15
Notes on Inequality and Poverty Notes on Inequality and Poverty Surjit Surjit S S Bhalla Bhalla Prepared for NBER – Neemrana Conference, Jan 10-13, 2009

Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Notes on Inequality and PovertyNotes on Inequality and PovertySurjitSurjit S S BhallaBhalla

Prepared for NBER – NeemranaConference, Jan 10-13, 2009

Page 2: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Trends in Consumption Inequality Trends in Consumption Inequality –– Belief and EvidenceBelief and Evidence

BeliefWidespread belief that inequality has worsened, and worsened considerably, in India

Since 1980, India has enjoyed 28 years of high and sustained growth in per capita growth (average of 5.9 % per annum, with volatility of only 2.0 percent)

This is fertile ground for the Kuznets curve to have had its effect.

EvidenceTrends in nominal inequality confirm this expectation – after staying constant at around 30.3 during the first decade 1983-93, the Gini had increased by 16 plus percent to 35.3 in 2004/5.

Trends in real inequality, the real indicator, show only a marginal increase from 1983 – a Gini increase from 29.5 in 1983 to 30.8 in 2004/5, a 4.4 percent deterioration.

These trend changes are put in perspective by noting what a pure difference in measurement (estimates for some consumer durables are on a 365 day basis (the mixed recall method) rather than a 30 day basis (uniform recall method).

This pure difference was 7.2 % in 1983 and 1993 and 4.6 percent in 2004/5. In other words, the entire change in inequality over 21 years is less than the definitional difference in inequality in any given year!

Page 3: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

NSS Consumption Inequality (NSS Consumption Inequality (GiniGini) in India 1983) in India 1983--2004/052004/05

Notes : 1. Uniform recall method: all consumption on a 30 day basis; mixed-recall : some items on a 365 day basis

2. Adjusted to NA means that NSS consumption is matched to equal national accounts consumption; for method of adjustment, see text.

Year 1983 1993/94 1999/00 2004/05 % change 1983-2004/05

Measure, Nominal

Uniform Recall 32.6 32.7 36.8 12.7

Mixed Recall 30.4 30.3 32.3 35.1 15.5

Adjusted to NA 36.0 37.8 36.5 43.4 20.7

Measure, Real

Uniform Recall 31.9 30.4 32.8 2.6

Mixed Recall 29.5 27.8 29.0 30.8 4.6

Adjusted to NA 35.4 35.5 33.2 39.8 12.6

Page 4: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Trends in Income Inequality Trends in Income Inequality –– EvidenceEvidence

NSS does not collect data on income distribution – but NCAER does.

Indeed it has been doing so since the mid-1960s.

Nominal income inequality increases between 1975 and 1994/95 (Gini=39.3 & 43.3, respectively), and increases to 46.6 in 2004/5

Real income inequality increases from 40.0 in 1994/5 to 44.0 in 2004/5

Table on Income & Consumption Inequality(Gini) NCAER –Jan 9,2009

Year 1975/76 1994/95 2004/05

Income Inequality

Nominal 39.3 43.3 46.6

Real 40.0 44.0

Consumption Inequality

Nominal 35.1

Real 31.8

Page 5: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

MMissing consumption in NSS data: how serious?issing consumption in NSS data: how serious?

Unfortunate trend in NSS consumption data of capturing less and less of consumption according to National Accounts data

The Survey to National Accounts ratio (mean consumption NSS survey/mean consumption national accounts) has been steadily declining from a near parity level of 97 percent in 1972 to 49.5 percent in the latest 2004/5 survey. This steep drop has implications for inequality and poverty measurement.

The present ratio, 49.5 percent, is among the 15 lowest ratios among all countries with household surveys in the post-war period. The five worst are Papua New Guinea, 1996, 18.7 percent; Peru, 1997, 22.8 percent; Albania, 2002, 29.3 percent; Ecuador, 1988, 40.4 percent and Ukraine, 1996, 40.4 percent.

Page 6: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

MMissing consumption in NSS data: implications for poverty? issing consumption in NSS data: implications for poverty?

Some idea of the impact on growth rates in real consumption [an essential ingredient in the analysis of poverty] can be gleaned from the table on NSS per capita growth and NAS per capita growth.

Notes : All values in 2004/5 rural prices and in Rs. per capita per month.

Year 1983 1993/94 1999/00 2004/05 % change 1983-2004/05

Measure, Real

Uniform Recall 471.7 521.6 604.4 28.1

Mixed Recall 467.9 530.9 575.3 628.9 34.4

Adjusted to NA 753.7 856.2 1022.3 1228.1 62.9

Page 7: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Survey to National Accounts Ratio in IndiaSurvey to National Accounts Ratio in India

Notes : The survey and NA figures are in current Rs. (000) per person per year.

The current NA figures are taken from the NA estimates prevailing at the time of the survey.

Year Survey National Accounts Survey/NA Ratio

1972 1.6 1.6 96.9

1977 2.4 2.6 91.3

1983 4.1 5.0 80.7

1987 6.0 8.9 67.0

1993 10.8 17.5 61.8

1999 19.3 34.8 55.5

2004 23.0 46.5 49.5

Page 8: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Adjustment for missing consumption dataAdjustment for missing consumption data

The NSS data can be taken as accurate if the missed consumption is proportionally distributed; Is it?The null hypothesis is that the missing consumption is mostly among the rich; in particular, scholars have pointed to the FINSIM category (spell out) as a major source of bias in the measurement of inequality.But FINSIM only accounted for less than 2 percent of total consumption in 2004/5, and less than 1 percent in 1993/94.More likely a source of bias is consumption of housing which is mostly canvassed for urban households; in 1993/94 only 24 percent reported any rental value of housing; in 2004/5 the fraction was considerably lower at 15 percent. For the rest of the population, housing rental consumption is imputed to be zero by the NSS.The survey procedure on housing should impart a downward bias to inequality and it does; imputing a rental value for all households (based on a regression relating fraction of expenditure on housing to total expenditure) results in the Gini increasing by 1 % in 1983 and 1993, 3 % in 1999 and 2004/5. Two major items in food and non-food category [spending in hotels and restaurants and spending on banking, insurance, and financial services] are not adequately covered in the NSS surveys. Allocating such expenditures on the basis of actual food and non-food expenditures yields almost zero change in inequality indices.

Page 9: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Adjusting NSS consumption to NAS consumptionAdjusting NSS consumption to NAS consumption

A suggested method for matching NSS data to the NAS data [this assumes that the latter is more accurate] is to break up total consumption into 12 food and 15 non-food categories.

For each category e.g. cereals and pulses, fruits and vegetables, clothing and footwear, education and medical care etc. , the survey mean is matched to the national accounts mean.

For example, in 2004/5, NSS per capita per month consumption of cereals and pulses was Rs. 104; the NAS was 128. Hence, each individual household’s consumption is multiplied by a factor of 1.23[the ratio of 128 and 104]. For education and medical care, the ratio in 2004/5 was 1.63. Note that this procedure allocates the missing consumption on a proportionate basis to actual consumption; if actual consumption was zero, so would the adjusted consumption be; if actual consumption is small say Rs. 2 per month, the adjusted consumption would be Rs. 3.26 for education and medical care.

Page 10: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

On NSS/NAS fractionsOn NSS/NAS fractions

The 27? consumption items are combined into five categories (essential food, non-essential food, essential non-food, education and medical care, and non-essential non-food).The fractions captured by the NSS in 1983, 1993/94 and 2004/5 are as follows, in percent.

Year 1983 1993/94 1999/00 2004/05

Essential food 83.3 87.3 77.4 79.0

Non-Essential food 60.0 59.9 51.5 48.2

Essential non-food 47.3 46.8 58.0 49.8

Education & Medical Care 68.2 92.0 63.1 54.1

Non-Essential non-food 45.7 42.8 36.6 31.8

Page 11: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Inequality trends with adjusted dataInequality trends with adjusted data

Two major results emerge with adjusted data. First, that inequality at any point in time is about 20 percent higher with adjusted data. Second, there is no trend in adjusted inequality between 1983 and 1999/00; but there is a significant 12 percent increase in inequality between 1983 and 2004/5. Third, that this increase is about equal to the difference in inequality at any point in time between two definitions of consumption (uniform and mixed recall).

Collecting all the results, inequality in India has increased by about 10 percent since 1983. This change in inequality (only inequality as measured by survey data are comparable) is considerably less than that of China; a noteworthy result since some scholars (Sen, Bardhan) believe that inequality change in the Communist dictatorship and democracy are comparable.

Page 12: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Poverty Reduction (Head Count Ratio, %)Poverty Reduction (Head Count Ratio, %)

Depending on the recall period (all items 30 days or 30/365 days for some items – uniform and mixed recall, respectively) the level of poverty at any point in time can be calculated. In 1983, the difference in the two estimates is around 3 percentage points; in 2004/5, the difference is about double or 6 percentage points. Over the period 1983-2004, the uniform recall method shows an 18 percentage point decline; the mixed recall method a 21 pptdecline.

These declines are based on an average per capita consumption growth rate of 1.15 percent per year (uniform recall) and 1.37 percent (mixed recall). The (adjusted) log growth rate in per capita consumption is 2.27 percent a year, full 1 percentage point a year lower.

Year 1983 1993/94 1999/00 2004/05 % change 1983-2004/05

Measure, Nominal

Uniform Recall 44.3 34.6 24.7 26.5 -40.2

Mixed Recall 41.6 29.3 24.7 20.8 -50.0

Adjusted to NA 17.1 10.5 3.4 3 -82.5

Page 13: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Elasticity of poverty reductionElasticity of poverty reduction

Unfortunately, there is a mix-up in the literature on poverty reduction – the growth in NA consumption is compared with the reduction in survey poverty. When doing this, India’s poverty performance seems mediocre, and growth not pro-poor. Further, researchers derive the elasticity of poverty reduction based on growth in per capita consumption and decline in the head-count ratio of poverty. This is incorrect. The correct elasticity (methodology detailed in Bhalla – Imagine…) is to estimate the elasticity based on the distribution of consumption in the vicinity of the poverty line. This distribution is independent of overall inequality.

For example, the elasticity of poverty reduction based on adjusted consumption data in India in 2004/5 is only 0.1 compared to an elasticity based on measured survey data of 0.9! At 0.9, the Indian poverty reduction elasticity is among the highest in the world in 2004/5; in contrast, China had an elasticity only one-third that of India.

Year 1983 1993/94 1999/00 2004/05 % change 1983-2004/05

Measure, Nominal

Uniform Recall 0.9 0.9 0.9 0.9 0.0

Mixed Recall 0.8 1 0.9 0.9 12.5

Adjusted to NA 0.4 0.5 0.2 0.2 -50.0

Page 14: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Level of poverty in IndiaLevel of poverty in India

If survey data are to be believed, then there has been near zero change in inequality during the high growth period of 1983 to 2004/5. And poverty reduction has been impressive, but the level of poverty is still very high in India – approximately 20 percent. This level is based on a dollar a day poverty line (the Indian poverty line and the World Bank $1.08 a day poverty line, 1993 PPP prices are identical till the second decimal place!). At 20 percent poverty, India’s poverty level is among the highest in the developing world –with Pakistan, Bangladesh both lower.

Page 15: Notes on Inequality and Poverty Surjit S Bhallademo.ncaer.org/.../EventDetails/E10Jan2009/neemrana-presentation… · Notes on Inequality and Poverty Surjit S Bhalla Prepared for

Time to raise the poverty lineTime to raise the poverty line

� Poverty in India is changing from absolute to relative� The NSS consumption data has a lot of problems� Time to raise the poverty line by about 50 percent, and

go back to the old method of adjusting survey consumption by NA consumption

� Alternatively, raise the poverty line and keep the S/NA ratio fixed at the 1993/1994 level of 61.8 percent; this would mean that the growth rates in consumption are not understated; the “error” from a declining S/NA ratio will no longer affect the results.