Growth, Inequality and Population Effects on Poverty Reduction

This blog, meant for a general readership, is about my new paper Decomposing Poverty Change: Deciphering Change in Total Population and Beyond. This is published in The Review of Income and Wealth that is available under Early View and is also open access. A slide share explaining the method is also uploaded as Decomposing Poverty Change: Within- and Between-group Effects at my bepress page.

It is believed that growth in the economy would address poverty reduction through the trickle-down effect. At the same time, equitable distribution or lowering inequality will also help reduce poverty while also leading us towards a welfare state.

A question that assumes importance is that if in a society the incidence of poverty has reduced (say, from 50 per cent to 40 per cent) then then how much of this is on account of growth in the economy and how much is on account of reductions in inequality. 

Conventionally, poverty reduction has tried to compute these two effects by holding the other constant. What is the growth effect, if inequality is held constant; and what is the inequality effect, if there is no growth.

While dealing with the changes in the proportion of poor, the population remains hidden. However, if the population consists of different subgroups (say, rural and urban) and the overall incidence of poverty is an weighted average then the changes in population shares between the subgroups could also have implications in our understanding of poverty reduction. 

The literature refers to this change in population share as between-group effect (say, migration of people from rural to urban regions has helped reduced poverty in the rural regions, but it has increased it in urban regions while at an aggregate level it has helped reduce poverty). As against this, the growth and inequality effects are referred to as within-group effects.

Further, with the population being hidden, the conventional computations for delineating the effect of growth on poverty reduction have implicitly assumed that there is no change in total population. This is contrary to ground reality as also public policy concerns where increasing population has been an important aspect in designing appropriate poverty reduction strategies. 

Keeping the change in total population in mind, I propose an alternative measure where growth, inequality and population can be considered as within-group effects and this would be independent of the between-group effect on account of changes in population shares across subgroups.

Using the method to Indian data for 2004-05 and 2009-10, one observes the following:

• At the aggregate all India level, poverty reduced from 37.14 per cent to 29.77 per cent (-7.37 percentage points).
• Growth effect led to a reduction in poverty by 187 per cent (-13.77 percentage points); from this, 69 per cent is from rural (-9.45 percentage points) and 31 per cent is from urban (-4.32 percentage points).
• Inequality effect led to an increase in poverty by 2 per cent (0.13 percentage points); from this there was a reduction in poverty by 308 per cent in rural India (-0.40 percentage points) and an an increase in poverty by 408 per cent in urban India (0.53 percentage points).
• Population effect (total change) led to an increase in poverty by 88 per cent (6.51 percentage points); from this, 63 per cent is from rural India (4.13 percentage points) and 37 per cent is from urban India (2.39 percentage points). The percentage point increase in rural and urban in this as also in some other instances may not add up to the aggregate level due to rounding off.
• Total within-group effects (growth + inequality + total population change) led to a reduction in poverty by 97 per cent (-7.13 percentage points); from this, 80 per cent is from rural India (-5.72 percentage points) and 20 per cent is from urban India (-1.41 percentage points).
• Between-group effect (change in population shares between rural and urban regions) led to a reduction of poverty by 3 per cent (-0.25 percentage points), from this there was a reduction of poverty in rural India by 263 per cent (-0.63 percentage points) and an increase in poverty in urban India by 163 per cent (0.39 percentage points). By definition, this will have a negative impact on one region and a positive impact on the other region. The results imply that at an aggregate level rural to urban migration is likely to have contributed to reductions in poverty.
• From the aggregate all India level reduction (-7.37 percentage points), the contribution of rural India was 86 per cent (-6.35 percentage points) and that of urban India was 14 per cent (-1.02 percentage points).

Getting back to the method, we mention that the computation of the three within-group effects will depend on the choice of the base period and the sequence of computations. Further, the three within group effects as also the between group effects, as shown in our results for India, can all be mutually exclusive. These are discussed in the paper. 

The link to my open access paper Decomposing Poverty Change: Deciphering Change in Total Population and Beyond published in The Review of Income and Wealth. Also see the slide share explaining the method Decomposing Poverty Change: Within- and Between-group Effects that can be downloaded from my bepress page.

Some other recent blogs of mine on Poverty are:

New Poverty Estimate: Some Concerns

Poor in India: Region-wise Estimates

Group Differential Measure

This blog is meant for a lay explanation of the paper Group Differential for Attainment and Failure Indicators (see Enhanced HTMLview) in the Journal of International Development. It develops on the premise that if a society is progressing then it ought to reduce gaps across groups. In other words, as society progresses, it ought to be become increasingly inequity conscious.

For instance, if there exists a society, with equal proportion of female and male population, where literacy rate for females is 40 per cent and that for males is 50 per cent. After a decade, there is not much change in the population composition, but due to public policy interventions and the literacy rate for females increases to 50 per cent and that for males increases to 60 per cent. This increased attainment in literacy rates is commendable, but it is equally worrisome that the gap in literacy rates between the two groups remains the same at 10 percentage points. Thus, we would state that such an increase an attainment of literacy rate has fails a simple difference based level sensitivity. It is in this sense that this higher level of attainment is not commensurate with the society being increasingly inequity conscious.

A group differential measure that takes the simple difference of the measures for the two groups would fail this level sensitivity test.  

To address this, one could suggest, squaring of the literacy rates and then taking the difference and then after manipulation (for a comparable value that lies between 0 and 100) gives us a group differential measure that is {(50^2)-(40^2)}/100=9 percentage points in the first scenario and 11 percentage points in the second scenario. This satisfies level sensitivity, as it gives a lower value at the lower level of attainment. However, it fails to be policy sensitive at the lower level of attainment. In other words, when the actual gap between the literacy rates is 10 percentage points, our measure shows a 9 percentage point difference. Thus, it is possible that while this measure is not a representation of the actual percentage point difference, but it could lead to complacency at lower levels of attainment.

One possible way out is to take a ratio of the simple difference to the simple difference between maximum attainment and half of the attainment for female literacy rate and then manipulate to obtain a comparable value, that is {(50-40)/(100-20)}*100=12.5 percentage points in the first scenario and 13.3 percentage points for the second scenario.

Thus, we come up with a group differential measure that imposes greater inequity consciousness at higher levels of attainment and also is sensitive to policy implications. The discussion in the paper also shows that this approach also satisfies normalisation and montonocity properties. It also proposes an alternative measure that satisfy similar properties for failure indicators. The paper has examples using an attainment and a failure indicator that are relevant for the Millennium Development Goals (MDGs).

Other related papers

On measuring group differentials displayed by socio-economic indicators: an extension, Applied Economics Letters (2008)

On measuring group-differentials displayed by socio-economic indicators. Applied Economics Letters (2006)

Working Paper version of the Current Paper

Group Differential for attainment and failure

Poverty Estimates in India

Poverty Estimates in India: Old and New Methods, 2004-05 is the title of a new working paper published from the Indira Gandhi Institute of Development Research (IGIDR), Mumbai. It has been co-authored by Durgesh C. Pathak, a post-doctoral fellow whom I have been mentoring for the last two years, and myself. This paper is dedicated to the memory of Late Professor Suresh D. Tendulkar who passed away recently on 21 June 2011. Below I give excerpts that draw from the two quotations  that the paper begins with, the abstract, introduction and concluding remarks.

The poor are a part of necessary furniture of the earth, a sort of perpetual gymnasium where the rich can practice virtue when they are so inclined. - Francesco Guicciardini (Discorsi Politici)

But I, being poor, have only my dreams;
I have spread my dreams beneath your feet;
Tread softly because you tread on my dreams...
- W. B. Yeats

Abstract

This paper provides estimates of poverty and inequality across states as also for different sub-groups of population for 2004-05 by using the old and new methods of the Planning Commission. The new method is critically evaluated with the help of some existing literature and its limitations discussed with regard to doing away with calorie norm, use of median expenditure as a norm for health and education when the distribution is positively skewed, difficulty in reproducing results for earlier rounds acting as a constraint on comparisons, and using urban poverty ration of the old method as a starting point to decide a consumption basket. More importantly, it discusses the implications on financial transfers across states if the share of poor is only taken into account without accounting for an increase in the total number of poor. Despite these limitations, on grounds of parsimony and prudence the state-specific poverty lines suggested in the new method, as also in the old method, are used to calculate incidence, depth (intensity) and severity (inequality among poor) estimates of poverty for different sub-groups of population, viz., NSS regions, social groups and occupation groups.

Introduction

In India, the quinquennial rounds of national sample survey (NSS) of consumption expenditure have been instrumental in providing us with an estimation of head count ratio. The Report of the Task Force on Projections of Minimum Needs and Effective Consumption Demands (Government of India, 1979) looked into the age, sex and activity specific nutritional requirements and arrived at a per capita norm of 2400 calorie for rural and 2100 calorie for urban and based on this a monthly per capita expenditure (MPCE) of Rs.49.09 in rural and Rs.56.64 in urban was identified as the poverty line for 1973-74. This was updated to accommodate price changes over time. The Report of the Expert Group on Estimation of Proportion and Number of Poor (Government of India, 1993) proposed the use of independent poverty lines for each state and updating them by looking into the state specific changes in prices. This formed the basis for official estimates of poverty provided by the Planning Commission till recently (hereafter, old method).

Some of the criticism of this approach is that the updated prices may not represent the calories norm that they were initially pegged to,  that the calorie norms should change because of demographic shifts in age and sex and change in occupational patterns, that basic requirements like health, education, sanitation and housing are not included in the calculation of poverty line, that a reference period of 30 days may not be appropriate for low frequency items of consumption expenditure among others. These have been partly addressed in the Report of the Expert Group to Review the Methodology for Estimation of Poverty (Government of India, 2009) leading to a new set of poverty estimates for the year 2004-05 that have now been accepted by the Planning Commission (hereafter, new method).

The current exercise focuses on three points. First, it discusses critically the new methodology in the light of a brief review of some recent literature by various scholars. Second, it analyses the change in shares of poverty across states and union territories (hereafter, states) that will occur due to this shift. It also tries to briefly hint the possible repercussions of these changes on poverty reduction efforts in states.  Third, it provides estimates of proportion of poor (head count ratio or incidence of poverty), depth (poverty gap or intensity) and the severity (poverty gap squared or inequality among the poor) at various levels of disaggregation like states, NSS regions, social groups and occupational categories.
...

Concluding Remarks

The Planning Commission accepted the suggestions by an Expert Group that it had constituted leading to a new method for estimating poverty in India using NSS's consumption expenditure data for 2004-05. The new method replaces the uniform recall of 30 days for all consumption items to a mixed recall where consumption of five low frequency items were collected for the last year (365 days) and appropriately adjusted to get a monthly per capita expenditure. It also takes into consideration health and education needs that the old method had not incorporated in its calorie norm. While doing these, it also opened up a number of other issues.

First, it did away with the benchmarking of a poverty line with a calorie norm that the old method was based on. They did not let the calorie norm go away totally. A reference is made to an FAO (Food and Agriculture Organization) calorie norm being achievable around its poverty line, but then this norm is for light and sedentary activities that may not adequately capture the energy needs of the poor who put in hard labour.  Second, while factoring in health and education expenditure is a positive step, using median expenditure as a norm for a positively skewed expenditure distribution may not represent the actual requirement of a poor person.  Third, having done away with a calorie norm, it begins with the poverty ratio for urban India from the old method as given. Using this ratio on the mixed recall it generates a consumption basket at the aggregate level for urban India and then uses this to generate a poverty line for states around this basket. This means that instead of using state estimates to compute a weighted all India average, it begins with the latter. A bottom-up method is replaced with a top-down approach. Fourth, the computation of consumption basket requires use of data from other rounds of NSS as also from other sources. The whole procedure is quite cumbersome and replicating it for earlier rounds or even for thin rounds is difficult and in many cases not possible. This will also have implications on the usage of time series poverty trends in macro modelling.

From a policy perspective, the new method will lead to change in share of poor. If financial transfers across states do not account for an increase in the number of poor or have a budget constraint then this means that the poorer states would end up getting less.

Despite these limitations, on account of pragmatic considerations as also for parsimony and prudence, the state-specific poverty lines have been used for computation of poverty at various sub-groups. This has been attempted in this paper for NSS regions, social groups and occupation groups for both the old and new methods. The relatively higher incidence of poverty among scheduled tribes in rural areas and scheduled castes in urban areas for social groups and that of agricultural labourers and other labourers in rural areas and casual labourers in urban areas for occupation groups have been discussed.

Though they do not play any active role in poverty estimation, yet the poor have maximum stake in poverty analysis as they are at the receiving end. Thus, a move towards a bottom-up approach where the poor get involved in the understanding of vulnerability, particularly in the implementation of policies (including on identification of poor and poverty alleviation) so as to bring in greater accountability and transparency is called for . In its absence, every attempt to define and measure poverty is like treading on the dreams of poor. If poverty measure chosen is going to help them, at least some of these dreams would become a reality. Otherwise they dry like leaves fallen from trees.

For details see the paper, Poverty Estimates in India: Old and New Methods, 2004-05.

'People first' stressed on 20th Anniversary edition of HDR

People are the real wealth reiterates the 20th anniversary edition of Human Development Report (HDR) 2010. It also comes up with three new measures for discussing poverty and inequality.

Cover Page, HDR 2010, © HDR 2010, UNDP.

 

The first HDR in 1990 by United Nations (UN) started a new era in development thinking. It put people at the centre and includes the processes of enhancing their choices as well as improvement in their well-being. Human beings are the ends. It is for this that human development is considered to be different from the following approaches.

* Economic growth is a means and not an end of development. Moreover, high GDP growth does not necessarily translate to progress in human development. Global experience has shown that income and human development are not always perfect companions, where some countries display relatively high levels of human development for their income and vice versa.

* Theories of human capital formation and human resource development view human beings as means to increased income and wealth rather than as ends. These theories are concerned with human beings as inputs to increasing production;

* The human welfare approach looks at human beings as beneficiaries rather than participants in the development process;

* The basic needs approach concentrates on the bundle of goods and services that deprived population groups need - food, shelter, clothing, health care and water. It focuses on the provision of these goods and services rather than their implications on human choices.

It can however encompass the above and it is with this broad thinking that the HDR beyond an income-based measure to the human development index (HDI) that had health, education and standard of living as its components. Over time, the index for each component as well as the measure has evolved. Nevertheless, one important criticism of the HDI has been the linear aggregation of three components. The current report takes care of this by proposing a geometric mean. Another alternative, which has not been used in the report, is to calculate the shortfall from the ideal and take its inverse.

There are also some further departures in the way HDI has been calculated. The components of education are mean years of schooling for adults of 25 years and above and expected years of schooling for children of school going age (earlier they were literacy rate for adults and gross enrollment ratio for school children). For decent standard of living the 2010 report uses per capita gross national income in purchasing power parity dollars (earlier it was gross domestic product). Further, the global/expected maximum and minimum are taken from observations or estimations from the last forty years (1970-2010). With regard to education component where minimum can be zero, one is added to all observations to avoid problems in the calculation of a geometric mean.

In addition, there are three more measures in the report

* The Inequality-adjusted Human Development Index (IHDI) adjusts the Human Development Index (HDI) for inequality in distribution of each dimension across the population. The IHDI accounts for inequalities in HDI dimensions by “discounting” each dimension’s average value according to its level of inequality. The IHDI equals the HDI when there is no inequality across people but is less than the HDI as inequality rises. In this sense, the IHDI is the actual level of human development (accounting for this inequality), while the HDI can be viewed as an index of “potential” human development (or the maximum level of HDI) that could be achieved if there was no inequality. The “loss” in potential human development due to inequality is given by the difference between the HDI and the IHDI and can be expressed as a percentage.

* The Gender Inequality Index (GII) reflects women’s disadvantage in three dimensions—reproductive health, empowerment and the labour market—for as many countries as data of reasonable quality allow. The index shows the loss in human development due to inequality between female and male achievements in these dimensions. It ranges from 0, which indicates that women and men fare equally, to 1, which indicates that women fare as poorly as possible in all measured dimensions.

* The Multidimensional Poverty Index (MPI) identifies multiple deprivations at the individual level in health, education and standard of living. It uses micro data from household surveys, and—unlike the Inequality-adjusted Human Development Index—all the indicators needed to construct the measure must come from the same survey. Each person in a given household is classified as poor or nonpoor depending on the number of deprivations his or her household experiences. These data are then aggregated into the national measure of poverty.

Using the previous method of calculating HDI there has been substantial progress in the last 40 years for 135 countries for which comparable data are available, and what is more, this has been possible through diverse path ways. The top mover is Oman that invested heavily in education and public health. The next nine movers are China, Nepal, Indonesia, Saudi Arabia, Laos, Tunisia, South Korea, Algeria and Morocco. Incidentally, China is the only one that made the gains because of income as they had earlier made investments in education and public health. Some of the important low-income but substantial gainers are Ethiopia, Cambodia and Benin.

Across regions, the gainers were East Asia, primarily because of China and Indonesia and the Arab countries. The laggards were former Soviet Union and Sub-Saharan Africa and from these three in the former (Belarus, Ukraine and the Russian Federation) and six in the latter (the Democratic Republic of the Congo, Lesotho, South Africa, Swaziland, Zambia and Zimbabwe) showing reductions in life expectancy.

Despite the above-mentioned setbacks in life expectancy,

"The dominant trend in life expectancy globally is convergence, with average life spans in most poor countries getting increasingly close to those in developed countries. In income, though, the pattern remains one of divergence, with most rich countries getting steadily richer, while sustained growth eludes many poor countries."

"Some countries have suffered serious setbacks, particularly in health, sometimes erasing in a few years the gains accumulated over several decades. Economic growth has been extremely unequal, both in countries experiencing fast growth and in groups benefiting from national progress. And the gaps in human development across the world, while narrowing, remain huge."

In South Asia, Iran fares the best (#70); Sri Lanka (#91), Maldives (#107), India (#119) and Pakistan (125) are in the middle HDI countries, whereas Bangladesh (#129), Nepal (#138) and Afghanistan (#155) are among low HDI countries. Data are not available for Bhutan.

As per the current measure of HDI done for 169 countries, the top ten are Norway, Australia, New Zealand, the United States, Ireland, Lichtenstein, the Netherlands, Canada, Sweden and Germany and the bottom ten are Mali, Burkina Faso, Liberia, Chad, Guinea-Bissau, Mozambique, Burundi, Niger, the Democratic Republic of the Congo and Zimbabwe.

IHDI has been applied to 139 countries and the average loss in HDI is 22 per cent. There are variations across regions and dimensions as indicated in the figure/chart given below.

Chart depicting inequality adjusted human development index across regions. © HDR 2010, UNDP

GII, calculated for 138 countries, also varies across regions and dimensions, but the largest contributor to loss is reproductive health. The loss on account of lower empowerment to females is also relatively higher for South Asia and Arab states, as indicated in the figure/chart given below.

Chart depicting loss on account of gender inequality across region and by dimensions (labour market, empowerment and reproductive health). © HDR 2010, UNDP

MPI has been calculated for 104 developing countries using data from household surveys. This as also some other measures used in HDR 2010 has already been discussed in some Oxford Poverty and Human Development Initiative (OPHI) papers. There is also a very lively debate under Let's talk HD. There are variations that one could observe between income-based measures and MPI, as indicated in the figure/chart given below.

Comparing multidimensional and income poverty for selected countries. © HDR 2010, UNDP

(This write-up was first put up in Digital Journal, 5 November 2010, http://www.digitaljournal.com/article/299827.)