You are on page 1of 9

Looking at Human Development Index through the lens of Intensity of Multi-

Dimensional poverty and income

Bhupendra Badgaiyan

Introduction

In this note we attempt to examine the relationship between Human


Development Index (HDI) used by United Nations Development
Programme (UNDP) to assess and monitor the progress of social
and economic development within and across nations and the Intensity of
Multi-Dimensional poverty as developed by Oxford Poverty and Human
Development Initiative [1] along with the mean per capita income. It may
be noted that all the three are average variables for a state or a nation.
Further it may be noted that HDI is sort of outcome (result) variable so far
as development is concerned, while Multi-Dimensional poverty Index
(MDI) comprises of input or process variables for non-monetary
interventions, the deprivations of which may affect human development.
Thus for example malnutrition, school attendance, ownership of adequate
assets, sanitation, drinking water and electricity etc which are captured in
MDI may further or hinder development, so it is quite reasonable to expect
a relationship between multi-dimensional poverty (intensity) and HDI. The
intensity of MDI is defined as the average deprivation score among the
multi-dimensionally poor individuals. Also, state-wise mean per capita
Gross State Domestic Product has been included in the analysis in order to
capture the average income part of the HDI.

The method deployed is that of linear regression with HDI as dependent


variable and Intensity of multi-dimensional poverty and average per capita
State GDP as independent variables.
Description of variables

The Human Development Index is a summary measure of achievements in


three key dimensions of human development: a long and healthy life,
access to knowledge and a decent standard of living. The HDI is the
geometric mean of normalized indices for each of the three dimensions
[5]. The indices for each of these dimensions are respectively life
expectancy at birth, mean and expected years of schooling and Gross
National Income per capita.

The Multi-Dimensional Poverty Index (MDI) captures the non-monetary


deprivations of a population along three dimensions – health, education
and standard of living using 10 indicators and measures the deprivation
along each of them. The indicators under each dimension are as follows :

Health – Nutrition (presence of malnourished child/men/ women of


specified age in the household); child and adolescent mortality in the last
five years; adequate antenatal care ( as specified) during the last child-
birth.

Education – years of schooling (no household member above the age of 10


years has completed six years of education) ; school attendance (any school
going child not attending school upto the age when he would complete
class 8th.).

Standard of living – Captured by - cooking fuel used (traditional like coal,


dung etc) ; unimproved or not having sanitation facility; drinking water (no
improved source within 30 minute walk for round trip); no electricity ; no
adequate housing ; not owning more than one of assets like TV, radio,
bicycle etc.; no household member has bank or post office account.
Each household member is assigned score of one if there is deprivation in
any of these indicators. All the three dimensions are given equal weight i.e.
1/3 and out of that, all indicators in a given dimension are given equal
weight e.g. the weight for both years of schooling and school attendance are
1/6 and similarly for other indicators. Based on this a weighted score of
each household member is computed and an individual is assigned as being
multi-dimensional poor if the weighted score is more than 33.3 percent.

The Intensity of multi-dimensional poverty is defined by the sum of


deprivation score of all multi-dimensionally poor individuals divided by the
total number of multi-dimensionally poor, thus it is an average extent of
deprivation suffered by the multi-dimensional poor. It may be noted MDI
is product of two factors viz. Head-count ratio (H) – proportion of persons
with multi-dimensional poverty and Intensity of multi-dimensional
poverty. The method has been discussed in reference [2] cited above.

Average GSDP per capita is obtained by Gross State Domestic Product


divided by the population.

Data source and unit of analysis

Here the analysis has been done for India using state-wise data available
from various agencies. State-wise HDI (for the year 2019) is available from
Global_Data_Lab, Institute for Mangement Research, Radboud University
[2], state-wise Intensity of multi-dimensional poverty for India is available
from publication of Niti Aayog, Government of India [3], which is based on
National Family Health Survey –IV (2015-16) and state-wise per capita
Gross State Domestic Product for the year 2018-19 is available from
Ministry of Statistics and Programme Implementation, Government of India
[4]. In our analysis mean per capita Gross State Domestic Product
(GSDPPCAP) has been taken in Lakh Rs.

Regression and the results

With the data thus obtained, a linear regression with a constant term has
been performed with HDI as dependent variable and with Intensity of
Multi-Dimensional poverty as one independent variable and GSDPPCAP
as another independent variable. The data and the regression results are
given in Annexure.

It is found that both the dependent variables have significant coefficients


and R2 is greater than 73%. The coefficient for GSDPPCAP is about 0.022
(p ~ 0.005) while that for Intensity its -0.940 (p ~ 0.00). The value of the
constant term is 1.05 ( p~0.00).

Both the coefficients have the right sign, as an increase in mean per capita
GSDP will raise the HDI and also as intensity of multi-dimensional poverty
increases the HDI decreases. Robust standard errors have been used and
the normality test for errors give a normal distribution (Shapiro–Wilk
test p~ 0.16).

HDI for India

The above results have been used for estimating HDI at all India level using
mean per capita GSDP of Rs 1.42328 lakh [4] and Intensity of 0.4713 [3]
for all India. The estimated value of HDI for all India is 0.641 against the
actual HDI of 0.645 [2]. The error in this estimate is just of the order of
0.6%.
Conclusion

Thus we see that besides being an important measure for estimating depth
of multi-dimensional poverty, the MDI (intensity part) can also be used to
estimate (predict) HDI. In fact, the relationship is expected, as both have
the non-monetary part of deprivations – on the one hand HDI essentially
captures the outcome side while on the other, Intensity of multi-
dimensional poverty measures mainly input/process side of
interventions. It can be therefore argued that such an association in a way
justifies and strengthens the methodology for calculation of MDI as done by
the Niti Aayog, Government of India [3] and as put forward by Alkire et
al.[1].

The limitation of the study is that though (baseline) MPI has been
calculated by Niti Aayog in 2021, it is based on data collected in National
Family Health Survey– IV conducted during 2015-16. Therefore as the new
MPI calculated from latest NFHS –V data is available it would be interesting
to see how if affects the coefficients of regression.

The data and the summary of estimation results (performed in Stata) are
given in Annexure.

______________________________________________________________________________________
References

1. Alkire, S. and Foster, J. 2011. Counting and Multidimensional Poverty


Measurement, Journal of Public Economics, 95(7-8), 476-487.

2. Global_Data_Lab,Institute for Mangement Research, Radboud


University
https://globaldatalab.org/shdi/shdi/IND/?levels=1%2B4&interpola
tion=1&extrapolation=0&nearest_real=0&years=2019%2B2014%2B
2009%2B2004%2B1999%2B1994%2B1990

3. National Multi-Dimensional Poverty Index-based on NFHS– 4 (2015-


16)
https://www.niti.gov.in/sites/default/files/2021-
11/National_MPI_India-11242021.pdf

4. State-wise per capita GSDP 2018-19


https://statisticstimes.com/economy/india/indian-states-gdp-per-
capita.php (source –MoSPI,Government of India)

5. Technical Notes – Human Development Index


https://hdr.undp.org/sites/default/files/hdr2020_technical_notes.p
df
Annexure

Table I – The regression variables

Sl No. State / UT Human Mean Gross State Intensity of Multi-


Development Domestic Product Dimensional
Index (HDI) per capita (In Poverty
Lakh Rs. per
annum)
0.741 - .41
Andaman
1.
0.649 1.68083 .43
Andhra Pradesh
2.
0.661 1.52716 .47
Arunachal Pradesh
3.
0.613 .92533 .48
Assam
4.
0.574 .44652 .51
Bihar
5.
0.776 3.5903 .43
Chandigarh
6.
0.611 1.05424 .45
Chhattisgarh
7.
0.663 - .45
Dadra Haveli
8.
0.708 - .44
Daman
9.
0.746 3.94216 .44
Delhi
10.
0.763 4.7637 .4
Goa
11.
0.672 2.25048 .45
Gujarat
12.
0.708 2.60287 .44
Haryana
13.
0.725 2.11325 .39
Himachal
14.
0.688 1.1089 .44
Jammu Kashmir
15.
0.599 .79937 .48
Jharkhand
16.
0.683 2.2768 .43
Karnataka
17.
0.782 2.25484 .39
Kerala
18.
0.751 - .36
Lakshadweep
19.
0.603 .99025 .47
Madhya
20.
0.697 2.16375 .44
Maharashtra
21.
0.697 .82796 .44
Manipur
22.
0.656 .96015 .48
Meghalaya
23.
0.704 1.87919 .47
Mizoram
24.
0.679 1.29978 .46
Nagaland
25.
0.606 1.11892 .46
Odisha
26.
0.74 2.25936 .39
Puducherry
27.
0.724 1.71906 .44
Punjab
28.
0.628 1.23343 .47
Rajasthan
29.
0.717 4.35199 .41
Sikkim
30.
0.709 2.15784 .4
Tamil
31.
0.669 2.25048 .43
Telangana
32.
0.658 1.25459 .45
Tripura
33.
0.596 .7068 .48
Uttar
34.
0.683 2.20257 .44
Uttarakhand
35.
0.641 1.10728 .45
West Bengal
36.
Table II - Regression Output

. reg hdin gsdppcap intensity, robust

Linear regression Number of obs =32


F( 2, 29) = 43.85
Prob > F = 0.0000
R-squared = 0.7339
Root MSE = .02947

Robust
hdin Coef. Std. Err. t P>|t| [95% Conf. Interval]
---------------------------------------------------------------------------------------------
gsdppcap | .0225741 .0074357 3.04 0.005 .0073665 .0377818
intensity | -.9403852 .2194149 -4.29 0.000 -1.389139 -.4916313
cons | 1.052171 .1076819 9.77 0.000 .8319372 1.272406
------------------------------------------------------------------------------------------------
. predict resid, residual
(4 missing values generated)

. swilk resid

Shapiro-Wilk W test for normal data

Variable | Obs W V z Prob>z


-------------------------------------------------------------------
resid | 32 0.95207 1.599 0.974 0.16500

--------------------------------------------------------------------------------------------------------------------

You might also like