REMOVING POVERTY THROUGH SPURIOUS STATISTICS
A REPORT in The Hindu (May 6) suggests that the Planning Commission is at its old game again which is to “remove” poverty in the country through a sleight-of-hand, by taking recourse to spurious statistics. Consumption expenditure estimates from National Accounts Statistics show a far higher level than those arrived at on the basis of the National Sample Survey. Since poverty is estimated on the basis of the NSS figures, it is being argued that poverty is being overestimated, as the actual So, the decline in poverty is claimed to be far greater than even the Planning Commission’s existing estimates, based a ludicrously low and progressively declining poverty line, suggest. consumption is higher than what the NSS shows. Let us examine this argument, which is bound to be presented with great stridency in the months to come. The macro-aggregates in any economy must exhibit certain relationships among themselves which follow from their very definitions. These relations are called “identities” by economists. For instance the value of the gross domestic product (GDP) must be identically equal to the expenditure upon it (counting retention of a product by its producer also as expenditure upon it). This expenditure can be of different kinds. So, the identity, that GDP at market prices must equal expenditure upon it at market prices, is sometimes written as: GDP = Consumption + Gross Investment + Government Expenditure + Net Exports of Goods and Services. The National Accounts statistics provide figures for each of these in such a way that the identity is shown to hold for the economy as a whole. The problem is that these items, other than consumption, are estimated separately and independently, and consumption is estimated as a residue. For instance, GDP is estimated by putting together the estimates of output, arrived at in various ways, of a whole lot of individual sectors. A lot of these ways are no better than pure guess work. But a GDP figure for the country has to be there; so somehow or the other a figure is produced. Balance of payments data (for estimating net exports) are a little more firm, as are government data (estimated from budgets); but investment estimates are arrived at “indirectly” by seeing how much cement, how much steel and how many machines etc. have been used over a certain period. Some fixed ratios are assumed (so much cement per unit of construction) and on that basis investment figures are arrived at. And then consumption figures are derived from all these as a residual. It follows that any error in estimating the GDP, or investment or any other component of expenditure would affect the consumption estimate. And with regard to both the GDP and investment (and even to an extent trade), the estimates are notoriously unreliable. Indeed, statisticians have argued that as much as 40 percent of the total GDP estimate for any year is based on “indirect” or “fictitious” statistics! Now, suppose the GDP is overestimated by 20 percent, i.e. instead of 100 which is “actual”, it is estimated to be 120. And suppose that the expenditure items are correctly estimated: investment at 35, government expenditure at 20, and net exports at -3 (since India typically has a deficit). Then instead of a correct consumption figure of 48, we get a spuriously inflated consumption figure of 68. In other words, any overestimation of the GDP, other things remaining unchanged, shows itself as an overestimation of consumption. What is more, while the extent of overestimation of GDP was 20 percent, the overestimation of consumption on account of it comes to 5/12 or 42 percent! In percentage terms, the consumption gets magnified to a far greater extent than the GDP estimate! While consumption estimates of the National Accounts Statistics are thus derived from the GDP and other estimates by using the “National Income Identities”, the NSS consumption estimates are directly arrived at. They are based directly on household consumption surveys conducted on an elaborately chosen representative sample of households. They are the largest regular sample survey anywhere in the world and use the most sophisticated statistical techniques in choosing the sample. They too no doubt are subject to error. An obvious one arises from the carelessness or the lackadaisical attitude of the field investigators. In addition, in any such sample survey, the top brackets can never be adequately represented. Since they are very exceptional in the sense of concentrating enormous amounts of wealth, income and expenditure in their hands (the top 1 percent of the world’s population for instance controls half the world’s total wealth), they tend to get under-represented in any sample of the population as a whole. Therefore if there is any underestimation involved in NSS data then it concerns the topmost households. For a long time however the government of India was pretending otherwise. The consumption figures from the National Accounts Statistics being higher than from the NSS is not a new phenomenon; it has always been the case. What the government used to do earlier is to take the total consumption expenditure from the NAS, which is the higher figure, and distribute it across size-classes in the same ratio as in the NSS expenditure figures, and then take the outcome of that exercise as the true picture. Not surprisingly, this, by raising all size-classes’ expenditures proportionately, tended to underestimate both the degree of consumption inequality (and hence income inequality) and also the extent of absolute poverty. Instead of either ignoring the excess of NAS consumption expenditure over the NSS figure altogether, or allocating this excess to the top size-classes where there was some cause for believing the NSS figure to be an underestimate, the government adopted a method that was tailor-made for showing a lower level of inequality and poverty. In 1993-94, an expert group suggested that this practice of pro rata distribution across all size classes of the excess of NAS over NSS consumption expenditure figure be discontinued; and that since the two were separately arrived at, they should be kept separate and not be mixed up. The poverty estimates, it argued, should be made entirely on the basis of NSS data. This recommendation was accepted and until now as a result, the poverty and consumption-inequality estimates are made entirely on the basis of the NSS data. It is this which the Planning Commission is now once more attempting to go back on. The excuse reportedly is that the private consumption expenditure figure arrived at on the basis of the national income identities in NAS is twice the corresponding figure arrived at by the NSS. But the first point that should have occurred to the Planning Commission owing to this big difference, is that the GDP figures are gross overestimates. And since the extent of the NAS-NSS difference has widened over time (it was about 30 percent in the 1970s), it follows that the growth rate figures are grossly exaggerated. The economy in short has not been growing as rapidly as successive governments have been claiming. Instead what is being suggested is that the growth rate figures are fine, the GDP estimates are fine, but the NSS figures, which are based on the direct compilation of data, the largest such compilation in the world, from a meticulously scientifically designed sample of the entire population of the country, is all wrong. Here in short we have a classic example of ideology triumphing over science. And now, no doubt the same ideological prejudices would dictate a recalculation of the degree of poverty to show how it has been virtually eliminated! But it is not only a matter of comparing NSS with NAS data. We have more direct evidence on poverty. The poverty line in India is defined as that level of expenditure in the base year at which a rural household accessed 2200 calories person per day and an urban household 2100 calories. And then this base year figure is brought up to date by taking a consumer price index. Let us ignore the question of whether the price index is an appropriate one, which has been a central issue of controversy. Let us focus on something else instead. In taking these figures in the base year, it is automatically implicitly assumed that all rural households having more than the poverty-line expenditure would be accessing more than 2200 calories, and all household having less would be accessing less; and likewise for the urban areas. In short, not only is poverty officially defined with respect to a calorie norm, but a monotonic relationship between expenditure level and calorie intake is implicitly assumed. Now, if poverty was actually declining then we must find at the macro-level a rise in per capita foodgrain consumption, since foodgrains ingested directly or indirectly (via processed foods and animal products into which they enter as feedgrains), are the main source of calories. Putting it differently, if we find that the per capita foodgrain consumption is declining over time, then we can infer with confidence that poverty cannot be going down. (It should be remembered here that the foodgrain required indirectly to produce a calorie, via meat or other animal products, is even higher than what is required directly). We do not however have figures for private consumption of foodgrains. The only information we have relates to foodgrain availability defined as net output minus net additions to stocks plus net imports; and even here we have no knowledge of net additions to private stocks. So if we take net output plus net imports minus net additions to government stocks, we get an approximation to foodgrain availability which in turn is the closest approximation to foodgrain consumption that we can have. Let us look at this variable in a historical perspective to get an idea of poverty trends in the country. Over the quinquennium 1897-1902, the average annual net per capita foodgrain availability in “British India” was 199 kilogrammes. It declined to 148.5 kilogrammes during 1939-44, and further to 136.8 in 1945-46. The post-independence dirigiste regime pushed it up to 177 kilogrammes for the country as a whole for the triennium ending 1991-92. After the introduction of neo-liberal “reforms” however there has been a fall: to 174.3 kilogrammes for the triennium ending 1994-95, 174.2 kilogrammes for the triennium ending 1997-98, and 155.7 kilogrammes for the triennium ending 2002-03, which was lower than the average figure (159.3) for 1933-38 . The figure for the triennium ending 2006-07 was in fact 159.88, almost the same as for 1933-38, and for the triennium ending 2011-12, 163 kilogrammes. The net per capita foodgrain availability in the first quinquennium of this century in short has been lower than in “British India” on the eve of the war, and not much higher thereafter. To claim that poverty has declined under these conditions defies reason.