle
Inco
me
coun
tries
a, e
, h, i
Mor
e va
riatio
n in
stun
ting
prev
alen
ce w
as c
aptu
red
usin
g th
e H
arttg
en &
Vol
mer
pre
dict
ed in
com
e ap
proa
ch
(38%
) com
pare
d to
wea
lth q
uint
iles (
20%
)It
is p
ossi
ble
to c
ompa
re h
ealth
out
com
es ta
king
inco
me
(or w
ealth
) ine
qual
ity in
to a
ccou
nt u
sing
tool
s suc
h as
eq
uipl
ots w
ith th
is a
ppro
ach,
reve
alin
g co
untri
es w
hich
ha
ve si
mila
r out
com
es a
t any
giv
en in
com
e le
vel,
and
othe
rs th
at a
re p
erfo
rmin
g po
orly
at a
giv
en in
com
e le
vel
Har
ttgen
and
Vol
lmer
(2
013)
PCA
; Pre
dict
ed
Inco
me
Econ
omic
sC
ross
-sec
tiona
l su
rvey
No
Bol
ivia
; Ind
ones
ia,
Zam
bia
a, f,
h, i
It sh
ould
be
poss
ible
to u
se a
PCA
-der
ived
ass
et in
dex
to
sim
ulat
e ho
useh
old
inco
me
distr
ibut
ion
whe
re n
atio
nal
inco
me
is lo
g-no
rmal
ly d
istrib
uted
and
hou
seho
ld ra
nks
are
the
sam
e fo
r bot
h th
e as
set i
ndex
and
inco
me
Pool
ing
asse
ts o
ver t
ime
wou
ld th
eore
tical
ly a
llow
ab
solu
te d
iffer
ence
s in
wea
lth to
be
com
pare
d ov
er
time,
but
this
is c
ompl
icat
ed b
y th
e fa
ct th
at th
e so
cial
si
gnifi
canc
e of
cer
tain
ass
ets s
uch
as la
ndlin
es, r
adio
s, an
d bi
cicl
es c
hang
es si
gnifi
cant
ly o
ver t
ime
Hom
enau
th e
t al.
(201
7)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Tanz
ania
c, f,
gTh
ree
alte
rnat
ive
wea
lth in
dice
s wer
e co
mpa
red
to a
st
anda
rd D
HS
wea
lth in
dex
in th
eir a
bilit
y to
pre
dict
ve
ctor
-bor
ne d
isea
se ri
sk. A
utho
rs su
gges
t tha
t ind
ices
th
at c
onta
in d
urab
le a
sset
s bes
t pre
dict
SES
-rel
ated
ris
k, b
ut th
ere
are
clea
rly e
rror
s in
cons
truct
ing
the
refe
renc
e D
HS-
style
wea
lth in
dex
31Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Hou
wel
ing
et a
l. (2
003)
PCA
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
oB
oliv
ia; B
razi
l; In
done
-si
a; C
amer
oon;
Cha
d;
Ken
ya; M
alaw
i; Pa
kist
an; T
anza
nia;
U
gand
a
c, e
, gC
ompa
rison
of a
Wor
ld B
ank
PCA
-der
ived
wea
lth in
dex
usin
g D
HS
data
to th
ree
alte
rnat
ive
indi
ces t
hat e
xclu
de
dire
ct d
eter
min
ants
of h
ealth
and
fact
ors p
rovi
ded
at
the
com
mun
ity-le
vel,
in w
hich
18%
of h
ouse
hold
s wer
e ca
tego
rized
in a
diff
eren
t wea
lth q
uint
ile, w
ith m
ost o
f th
ese
shift
ing
to a
n ad
jace
nt q
uint
ile60
% c
hang
e in
the
rela
tive
inde
x of
ineq
ualit
y fo
r five
he
alth
out
com
es w
ith a
ltern
ativ
e PC
A-d
eriv
ed w
ealth
in
dice
s, al
thou
gh th
e di
rect
ion
of c
hang
e ap
pear
ed to
be
rand
om a
nd w
as n
ot si
gnifi
cant
in so
me
coun
tries
How
e et
al.
(200
9)PC
A, C
onsu
mpt
ion
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
oA
lban
ia; B
razi
l; C
ôte
d’Iv
oire
; G
hana
; Gua
tem
ala;
In
done
sia;
Jam
aica
; M
adag
asca
r; M
alaw
i; M
exic
o;
Moz
ambi
que;
Nep
al;
Nic
arag
ua; P
akist
an;
Pana
ma;
Pap
ua N
ew
Gui
nea;
Per
u; S
outh
A
fric
a; T
anza
nia;
U
gand
a; V
ietn
am;
Zam
bia
b, i
SES
is a
t lea
st pa
rtial
ly d
epen
dent
upo
n lo
ng-te
rm e
arn-
ings
, sha
red
asse
ts, c
onsu
mpt
ion,
and
con
sum
ptio
n sm
ooth
ing.
The
re is
evi
denc
e th
at c
onsu
mpt
ion
data
tra
cks a
sset
indi
ces m
ore
clos
ely
in m
iddl
e in
com
e co
untri
es, a
nd e
spec
ially
if a
gre
ater
var
iety
of a
sset
s ar
e in
clud
edW
eak
to m
oder
ate
asso
ciat
ion
betw
een
asse
t ind
ices
an
d ex
pend
iture
dat
a, a
lthou
gh th
ese
findi
ngs m
ay b
e di
sput
ed d
ue to
the
rela
tivel
y sm
all s
ampl
e si
ze (1
7 stu
dies
), in
clus
ion
of a
ll in
dex
cons
truct
ion
met
hods
, an
d ar
bitra
ry c
ut o
ffs o
f effe
ctiv
enes
s (as
ack
now
ledg
ed
by th
e au
thor
s the
mse
lves
)
32 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
How
e et
al.
(201
2)PC
A; C
onsu
mpt
ion,
In
com
e; P
ar-
ticip
ator
y w
ealth
ra
nkin
g
Glo
bal H
ealth
Met
hods
No
N/A
a, b
, c, d
, e,
f, g
The
adve
nt o
f PCA
can
be
trace
d to
the
chal
leng
e of
de
velo
ping
a m
etho
d of
con
verti
ng a
serie
s of o
wne
r-sh
ip v
aria
bles
, man
y of
whi
ch w
ere
bina
ry (y
es/n
o) o
r ca
tego
rical
(roo
f mat
eria
l, e.
g.),
into
a c
ontin
uous
SES
gr
adie
nt. I
nitia
l app
roac
hes m
ostly
relie
d on
sim
ple
sum
s of a
sset
ow
ners
hip
such
as h
ousi
ng q
ualit
y, d
ura-
ble
asse
t ow
ners
hip,
or p
ublic
util
ity a
cces
s. H
owev
er,
this
impl
icitl
y ga
ve e
qual
wei
ght t
o al
l ass
ets,
whe
ther
it
was
a re
lativ
ely
rare
maj
or e
xpen
se su
ch a
s a c
ar, o
r a
near
ly u
biqu
itous
com
mod
ity su
ch a
s a ra
dio
Ass
et in
dice
s are
gen
eral
ly v
iew
ed a
s mea
sure
s of l
ong-
term
wea
lth o
r SES
, but
not
of s
hort-
term
pov
erty
, in
com
e, o
r con
sum
ptio
nA
sset
indi
ces a
re fr
eque
ntly
com
pare
d to
con
sum
ptio
n da
ta b
y re
sear
cher
s tha
t arg
ue th
at it
is th
e m
ost a
cces
-si
ble
and
clos
ely
rela
ted
com
para
tor w
ith w
hich
to
mea
sure
the
perfo
rman
ce o
f PCA
Ther
e is
a fu
ndam
enta
l flaw
in th
e ap
plic
atio
n of
ord
inar
y PC
A to
the
type
s of v
aria
bles
nee
ded
to c
onstr
uct a
n as
set i
ndex
—th
e te
chni
que
is n
ot m
eant
to b
e ap
plie
d to
bi
nary
and
cat
egor
ical
var
iabl
esJe
an e
t al.
(201
6)PC
A, C
onsu
mpt
ion
Econ
omic
sM
achi
ne le
arn-
ing
No
Mal
awi;
Nig
eria
; R
wan
da;T
anza
nia;
U
gand
a
b, f,
iA
con
volu
tiona
l neu
ral n
etw
ork
train
ed o
n gr
ound
im
ager
y is
abl
e to
pre
dict
37–
55%
of v
aria
tion
in
cons
umpt
ion
and
55–7
5% o
f var
iatio
n in
ass
et w
ealth
if
train
ed se
para
tely
for e
ach
coun
try, a
lthou
gh th
is d
rops
to
19–
52%
and
24–
71%
if a
pplie
d to
oth
er c
ount
ries.
This
is sl
ight
ly m
ore
pred
ictiv
e th
an p
hone
use
esti
ma-
tion
mod
els,
and
the
diffe
renc
e in
esti
mat
ion
is li
kely
du
e to
the
area
’s w
ealth
itse
lf ra
ther
than
dire
ctly
id
entif
ying
hou
seho
ld fe
atur
es d
irect
ly
33Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
John
ston
and
Abr
eu
(201
6)PC
AEc
onom
ics
Met
hods
No
All
Afr
ican
cou
ntrie
sa,
b, d
, e,
f, i
Ther
e is
a v
illag
e in
Gui
nea-
Bis
sau
whe
re p
orta
ble
gas
stove
s are
a n
orm
al g
ood,
but
bec
ause
that
vill
age
is
rela
tivel
y po
or c
ompa
red
to o
ther
vill
ages
, the
fact
or
load
ing
is n
egat
ive
A h
eate
d de
bate
ove
r whe
ther
an
Afr
ican
gro
wth
“m
ira-
cle”
occ
urre
d w
as sp
arke
d du
e to
com
paris
ons o
f ass
et
wea
lth w
ith n
atio
nal a
ccou
nts.
Fact
ors s
uch
as n
ew
chea
p im
ports
of h
ouse
hold
dur
able
s fro
m A
sia
and
the
tend
ency
of h
oush
old
asse
t pric
es to
dro
p ov
er ti
me
wer
e no
t ful
ly a
ccou
nted
for,
lead
ing
to m
any
diss
ent-
ing
opin
ions
and
unc
erta
inty
ove
r whe
ther
wel
fare
had
tru
ly im
prov
edJo
seph
et a
l. (2
018)
PCA
, Pre
dict
ed
Inco
me
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
o10
0 lo
w- a
nd m
iddl
e-in
com
e co
untri
esa,
e, h
, iPr
edic
ted
abso
lute
inco
me
bette
r pre
dict
ed sk
illed
birt
h de
liver
y in
a la
rge
100-
coun
try st
udy,
with
log-
norm
al-
ized
pre
dict
ed in
com
e ex
plai
ning
51.
6% o
f var
iatio
n,
DH
S w
ealth
qui
ntile
s pre
dict
ing
22.0
%, a
nd m
ean
wea
lth q
uint
ile P
CA sc
ore
pred
ictin
g 12
.8%
Com
paris
ons o
f all
coun
tries
ove
r tim
e re
veal
s im
porta
nt
trend
s suc
h as
cou
ntrie
s tha
t hav
e su
ccee
ded
in in
crea
s-in
g sk
illed
birt
h at
tend
ance
in sp
ite o
f sta
lled
inco
me
grow
th, a
nd th
ose
that
hav
e no
t im
prov
ed o
utco
mes
ev
en in
tim
es o
f sus
tain
ed e
cono
mic
gro
wth
34 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Kol
enik
ov a
nd A
ngel
es
(200
9)PC
A, P
olyc
horic
PC
AEc
onom
ics
Cro
ss-s
ectio
nal
surv
ey; S
imu-
latio
n
No
Ban
glad
esh;
sim
ulat
ed
data
c, d
Man
y re
sear
cher
s hav
e po
inte
d ou
t a fu
ndam
enta
l flaw
in
the
appl
icat
ion
of o
rdin
ary
PCA
to th
e ty
pes o
f va
riabl
es n
eede
d to
con
struc
t an
asse
t ind
ex; n
amel
y,
that
the
tech
niqu
e is
not
mea
nt to
be
appl
ied
to b
inar
y an
d ca
tego
rical
var
iabl
es. S
ince
an
aver
age
of 6
0%
of h
ouse
hold
surv
ey q
uesti
ons u
sed
to c
onstr
uct a
sset
in
dice
s are
bin
ary,
this
is n
o m
inor
lim
itatio
nFi
lmer
and
Prit
chet
t’s te
chni
que
crea
tes s
purio
us c
or-
rela
tion
thro
ugh
the
intro
duct
ion
of d
umm
y va
riabl
es
and
for l
osin
g di
rect
iona
lity
of o
rdin
al d
ata,
Kol
enik
ov
and
Ang
eles
pro
pose
the
use
of a
slig
htly
am
ende
d m
ultiv
aria
te te
chni
que,
orig
inal
ly d
eriv
ed b
y th
e sa
me
stat
istic
ian
as o
rdin
ary
PCA
. Not
onl
y is
ther
e gr
eate
r st
atist
ical
fide
lity,
but
the
stat
us o
f not
ow
ning
an
asse
t is
als
o ta
ken
into
acc
ount
The
key
findi
ngs o
f thi
s pro
of-o
f-co
ncep
t stu
dy w
ere
that
po
lych
oric
PCA
dem
onstr
ated
low
er m
iscl
assi
ficat
ion
rate
s, ex
plai
ned
a hi
gher
pro
porti
on o
f var
ianc
e, is
m
ore
robu
st to
the
num
ber o
f cat
egor
ies u
sed,
and
is
mor
e ro
bust
to c
hang
es in
var
iabl
e co
ding
sche
me
than
th
e Fi
lmer
and
Prit
chet
t pro
cedu
re T
he tw
o m
etho
ds
dem
onstr
ate
dive
rgen
t cla
ssifi
catio
ns a
t the
low
er e
nd
of th
e SE
S sp
ectru
m w
ith in
crea
sing
agr
eem
ent o
f cla
s-si
ficat
ion
on th
e up
per e
nd o
f the
SES
spec
trum
Lind
elow
(200
6)PC
A, C
onsu
mpt
ion
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyYe
sM
ozam
biqu
eb,
eIn
gen
eral
, the
re is
som
e ev
iden
ce th
at a
sset
mea
sure
s m
ay in
crea
se th
e si
gnifi
canc
e of
hea
lth in
equa
litie
s. Pr
o-ric
h in
equa
litie
s in
imm
uniz
atio
ns, m
ater
nity
ca
re, i
nstit
utio
nal d
eliv
erie
s, an
d ho
spita
l vis
its w
ere
grea
ter w
hen
mea
sure
d w
ith a
PCA
-der
ived
inde
x th
an
cons
umpt
ion
data
in M
ozam
biqu
e
35Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Man
thal
u et
al.
(201
0)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Mal
awi
e, f
A M
alaw
ian
team
wen
t fro
m h
ealth
to w
ealth
, and
foun
d al
mos
t no
diffe
renc
e be
twee
n us
ing
a PC
A in
dex
or a
n in
dex
com
pose
d of
the
perc
enta
ge o
f chi
ldre
n w
ith st
untin
g pe
r dist
rict w
ith g
over
nmen
t allo
catio
ns
(spe
arm
an ra
nk c
orre
latio
n = 0.
96)
McK
enzi
e (2
005)
PCA
Econ
omic
sC
ross
-sec
tiona
l su
rvey
Yes
Mex
ico
a, d
, f, h
One
of t
he e
arlie
st re
sear
cher
s to
inve
stiga
te th
e us
eful
-ne
ss o
f an
asse
t ind
ex in
ineq
ualit
y re
sear
ch c
oncl
uded
th
at P
CA-b
ased
mea
sure
s in
com
bina
tion
with
bo
otstr
ap p
redi
ctio
n m
etho
ds w
ere
high
ly c
orre
late
d to
ex
pend
iture
-bas
ed in
equa
lity
mea
sure
s in
Mex
ico;
and
im
porta
ntly
, tha
t an
inde
x ba
sed
on d
urab
les i
s mor
e hi
ghly
cor
rela
ted
with
thes
e m
easu
res t
han
indi
ces
base
d on
hou
sing
, util
ities
, foo
d ex
pend
iture
, or a
co
mbi
natio
n of
thes
e va
riabl
esM
iche
lson
(201
3)PC
A, M
CA,
Inco
me
Econ
omic
sC
ross
-sec
tiona
l su
rvey
No
Mal
awi;
Tanz
ania
; M
ali;
Gha
naa,
c, f
, gA
n ex
pand
ed se
t of a
sset
s col
lect
ed to
mea
sure
pro
gres
s in
the
Mill
eniu
m V
illag
es P
roje
ct d
id n
ot p
erfo
rm b
ette
r th
an th
e st
anda
rd D
HS
asse
t mix
MCA
and
PCA
wer
e no
t sig
nific
antly
diff
eren
t in
year
ov
er y
ear c
hang
e an
d w
ere
both
mor
e str
ongl
y au
toco
r-re
late
d th
an h
ouse
hold
inco
me
Tran
sitio
ns o
ut o
f pov
erty
occ
ur a
t the
sam
e ra
te u
sing
as
set i
ndic
es a
nd h
ouse
hold
inco
me,
with
app
roxi
-m
atel
y 12
–20%
of t
he lo
wes
t qua
rtile
hou
seho
lds c
las-
sifie
d as
hig
hest
quar
tile
hous
ehol
ds a
fter t
wo
year
sPC
A in
dex
was
mor
e str
ongl
y co
rrel
ated
with
loca
lly
iden
tified
fact
ors i
ndic
atin
g po
verty
(fem
ale-
head
edne
ss
of h
ouse
hold
, dep
ende
ncy
ratio
s, an
d ho
useh
old
food
in
secu
rity)
than
hou
seho
ld in
com
e
36 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Moh
anty
(200
9)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Indi
ac,
e, f
Ass
et in
dice
s may
be
mor
e di
rect
ly a
ssoc
iate
d w
ith
thes
e ou
tcom
es th
an h
ouse
hold
exp
endi
ture
or i
ncom
e be
caus
e he
alth
and
edu
catio
n ar
e m
ore
repr
esen
tativ
e of
lo
ng-r
un h
ouse
hold
SES
than
mon
etar
y hi
ghs o
r low
sU
nmet
hea
lthca
re n
eed
can
be u
nder
estim
ated
in ru
ral
area
s and
ove
resti
mat
ed in
urb
an a
reas
due
to th
e ur
ban
bias
inhe
rent
in P
CA in
dice
sB
uild
ing
a na
tiona
l ind
ex c
ompa
red
to c
ombi
ning
se
para
te u
rban
and
rura
l PCA
-der
ived
indi
ces i
n In
dia
resu
lted
in 3
9% o
f hou
seho
lds b
eing
cla
ssifi
ed in
the
sam
e qu
intil
e, 5
0% to
hav
e m
oved
to a
n ad
jace
nt q
uin-
tile,
and
10%
to h
ave
mov
ed to
the
farth
est q
uint
ileM
ohse
na e
t al.
(201
0)PC
A, C
ount
m
easu
reG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Ban
glad
esh
c, e
, fU
sing
a b
asic
cou
nt m
easu
re o
utpe
rform
s PCA
in
disc
rimin
atin
g ho
useh
olds
mor
e at
risk
for s
tunt
ing,
w
astin
g, a
nd u
nder
wei
ght.
Thei
r use
of a
sim
ple
coun
t of
radi
o, te
levi
sion
, bic
ycle
, mot
orcy
cle,
tele
phon
e,
and
elec
trici
ty to
con
struc
t wea
lth q
uint
iles r
esul
ted
in
49.1
% o
f hou
seho
lds i
n th
e po
ores
t qui
ntile
and
onl
y 4.
2% o
f the
rich
est q
uint
ile h
avin
g al
l thr
ee o
utco
mes
of
inte
rest,
whi
le th
e PC
A in
dex
prod
uced
equ
ival
ent
perc
enta
ges o
f 28.
6% a
nd 1
1.4%
, res
pect
ivel
yM
oser
and
Fel
ton
(200
7)Po
lych
oric
PCA
Econ
omic
sC
ross
-sec
tiona
l su
rvey
No
Ecua
dor
e, f,
gRe
porti
ng e
rror
s are
kno
wn
to a
ffect
eve
n th
e m
ost c
are-
fully
pla
nned
and
exe
cute
d ho
useh
old
cons
umpt
ion
surv
eys d
ue to
reca
ll er
ror,
excl
usio
n of
som
e ex
pens
es,
choi
ce o
f defl
ator
, and
cur
renc
y ex
chan
ge fl
uctu
atio
nsIn
door
plu
mbi
ng, w
hich
may
onl
y be
mis
sing
from
a
smal
l per
cent
age
of th
e po
ores
t hou
seho
lds o
f a p
opul
a-tio
n ca
n sti
ll be
a v
ery
usef
ul p
redi
ctor
of h
ouse
hold
SE
S
37Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Ngo
and
Chr
istia
ense
n (2
018)
PCA
, DIH
OPI
T,
Con
sum
ptio
nEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyN
oG
hana
, Mal
awi,
Rw
anda
, Tan
zani
a,
and
Uga
nda
b, c
, f, g
, iPC
A c
orre
late
s mor
e str
ongl
y w
ith c
onsu
mpt
ion
than
in
vers
e fr
eque
ncy
wei
ghtin
g or
DIH
OPI
T re
gres
sion
, bu
t the
se d
iffer
ence
s are
not
sign
ifica
ntA
ddin
g a
few
bin
ary
cons
umpt
ion
varia
bles
on
food
and
cl
othi
ng p
urch
ases
incr
ease
d id
entifi
catio
n of
the
poor
in
rura
l set
tings
by
9%, b
ut m
ade
no d
iffer
ence
in u
rban
se
tting
sW
eake
r cor
rela
tion
amon
g hi
ghes
t inc
ome
coun
try a
nd in
ru
ral s
ampl
eN
konk
i et a
l. (2
011)
PCA
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyYe
sSo
uth
Afr
ica
a, e
Non
-sig
nific
ant a
ssoc
iatio
n of
mot
her-t
o-ch
ild H
IV
trans
mis
sion
and
chi
ld m
orta
lity
to a
PCA
-der
ived
ass
et
inde
xN
war
u et
al.
(201
2)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Chi
nac,
eLo
w to
mod
erat
e co
rrel
atio
n w
ith m
ater
nal a
nd c
hild
he
alth
indi
cato
rs a
s an
outc
ome
of in
tere
st; a
lthou
gh
both
an
occu
patio
nal i
ndex
and
edu
catio
nal i
ndex
foun
d eq
ually
, if n
ot c
ontra
dict
ory,
wea
k as
soci
atio
nsO
puni
et a
l. (2
011)
PCA
, Con
sum
ptio
nG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
Yes
Tanz
ania
b, c
, e, i
Cou
ld n
ot id
entif
y a
patte
rn o
f AID
S di
strib
utio
n in
K
enya
whe
ther
PCA
-der
ived
inde
x, h
ouse
hold
con
-su
mpt
ion,
or “
tradi
tiona
l wea
lth”
asse
t ind
ices
wer
e us
edC
ontra
ry to
cla
ims t
hat a
sset
indi
ces d
eriv
ed fr
om st
and-
ardi
zed
hous
ehol
d su
rvey
s can
not b
e re
lied
upon
in
coun
tries
with
trad
ition
al c
once
ptio
ns o
f wea
lth, a
“tra
-di
tiona
l wea
lth in
dex”
con
struc
ted
to re
pres
ent K
enya
n cu
ltura
l con
struc
tions
of w
ealth
was
nea
rly id
entic
al to
st
anda
rd P
CA o
f hou
seho
ld a
sset
s
38 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Reid
path
and
Ahm
adi
(201
4)Po
lych
oric
PCA
, M
SA, C
onsu
mp-
tion
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyYe
sV
ietn
amb,
c, e
, gM
okke
n Sc
ale
Ana
lysi
s is a
non
para
met
ric te
chni
que
whi
ch re
lies o
n G
uttm
an sc
ales
of i
tem
s whi
ch a
re
incr
easi
ngly
“ha
rder
” to
ans
wer
. Usi
ng a
com
bina
-tio
n of
pos
itive
ow
ners
hip
of a
sset
s with
the
diffi
culty
of
elic
iting
a p
ositi
ve re
spon
se, M
SA is
abl
e to
rank
ho
useh
olds
alo
ng a
late
nt S
ES g
radi
ent
Key
ass
umpt
ions
incl
ude
unid
imen
sion
ality
of S
ES, l
ocal
in
depe
nden
ce o
f var
iabl
es, m
onot
onic
ity o
f res
pons
es,
and
inva
riant
item
ord
erin
g. A
n em
piric
al a
pplic
atio
n of
th
e te
chni
que
foun
d ve
ry h
igh
Pear
son
prod
uct m
omen
t co
rrel
atio
n w
ith a
pol
ycho
ric P
CA in
dex
(r =
0.96
) with
a
low
er c
orre
latio
n to
hou
seho
ld e
xpen
ditu
re (r
= 0.
59)
Redu
cing
a 1
7 ite
m a
sset
inde
x to
11
item
s with
pol
y-ch
oric
PCA
resu
lted
in m
uch
low
er c
onco
rdan
ce w
ith
both
exp
endi
ture
(r =
0.57
vs.
r = 0.
41) a
nd a
n M
SA
deriv
ed a
sset
inde
x (r
= 0.
98 v
s. r =
0.68
)Ro
hner
et a
l. (2
012)
PCA
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
oIv
ory
Coa
stc,
eU
sed
rigor
ous b
iom
etric
mea
sure
s of n
utrit
iona
l sta
tus
whi
le a
ccou
ntin
g fo
r the
effe
ct o
f mal
aria
l inf
ectio
n,
age,
and
resi
denc
y. T
he P
CA-d
eriv
ed p
over
ty in
dex
outp
erfo
rmed
the
coun
t sco
re b
y ca
ptur
ing
sign
ifica
nt
soci
oeco
nom
ic in
equa
litie
s in
anem
ia, s
tunt
ing,
and
w
astin
g in
chi
ldre
n an
d w
omen
of r
epro
duct
ive
age
Sahn
and
Stif
el (2
003)
PCA
, Con
sum
ptio
nEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyYe
sC
ote
d’Iv
oire
, Gha
na,
Jam
aica
, Mad
agas
car,
Nep
al, P
akist
an,
Papu
a N
ew G
uine
a,
Peru
, Sou
th A
fric
a,
Vie
tnam
b, e
Con
sum
ptio
n da
ta, s
uch
as th
at m
easu
red
by th
e Li
ving
St
anda
rds a
nd M
easu
rem
ent S
tudi
es, c
an b
e ex
trem
ely
time
cons
umin
g an
d ex
pens
ive
to c
olle
ctM
ulti-
coun
try c
ompa
rativ
e stu
dies
sugg
ests
that
the
use
of P
CA-d
eriv
ed a
sset
indi
ces r
esul
ted
in sm
ooth
er
decl
ines
in st
untin
g by
wea
lth q
uint
iles w
hen
com
pare
d to
pre
dict
ed h
ouse
hold
con
sum
ptio
n in
10
coun
tries
39Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Smits
and
Ste
endi
jk
(201
3)PC
A, M
CA, F
A,
Cat
egor
ical
PCA
Econ
omic
sC
ross
-sec
tiona
l su
rvey
No
97 d
evel
opin
g co
untri
esa,
c, d
, e,
h, i
Can
ove
rcom
e th
e di
fficu
lty o
f inc
ompa
rabi
lity
of m
any
surv
ey it
ems b
y gr
oupi
ng a
cces
sorie
s int
o ch
eap
and
expe
nsiv
e ut
ensi
l cat
egor
ies.
An
asse
t ind
ex a
pplic
able
to
165
hou
seho
ld su
rvey
s acr
oss 9
7 LM
IC is
dev
elop
ed.
Alth
ough
thes
e re
sults
wer
e ob
tain
ed w
ith o
rdin
ary
PCA
, sen
sitiv
ity c
heck
s with
MCA
, fac
tor a
naly
sis,
and
cate
goric
al P
CA d
id n
ot c
hang
e th
e re
sults
. Thi
s int
er-
natio
nal w
ealth
inde
x is
robu
st to
rem
oval
of a
ny re
gion
fro
m a
naly
sis (
Pear
son
corr
elat
ion
coeffi
cien
t ≥ 0.
996)
, re
mov
al o
f any
tim
e pe
riod
(Pea
rson
cor
rela
tion
coef
-fic
ient
≥ 0.
997)
, and
rem
oval
of a
ny o
ne a
sset
(Pea
rson
co
rrel
atio
n co
effici
ent ≥
0.98
6)G
ood
agre
emen
t bet
wee
n th
e in
tern
atio
nal w
ealth
inde
x an
d D
HS
coun
try-s
peci
fic in
dice
s, co
untry
-spe
cific
po
verty
leve
ls, l
ife e
xpec
tanc
y, a
nd m
ost s
trong
ly w
ith
the
Hum
an D
evel
opm
ent I
ndex
. The
aut
hors
ass
ert t
hat
reas
onab
le e
stim
ates
of p
urch
asin
g po
wer
par
ity (P
PP)
pove
rty le
vels
can
be
plac
ed a
t 30t
h pe
rcen
tile
of th
e in
dex
equi
vale
nt to
PPP
$1.2
5 a
day
and
50th
per
cent
ile
at P
PP$2
.00
a da
yTu
sting
et a
l. (2
016)
PCA
, Inc
ome
Glo
bal H
ealth
Pros
pect
ive
coho
rtN
oU
gand
aa,
c, e
, gD
ropp
ing
hous
ehol
d co
nstru
ctio
n va
riabl
es fr
om P
CA
inde
x re
sulte
d in
a si
gnifi
cant
ass
ocia
tion
with
hum
an
bitin
g ra
te b
ecom
ing
insi
gnifi
cant
, eve
n th
ough
the
two
indi
ces a
re h
ighl
y co
rrel
ated
(Spe
arm
an’s
rho =
0.93
)U
car (
2015
)PC
A, I
ncom
e,
Con
sum
ptio
nEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyN
oTu
rkey
a, b
Ass
et in
dex
was
mor
e m
oder
atel
y as
soci
ated
with
con
-su
mpt
ion
and
inco
me,
with
54.
1% b
eing
in th
e lo
wes
t qu
intil
e fo
r bot
h as
set i
ndex
and
con
sum
ptio
n, a
n 47
.1%
in
the
low
est q
uint
ile fo
r bot
h as
set i
ndex
and
inco
me
Van
Leth
et a
l. (2
011)
PCA
, Inc
ome,
C
onsu
mpt
ion
Glo
bal H
ealth
Cro
ss-s
ectio
nal
surv
eyN
oB
angl
ades
h, K
enya
, Ph
ilipp
ines
, Vie
tnam
a, b
, eFo
cuse
d re
view
of t
he u
se o
f PCA
to d
eriv
e as
set i
ndic
es
spec
ifica
lly fo
r tub
ercu
losi
s sur
veys
con
clud
ed th
at th
e m
etho
d m
ore
cons
isten
tly id
entifi
ed in
equi
ties i
n he
alth
th
an in
com
e or
exp
endi
ture
surv
eys
40 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Vand
emoo
rtele
(201
4)PC
A, I
RTEc
onom
ics
Cro
ss-s
ectio
nal
surv
eyN
oM
alaw
ic,
fTh
e ba
sic
idea
of I
RT is
that
obs
erve
d as
sets
(whe
ther
th
ey a
re d
icho
tom
ous,
poly
tom
ous,
nom
inal
, or o
rdin
al)
whi
ch d
emon
strat
e th
e m
ost d
iscr
imin
atio
n ac
cord
ing
to a
late
nt tr
ait (
SES)
are
giv
en la
rger
wei
ghts
, whi
ch
are
then
ass
esse
d fo
r rel
iabi
lity
with
a n
on-p
aram
etric
bo
otstr
apTh
e tw
o ke
y as
sum
ptio
ns o
f nor
mal
dist
ribut
ion
of d
ata
and
inde
pend
ence
of v
aria
bles
offe
r no
impr
ovem
ent t
o th
e ex
istin
g PC
A a
ppro
ach,
and
is a
ckno
wle
dged
to b
e m
ore
time
cons
umin
gH
igh
corr
elat
ion
with
a P
CA in
dex
(Spe
arm
an ra
nk c
or-
rela
tion =
0.88
)V
u et
al.
(201
1)PC
AG
loba
l Hea
lthC
ross
-sec
tiona
l su
rvey
No
Vie
tnam
d, e
, fPC
A-d
eriv
ed a
sset
qui
ntile
s are
cor
rela
ted
with
low
bi
rthw
eigh
t, ed
ucat
ion
leve
l, an
d oc
cupa
tion
in th
e V
ietn
ames
e co
ntex
tV
yas a
nd K
umar
anay
ake
(200
6)PC
AG
loba
l Hea
lthM
etho
dsN
oB
razi
l, Et
hiop
iaa,
b, c
, f, g
The
supp
osed
urb
an b
ias o
f PCA
is n
ot m
iscl
assi
ficat
ion,
bu
t an
accu
rate
repr
esen
tatio
n of
the
rela
tive
afflue
nce
of u
rban
hou
seho
lds
41Approaches and Alternatives to the Wealth Index to Measure…
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
War
d (2
014)
Poly
chor
ic P
CA,
Inco
me
Econ
omic
sC
ross
-sec
tiona
l su
rvey
Yes
Chi
naa,
f, g
, h, i
A la
rge
mul
tiyea
r poo
led
anal
ysis
in C
hina
. Fou
nd n
ega-
tive
fact
or w
eigh
ts fo
r all
agric
ultu
ral a
sset
s, le
adin
g th
em to
sugg
est t
hat s
econ
dary
prin
cipa
l com
pone
nts
wei
ghts
(whi
ch w
ere
posi
tive)
cou
ld b
e us
ed in
thes
e ca
ses.
Alte
rnat
ivel
y, u
rban
and
rura
l hou
seho
lds w
ere
disa
ggre
gate
d an
d se
para
te in
dice
s wer
e co
nstru
cted
, w
hich
led
to a
gric
ultu
ral a
sset
s hav
ing
posi
tive
wei
ghts
fo
r rur
al h
ouse
hold
s and
neg
ativ
e w
eigh
ts fo
r urb
an
hous
ehol
dsEv
en re
mov
al o
f all
agric
ultu
ral a
sset
s has
bee
n sh
own
to
have
littl
e eff
ect o
n ho
useh
old
rank
ings
in C
hina
, with
90
per
cent
of h
ouse
hold
s sta
ying
in th
e sa
me
wea
lth
quin
tile
Usi
ng v
aria
nce
of P
CA to
trac
k in
equa
lity
of S
ES h
as
been
app
lied
to C
hine
se su
rvey
dat
a, w
hich
app
ears
to
hav
e gr
eat p
rom
ise
for i
nequ
ality
rese
arch
mov
ing
forw
ard
Witt
enbe
rg a
nd L
eib-
bran
dt (2
017)
PCA
Econ
omic
sC
ross
-sec
tiona
l su
rvey
No
Sout
h A
fric
aa,
f, g
, hC
are
mus
t be
take
n be
fore
app
lyin
g in
equa
lity
mea
sure
s to
ass
et in
dice
s—G
ini c
oeffi
cien
t can
onl
y be
app
lied
to th
e ab
senc
e or
pre
senc
e of
real
ass
ets,
and
Cow
ell-
Flac
haire
app
roac
hes c
an b
e ap
plie
d to
cat
egor
ical
va
riabl
esH
avin
g a
com
mon
ass
et w
ill u
sual
ly im
ply
a ne
gativ
e fa
c-to
r loa
ding
with
PCA
, MCA
, or f
acto
r ana
lysi
s, w
hich
co
uld
perv
erse
ly ra
nk th
e ho
useh
old
as p
oore
r tha
n on
e la
ckin
g th
e as
set a
t all
Ass
ets c
an b
e po
oled
ove
r tim
e if
care
is ta
ken
with
va
riabl
es li
ke la
ndlin
es a
nd c
ell p
hone
s, w
hich
can
be
com
bine
d in
to o
ne “
phon
e” a
sset
Ass
et in
equa
lity
fell
from
a G
ini c
oeffi
cien
t of 0
.47
to
0.29
from
199
3 to
200
8 in
Sou
th A
fric
a
42 M. J. P. Poirier et al.
1 3
Tabl
e 3
(con
tinue
d)
Arti
cle
SES
mea
sure
sD
isci
plin
eSt
udy
desi
gnSp
earm
an’s
ρ
retri
eved
Cou
ntrie
sK
ey id
easa
Mai
n co
ntrib
utio
ns to
mod
el d
evel
opm
ent
Zelle
r et a
l. (2
006)
PCA
, Inc
ome
Econ
omic
sC
ross
-sec
tiona
l su
rvey
No
Ban
glad
esh,
Kaz
akh-
stan
, Per
u, a
nd
Uga
nda
a, c
, f, i
PCA
gen
eral
ly h
as m
ore
erro
r in
clas
sify
ing
pove
rty
than
qua
ntile
, ord
inar
y le
ast s
quar
es, p
robi
t, an
d lin
ear
prob
abili
ty m
odel
s in
Ban
glad
esh,
Kaz
akhs
tan,
Per
u,
and
Uga
nda,
but
with
goo
d en
ough
per
form
ance
to b
e in
vesti
gate
d fu
rther
a a: In
com
e ve
rsus
Ass
ets;
b: C
onsu
mpt
ion
vers
us A
sset
s c: A
sset
inde
x co
mpa
rison
s; d
: Ass
ets a
nd E
duca
tion;
e: A
sset
s and
Hea
lth; f
: Urb
an–R
ural
Dyn
amic
s; g
: Rob
ustn
ess
to V
aria
ble
Loss
; h: I
ncom
e/C
onsu
mpt
ion
Ineq
ualit
y; i:
Wor
ldw
ide
Com
para
bilit
y
43Approaches and Alternatives to the Wealth Index to Measure…
1 3
References
Aaberge, R., & Melby, I. (1998). The sensitivity of income inequality to choice of equivalence scales. Review of Income and Wealth, 44, 565–569. https ://doi.org/10.1111/j.1475-4991.1998.tb002 99.x.
Ako-Arrey, D. E., Brouwers, M. C., Lavis, J. N., Giacomini, M. K., Haines, A., Dolea, C. M., et al. (2016). Health systems guidance appraisal—A critical interpretive synthesis. Implementation Science, 11, 9. https ://doi.org/10.1186/s1301 2-016-0373-y.
Amek, N., Vounatsou, P., Obonyo, B., Hamel, M., Odhiambo, F., Slutsker, L., et al. (2015). Using health and demographic surveillance system (HDSS) data to analyze geographical distribution of socio-economic status; an experience from KEMRI/CDC HDSS. Acta Tropica, 144, 24–30. https ://doi.org/10.1016/j.actat ropic a.2015.01.006.
Arsenault, C., Harper, S., Nandi, A., Mendoza Rodríguez, J. M., Hansen, P. M., & Johri, M. (2017). Moni-toring equity in vaccination coverage: A systematic analysis of demographic and health surveys from 45 Gavi-supported countries. Vaccine, 35, 951–959. https ://doi.org/10.1016/j.vacci ne.2016.12.041.
Aryeetey, G. C., Jehu-Appiah, C., Spaan, E., D’Exelle, B., Agyepong, I., & Baltussen, R. (2010). Identi-fication of poor households for premium exemptions in Ghana’s National Health Insurance Scheme: Empirical analysis of three strategies. Tropical Medicine & International Health, 15, 1544–1552. https ://doi.org/10.1111/j.1365-3156.2010.02663 .x.
Azzarri, C., Carletto, G., Davis, B., & Zezza, A. (2005). Monitoring poverty without consumption data: An application using the Albania Panel Survey. ESA working paper. https ://doi.org/10.2753/EEE00 12-87554 40103 .
Balen, J., McManus, D. P., Li, Y. S., Zhao, Z. Y., Yuan, L. P., Utzinger, J., et al. (2010). Compari-son of two approaches for measuring household wealth via an asset-based index in rural and peri-urban settings of Hunan province, China. Emerging Themes in Epidemiology, 7, 7. https ://doi.org/10.1186/1742-7622-7-7.
Benini, A. (2007). The wealth of the poor: Simplifying living standards measurements with Rasch scales? [Unpublished Manuscript], Washington, DC.
Blumenstock, B. J. E. (2018). Estimating economic characteristics with phone data † 72–76. https ://doi.org/10.1257/pandp .20181 033.
Boccia, D., Hargreaves, J., Howe, L. D., De Stavola, B. L., Fielding, K., Ayles, H., et al. (2013). The meas-urement of household socio-economic position in tuberculosis prevalence surveys: A sensitivity analy-sis. The International Journal of Tuberculosis and Lung Disease, 17, 39–45. https ://doi.org/10.5588/ijtld .11.0387.
Bollen, K. A., Glanville, J. L., & Stecklov, G. (2002). Economic status proxies in studies of fertility in developing countries: Does the measure matter? Population Studies (NY), 56, 81–96. https ://doi.org/10.1080/00324 72021 3796.
Booysen, F., van der Berg, S., Burger, R., Maltitz, M. Von, & Rand, G Du. (2008). Using an asset index to assess trends in poverty in seven Sub-Saharan African countries. World Development, 36, 1113–1130. https ://doi.org/10.1016/j.world dev.2007.10.008.
Boyko, J. A., Lavis, J. N., Abelson, J., Dobbins, M., & Carter, N. (2012). Deliberative dialogues as a mecha-nism for knowledge translation and exchange in health systems decision-making. Social Science and Medicine, 75, 1938–1945. https ://doi.org/10.1016/j.socsc imed.2012.06.016.
Chakraborty, N. M., Fry, K., Behl, R., & Longfield, K. (2016). Simplified asset indices to measure wealth and equity in health programs: A reliability and validity analysis using survey data from 16 countries. Global Health: Science and Practice, 4, 141–154. https ://doi.org/10.9745/GHSP-D-15-00384 .
Chasekwa, B., Maluccio, J. A., Ntozini, R., Moulton, L. H., Wu, F., Smith, L. E., et al. (2018). Measuring wealth in rural communities: Lessons from the sanitation, hygiene, infant nutrition efficacy (SHINE) trial. PLoS ONE, 13, 1–19. https ://doi.org/10.1371/journ al.pone.01993 93.
Chuma, J., & Molyneux, C. (2009). Estimating inequalities in ownership of insecticide treated nets: Does the choice of socio-economic status measure matter? Health Policy Plan., 24, 83–93. https ://doi.org/10.1093/heapo l/czn05 0.
Córdova, A. (2008). Methodological note: Measuring relative wealth using household asset indicators. AmericasBarometer Insights. https ://www.vande rbilt .edu/lapop /insig hts/I0806 en_v2.pdf.
Dixon-Woods, M., Agarwhal, S., Jones, D., Young, B., & Sutton, A. (2005). Synthesising qualitative and quantitative evidence: A review of possible methods. Journal of Health Services Research & Policy, 10, 45–53. https ://doi.org/10.1258/13558 19052 80180 4.
Dixon-Woods, M., Cavers, D., Agarwal, S., Annandale, E., Arthur, A., Harvey, J., et al. (2006). Conducting a critical interpretive synthesis of the literature on access to healthcare by vulnerable groups. BMC Medical Research Methodology, 6, 35. https ://doi.org/10.1186/1471-2288-6-35.
44 M. J. P. Poirier et al.
1 3
Doku, D., Koivusilta, L., & Rimpelä, A. (2010). Indicators for measuring material affluence of adolescents in health inequality research in developing countries. Child Indicators Research, 3, 243–260. https ://doi.org/10.1007/s1218 7-009-9045-7.
Ellen, M. E., Wilson, M. G., Vélez, M., Shach, R., Lavis, J. N., Grimshaw, J. M., et al. (2018). Addressing overuse of health services in health systems: A critical interpretive synthesis. Health Research Policy and Systems, 16, 1–14. https ://doi.org/10.1186/s1296 1-018-0325-x.
Ergo, A., Ritter, J., Gwatkin, D. R., & Binkin, N. (2016). measurement of health program equity made easier: Validation of a simplified asset index using program data from Honduras and Senegal. Global Health: Science and Practice, 4, 155–164.
Ferguson, B. D., Tandon, A., Gakidou, E., & Murray, C. J. L. (2003). Estimating permanent income using indicator variables, evidence and information for policy cluster. Geneva: World Health Organization.
Filmer, D., & Pritchett, L. H. (2001). Estimating wealth effects without expenditure data—or tears: An application to educational enrollment in states of India. Demography, 38, 115–132. https ://doi.org/10.1353/dem.2001.0003.
Filmer, D., & Scott, K. (2012). Assessing asset indices. Demography, 49, 359–392. https ://doi.org/10.1007/s1352 4-011-0077-5.
Fink, G. (2016). Estimated household income for DHS and MICS surveys [WWW Document]. Percentile level predictions for all countries. https ://www.hsph.harva rd.edu/gunth er-fink/data/. Accessed August 18, 2018.
Fink, G., Victora, C. G., Harttgen, K., Vollmer, S., Vidaletti, L. P., & Barros, A. J. D. (2017). Measuring socioeconomic inequalities with predicted absolute incomes rather than wealth quintiles: A compara-tive assessment using child stunting data from national surveys. American Journal of Public Health, 107(4), 550–555. https ://doi.org/10.2105/AJPH.2017.30365 7.
Gough, D., David, A., Oliver, S., & Thomas, J. (2017). An introduction to systematic reviews (2nd ed.). London: Sage.
Harttgen, K., & Vollmer, S. (2013). Using an asset index to simulate household income. Economic Letters, 121, 257–262. https ://doi.org/10.1016/j.econl et.2013.08.014.
Higgins, J. P., & Green, S. (2011). Cochrane handbook for systematic reviews of interventions (5.1.0.). Chichester: The Cochrane Collaboration. https ://doi.org/10.1002/97804 70712 184.
Homenauth, E., Kajeguka, D., & Kulkarni, M. A. (2017). Principal component analysis of socioeconomic factors and their association with malaria and arbovirus risk in Tanzania: A sensitivity analysis. Jour-nal of Epidemiology and Community Health, 71, 1046–1051. https ://doi.org/10.1136/jech-2017-20911 9.
Houweling, T. A. J., Kunst, A. E., & Mackenbach, J. P. (2003). Measuring health inequality among chil-dren in developing countries: Does the choice of the indicator of economic status matter? Interna-tional Journal for Equity in Health, 2, 8.
Howe, L. D., Galobardes, B., Matijasevich, A., Gordon, D., Johnston, D., Onwujeke, O., et al. (2012). Measuring socio-economic position for epidemiological studies in low- and Middle-income coun-tries: A methods of measurement in epidemiology paper. International Journal of Epidemiology, 41, 871–886. https ://doi.org/10.1093/ije/dys03 7.
Howe, L. D., Hargreaves, J. R., Gabrysch, S., & Huttly, S. R. (2009). Is the wealth index a proxy for con-sumption expenditure? A systematic review. Journal of Epidemiology and Community Health, 63, 871–877. https ://doi.org/10.1136/jech.2009.08802 1.
Jean, N., Burke, M., Xie, M., Davis, W. M., Lobell, D. B., & Ermon, S. (2016). Machine learning to pre-dict poverty. Science (80-.), 353, 790–794.
Johnston, D., & Abreu, A. (2016). The asset debates: How(not) to use asset indices to measure well-being and the middle class in Africa. African Affairs (Lond), 115, 399–418. https ://doi.org/10.1093/afraf /adw01 9.
Joseph, G., da Silva, I. C. M., Fink, G., Barros, A. J. D., & Victora, C. G. (2018). Absolute income is a better predictor of coverage by skilled birth attendance than relative wealth quintiles in a multi-country analysis: Comparison of 100 low- and middle-income countries. BMC Pregnancy Child-birth, 18, 104. https ://doi.org/10.1186/s1288 4-018-1734-0.
Kolenikov, S., & Angeles, G. (2009). Socioeconomic status measurement with discrete proxy variables: Is principal component analysis a reliable answer? Review of Income and Wealth, 55, 128–165.
Lindelow, M. (2006). Sometimes more equal than others: How health inequalities depend on the choice of welfare indicators. Health Economics, 15, 263–279. https ://doi.org/10.1002/hec.1058.
Manthalu, G., Nkhoma, D., & Kuyeli, S. (2010). Simple versus composite indicators of socioeconomic status in resource allocation formulae: The case of the district resource allocation formula in Malawi. BMC Health Services Research, 10, 6. https ://doi.org/10.1186/1472-6963-10-6.
45Approaches and Alternatives to the Wealth Index to Measure…
1 3
McKenzie, D. J. (2005). Measuring inequality with asset indicators. Journal of Population Economics, 18, 229–260. https ://doi.org/10.1007/s0014 8-005-0224-7.
Michelson, H. C. (2013). Measuring poverty in the millennium villages: The effect of asset index choice. World Development, 49, 917–935.
Moat, K. A., Lavis, J. N., & Abelson, J. (2013). How contexts and issues influence the use of policy-relevant research syntheses: A critical interpretive synthesis. Milbank Quarterly, 91, 604–648. https ://doi.org/10.1111/1468-0009.12026 .
Mohanty, S. K. (2009). Alternative wealth indices and health estimates in India. Genus, 65, 113–137. https ://doi.org/10.4402/genus -61.
Mohsena, M., Mascie-Taylor, C. G. N., & Goto, R. (2010). Association between socio-economic status and childhood undernutrition in Bangladesh; a comparison of possession score and poverty index. Public Health Nutrition, 13, 1498–1504. https ://doi.org/10.1017/S1368 98001 00017 58.
Moser, C., & Felton, A. (2007). The construction of an asset index measuring asset accumulation in Ecuador. Chronic Poverty Research Centre working paper 87. The Brookings Institution, Washing-ton, DC.
Ngo, D., & Christiaensen, L. (2018). The performance of a consumption augmented asset index in rank-ing households and identifying the poor. World Bank Policy Research working paper.
Nkonki, L. L., Chopra, M., Doherty, T. M., Jackson, D., & Robberstad, B. (2011). Explaining house-hold socio-economic related child health inequalities using multiple methods in three diverse settings in South Africa. International Journal for Equity in Health, 10(1), 13. https ://doi.org/10.1186/1475-9276-10-13.
Nwaru, B. I., Klemetti, R., Kun, H., Hong, W., Yuan, S., Wu, Z., et al. (2012). Maternal socio-economic indices for prenatal care research in rural China. The European Journal of Public Health, 22, 776–781. https ://doi.org/10.1093/eurpu b/ckr18 2.
Opuni, M., Peterman, A., & Bishai, D. (2011). Inequality in prime-age adult deaths in a high AIDS mortality setting: Does the measure of economic status matter. Health Economics, 20, 1298–1311. https ://doi.org/10.1002/hec.1671.
Reidpath, D. D., & Ahmadi, K. (2014). A novel nonparametric item response theory approach to measur-ing socioeconomic position: A comparison using household expenditure data from a Vietnam health survey, 2003. Emerging Themes in Epidemiology, 11(1), 9. https ://doi.org/10.1186/1742-7622-11-9.
Rohner, F., Tschannen, A. B., Northrop-Clewes, C., Kouassi-Gohou, V., Bosso, P. E., & Nicholas Mas-cie-Taylor, C. G. (2012). Comparison of a possession score and a poverty index in predicting anae-mia and undernutrition in pre-school children and women of reproductive age in rural and urban Côte d’Ivoire. Public Health Nutrition, 15, 1620–1629. https ://doi.org/10.1017/S1368 98001 20028 19.
Rutstein, S. O. (2008). The DHS Wealth Index: Approaches for rural and urban areas. Demographic and Health Survey working papers, Calverton, Maryland.
Sahn, D. E., & Stifel, D. (2003). Exploring alternative measures of welfare in the absence of expenditure data. Review of Income and Wealth, 49, 463–489. https ://doi.org/10.1111/j.0034-6586.2003.00100 .x.
Smits, J., & Steendijk, R. (2013). The International Wealth Index (IWI) (No. 12–107). NiCE working paper, Nijmengen, The Netherlands. https ://doi.org/10.1007/s1120 5-014-0683-x.
Spearman, C. (1904). The proof and measurement of association between two things. American Journal of Psychology, 15, 72–101. https ://doi.org/10.1177/03635 46578 00600 604.
Tusting, L. S., Rek, J. C., Arinaitwe, E., Staedke, S. G., Kamya, M. R., Bottomley, C., et al. (2016). Measur-ing socioeconomic inequalities in relation to malaria risk: A comparison of metrics in Rural Uganda. American Journal of Tropical Medicine and Hygeine, 94, 650–658. https ://doi.org/10.4269/ajtmh .15-0554.
Ucar, B. (2015). The usability of asset index as an indicator of household economic status in Turkey: Com-parison with expenditure and income data. Social Indicators Research, 121, 745–760. https ://doi.org/10.1007/s1120 5-014-0670-2.
United Nations. (2015). Transforming our world: The 2030 agenda for sustainable development, A/RES/70/1. https ://doi.org/10.1007/s1339 8-014-0173-7.2.
Van Leth, F., Guilatco, R. S., Hossain, S., Van’t Hoog, A. H., Hoa, N. B., Van Der Werf, M. J., et al. (2011). Measuring socio-economic data in tuberculosis prevalence surveys. The International Journal of Tuberculosis and Lung Disease, 15, S58–S63. https ://doi.org/10.5588/ijtld .10.0417.
Vandemoortele, M. (2014). Measuring household wealth with latent trait modelling: An application to Malawian DHS data. Social Indicators Research, 118, 877–891. https ://doi.org/10.1007/s1120 5-013-0447-z.
46 M. J. P. Poirier et al.
1 3
Vu, L., Tran, B., & Le, A. (2011). The use of total assets as a proxy for socioeconomic status in Northern Vietnam. Asia-Pacific Journal of Public Health, 23, 996–1004. https ://doi.org/10.1177/10105 39510 36163 8.
Vyas, S., & Kumaranayake, L. (2006). Constructing socio-economic status indices: How to use principal components analysis. Health Policy and Planning, 21, 459–468. https ://doi.org/10.1093/heapo l/czl02 9.
Ward, P. (2014). Measuring the level and inequality of wealth: An application to China. Review of Income and Wealth, 60, 613–635. https ://doi.org/10.1111/roiw.12063 .
Wittenberg, M., & Leibbrandt, M. (2017). Measuring inequality by asset indices: A general approach with application to South Africa. Review of Income and Wealth, 63, 706–730. https ://doi.org/10.1111/roiw.12286 .
Zeller, M., Houssou, N., Alcaraz, G. V, Schwarze, S., & Johannsen, J. (2006). Developing poverty assess-ment tools based on principal component analysis: Results from Bangladesh, Kazakhstan, Uganda, and Peru. In International association of agricultural economists conference, Gold Coast, Australia, (pp. 1–24).
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Affiliations
Mathieu J. P. Poirier1,2 · Karen A. Grépin3 · Michel Grignon4
1 School of Kinesiology and Health Science, York University, 4700 Keele St., Toronto, ON M3J 1P3, Canada
2 Department of Health Research Methods, Evidence, and Impact, McMaster University, Hamilton, Canada
3 Department of Health Sciences, Wilfrid Laurier University, Waterloo, Canada4 Department of Economics, McMaster University, Hamilton, Canada