Multidimensional poverty approach and development of poverty indicators: the case of Bangladesh

Abstract

The multidimensionality of poverty is often neglected at the policy-formulation stage in developing countries. As need priorities of local poor vary across countries and within regions of the same country, an understanding of finely tuned poverty indicators should be the starting point in formulating country-specific poverty-reduction strategies. By reviewing cross-country studies and existing poverty literature, this paper demonstrates the construction of a multidimensional poverty model to better inform poverty listing in the case of Bangladesh.

Keywords:

• multidimensional poverty
• economic functioning
• social functioning
• Bangladesh

JEL Classification:

• I380
• I390
• O290
• C190

Notes

1. In Kerala, India 16% of people are multidimensional poor compared to 81% in the Indian state of Bihar (Alkire and Santos 2010).

2. Book titled Life and Labour of the People in London.

3. Known as A Study of Town Life.

4. Sen (1984, 1993) also emphasized that indicators of poverty and well-being should be area specific and thus reflect the social values and culture of the local community.

5. According to Alkire and Santos (2010, 1), ‘Ethiopia reduced multidimensional poverty by improving nutrition and water, whereas Bangladesh improved it by sending children to school. Ghana improved several aspects of poverty at once.’

6. For details, see the works of Booysen (2002) and McGillivray and Noorbakhsh (2007).

7. Initially the concept of relative poverty was used for developed countries. However, the concept is now used widely also in less-developed and developing countries (Mabughi and Selim 2006).

8. Beginning with the Tanner Lecture ‘Equality of What?’, first delivered at Stanford University in 1979.

9. Sen defines capability in a broader sense, which comprises a large set of functioning and is why capability in most cases is used as a synonym to ‘capability set’ (Qizilbash 2005).

10. In DCI, poor households are defined as those with per capita energy intake less than the standard per capita requirement of energy (1805 and 2112 kilocalories for extreme and moderate poverty line), BBS (2007).

11. FEI method sets the poverty line as the income or consumption level at which ‘basic needs’ are met.

12. Upazillas are the lowest level administrative government in Bangladesh.

13. Districts are Lalmonirhat, Gaibandha, Kurigram and Nilphamari from the northern part, Potuakhali, Jessore, Barguna, Jhalokathi and Barisal from the southern part and Feni, Brahmanbaria and Lakshmipur from the central part of Bangladesh.

14. For example, outcomes of economic well-being are income per month, savings per month and use of sanitary latrine; outcomes of asset building is home ownership pattern, landholding size and landholding status (whether bought new or sold or owned by inheritance).

15. Acceptable range of MSA is above 0.50 (Hair et al. 2010, 132). Furthermore, an individual MSA value for the items ranges from 0.544 to 0.925, which is another positive indication of data adequacy.

16. Considering the total number of factors until the eigenvalue drops below 1.

17. Similar rules were followed in the marketing literature by Shimp and Sharma (1987), Bawa (2004), and in research methodology by Hair et al. (2010).

18. For a good fit, the preferred CFI value is closer to 0.95 and the RMSEA value should be less than or equal to 0.05.

19. For instance, small and significant chi-square values, p-values greater than .05, GFI and CFI greater than 0.90, RMSEA less than 0.05 and Hoelters value more than 200 (Byrne 2001).

20. All of these items have loading less than .30.

21. This problem factor (Empowerment) has been identified through a trial and error process of checking the goodness of fit values of the model by deleting one factor at a time along with its items.

22. For example, our survey results showed that about 96.3% of the respondents cast their vote regularly and 98% of these regular voters vote for their preferred candidate. This illustrates that from the voting point of view beneficiaries have quite a high level of empowerment, thus these items and the ‘Empowerment’ factor were identified as less relevant to the poverty model for Bangladesh.

23. Explanations on all correlations are not included here. We will be happy to provide the detailed results to any interested reader.

24. It includes factor loading testing, average variance extracted test and construct reliability measures. Detailed results can be provided upon request.

25. Interested readers are requested to contact the authors for further details.

26. χ² value dropped to 687.553 from 1115.859 of preliminary model. Hoelter's values are 304 and 320, which are greater than 200. Akaike's information criterion and expected cross-validation index are both lower than the independence model.

27. In case of the structural model, Cronbach's alpha is always under-reported (Arbuckle 2009), thus we used coefficient H value.