This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The UNDP (1997, 2006) computes the HPI by using the percentage of people not expected to survive to age 40 as the proxy for longevity, the percentage of adults who are illiterate as the proxy for knowledge, and the percentage of people without access to safe water, percentage of people without access to health services, and percentage of moderately and severely under weight children under five as the proxies for a decent standard of living. While the UNDP continues to use this approach to compute the HPI for developing countries (without the percentage of people without access to health services), it uses a slightly different approach for the OECD countries. For these countries, the HPI is computed as the unweighted average of four separate measures including the probability at birth of not surviving to age 60, the percentage of adults lacking functional literacy skills, the percentage of people living below the poverty line (defined as the 50 percent of median adjusted household disposable income), and long term (over 12 months) unemployment rate (UNDP 2005).
- 2.
Indicators such as life expectancy, underweight children, and long term unemployment, which are used by the UNDP, for example, cannot be used to meaningfully assess poverty at the individual (or household) level.
- 3.
These theoretically substantive issues will be revisited while discussing the operational issues later in the chapter.
- 4.
While the UNDP's (1997, 2005) human poverty index is constructed as a composite of economic well-being, capability, and social exclusion especially in case of OECD countries, the approach is largely simplistic and narrow in coverage. In most cases, for example, GDP, income poverty, or child malnutrition, and access to safe drinking water are used as the indicators of economic well-being, longevity and adult literacy are used as the indicators of capability, and long term unemployment is used as the indicator of social exclusion.
- 5.
Although more recent developments in the capability approach has seen attempts to incorporate the social dimensions of freedom (Sen 2000), this has been in response to the relevance of social inclusion explanations of poverty with the central role of social institutions.
- 6.
This applies specifically to the conventional approaches to poverty. While more recent, innovative approaches do not necessarily make such assumption, their theoretical distinction from other approaches tend to downplay the other dimensions of poverty.
- 7.
The framework provides the three social inclusion (sub)dimensions as stand-alone indicators of social inclusion, without any relationships among them as well as with other poverty dimensions. The actual operationalization developed later, however, will treat these (sub)dimensions as separate poverty dimensions with a complex web of their interrelationships.
- 8.
Appropriateness and completeness are two distinct concepts, however, with the former dealing with face validity and the latter dealing content validity. Because of the comprehensive nature of the poverty dimensions, it will be difficult to fully establish content validity with a given set of indicators, let alone in an empirical sense. The use of any set of indicators needs to be substantiated with appropriate theory. But the empirical test will be useful to establish face validity of each indicator.
- 9.
The existing income-based poverty lines appear to be absolutely political (Glennerster 2002). In the 1960s, for example, the official poverty line in the US was set out to be approximately one-half the median income, which by the 1990s was reduced to one-third of the median income. With growing housing, transportation, insurance, and childcare costs in the United States, even those focusing on income or consumption based poverty thresholds propose divergent arguments over what the actual poverty threshold should look like (Citro and Michael 1995; Dalaker 2005; Joassart-Marcelli 2005). As for the relative approach, on the other hand, whether to use 50 or 60 percent of the median income as the poverty line is debatable. While almost all European countries now use 60 percent standard, the UK uses the 50 percent standard for its official purposes.
- 10.
South Asia provides an interesting example. While the poverty headcount ratio declined in this region during the past two decades, the actual headcount of the poor did not, owing perhaps to population growth especially in families in poverty (Wagle 2007a).
- 11.
The multidimensional space of poverty can be based on five or three dimensions depending on whether one treats the three social inclusion (sub)dimensions as separate integrated dimensions. The three social inclusion (sub)dimensions are appropriately handled as separate dimensions in analyzing the poverty issues and in constructing and verifying the specific channels through which they are determined. At the same time, they can be appropriately aggregated as the overall social inclusion dimension for its use in identifying the poverty status of individuals. While the resulting poverty categorization on the social inclusion dimension is essentially multidimensional, working with three dimensions is both theoretically justified and operationally more manageable.
- 12.
It measures the difference between a poor person's income and the applicable poverty threshold.
- 13.
This is used to measure the length of time a person has been poor. Typically, 5 years is used as the applicable cutoff point.
- 14.
This does not mean, however, that this group ends up with less attention. Because who gets what from policies is determined through the political calculus, in reality, the most deserving poor may end up with the least amount of policy resources and attention (Berrick 2001; Sen 1995; Stone 2002).
- 15.
One may in fact include questions explicitly dealing with the perceived level of economic well-being. Like depression, stress, or mental health, however, these questions may not be very useful to identify the magnitude of well-being partly because of the misunderstanding of the concept or misreporting likely with these abstract and/or sensitive issues.
- 16.
These variables or dimensions are used here to be consistent with the multidimensional framework which includes them in one single model. Demographic variables are not included because they do not play a role in measuring any of the poverty dimensions and thus poverty.
- 17.
One can take a weighted or unweighted average of all of the indicators used with identification of poverty status occurring after this aggregation. Because the aggregates provide abstract estimates of the construct, it would be more difficult to use directly sensible, absolute approach to identify poverty status. Identifying relative poverty status would therefore be easier in this case.
- 18.
This would involve identifying poverty status using INC, WLTH, CONS, and SVIEWS separately. The income poverty status for the ith person would depend on whether or not INC i <INC *, where INC* represents the income poverty threshold. After the similar process is applied for WLTH, CONS, and SVIEWS, the multidimensional poverty status of the ith person would be Poor if INC i <INC *, WLTHi<WLTH *, and CONSi<CONS *, and Non-poor otherwise. Whether or not meeting the poverty conditions for all dimensions will qualify as the Poor, however, would depend on the specific rule applied.
- 19.
See the above footnotes for the unidimensional and multidimensional methods.
- 20.
See the above footnotes for the methods of aggregation.
- 21.
In this case, each equation would exclude the y vector containing the indicators of the dimension being regressed.
- 22.
This matrix would depend on the different covariance matrices including ψ(=E(ςς′)), B, Λ, and ϑe(=E(εε')).
- 23.
This matrix will include the covariances of the Y's only since these are the only observed variables used in the model.
- 24.
Berenger and Verdier-Chouchane (2007), Lelli (2001), Qizilbash (2003), and Qizilbash and Clark (2005), for example, identify the capability poverty statuses, including the definitely poor, definitely non-poor, and those in between. Person i would have the membership to the definitely poor if he or she is below the threshold \(D_j^* \) on all the indicators used (Si=1, if Dji<\(D_j^* \),\(\forall D_j \)) and to the definitely non-poor if he or she is above the threshold \(D_j^{**} \)on all the indicators used (Si=0, if Dji> \(D_j^{**} \),\(\forall D_j \)). In case of not having full membership to any of these groups, on the other hand, the degree of i's membership to the poor would depend on the weighted average of the membership scores on each of the indicators (0<Si<1, if Dij<\(D_j^* \) and Dij> \(D_j^{**} \) for at least one Dj). In this latter case, the degree of membership to the poor would be determined by the weighted average of\({{(D_j^{**} - D_{ij} )} \over {(D_j^{**} - D_j^* )}}\)for all indicators.
- 25.
Rather than identifying the poverty status on each of the dimensions, in this case, the dimension scores would have to be weighted prior to aggregation. The aggregate poverty status score would therefore be:U = \(\sum\limits_{d = 1}^3 {\eta _d Z_d } \), where Z's are the relative weights.The score U, therefore, would have to be evaluated using some predefined rule to derive poverty status.
- 26.
This alternative would necessitate extensive work to ascertain the contribution of each of the poverty dimensions and therefore their indicators to categorize one as poor or non-poor. It would also involve tremendous value judgments to decide how particular values for particular indicators could be compensatory for others, leaving the likelihood of the household of being poor unchanged. I have taken this alternative further in my more recent works focusing on capability poverty. See Wagle (2007d) for details.
- 27.
While the typical data input for the SEM include correlations (or covariances), the zero-order Pearson's correlation will not be appropriate in case of categorical variables. Here tetrachoric signifies the correlation between dichotomous variables, polychoric signifies the correlation between categorical variables, and polyserial signifies the correlation between categorical and continuous variables. These correlations requiring estimation of a series of Probit models for each categorical variable are often difficult and time-consuming to compute, especially when appropriate software are not readily available.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2008 Springer Science+Business Media, LLC
About this chapter
Cite this chapter
Wagle, U. (2008). Multidimensional Approach to Poverty. In: Multidimensional Poverty Measurement. Economic Studies in Inequality, Social Exclusion and Well-Being, vol 4. Springer, New York, NY. https://doi.org/10.1007/978-0-387-75875-6_3
Download citation
DOI: https://doi.org/10.1007/978-0-387-75875-6_3
Published:
Publisher Name: Springer, New York, NY
Print ISBN: 978-0-387-75874-9
Online ISBN: 978-0-387-75875-6
eBook Packages: Business and EconomicsEconomics and Finance (R0)