A Critical Look at Some Propositions on Consumption and Poverty Utsa Patnaik
A comment to the analysis and discussion in Angus Deaton and Jean Dreze, “Food and Nutrition in India: Facts and Interpretations” (EPW, 14 February 2009).
Utsa Patnaik ([email protected]
) is at the Centre for Economic Studies and Planning, Jawaharlal Nehru University, New Delhi.
1 Consumer Price Index Is Not a Cost of Living Index
eaton and Dreze (2009) have argued that the NSS consumption data show rising per capita real expenditure in both rural and urban India, over the same period that per capita calorie intake has been declining. As they put it (p 46), “…the decline in per capita calorie consumption has been accompanied by an increase in real average household per capita expenditure”. They have posed as a “puzzle” the steady “downward drift” in calorie intake for all spending classes despite this “rise” in real per capita expenditure, saying “per capita calorie consumption at a given level of per capita expenditure has steadily declined over the last 20 years” (ibid, p 46, all emphases in original). What do they mean by “real” expenditure and by a “given level of expenditure”? From the note to their Table 2 headed “Real Mean Per Capita Expenditure on All Goods and on Food”, they are deflating nominal rural and urban spending by the CPIAL and CPIIW, respectively, which is the routine procedure that economists follow. This is the same as using the implicit deflator from successive official poverty lines, since these poverty lines have been obtained by updating 1973-74 rural and urban poverty lines using the CPIAL and the CPIIW. However, can economists confine themselves to routine procedures alone, and continue to apply uncritically only the conventionally accepted methods of deflating current to constant values, when these methods have been shown already to be increasingly inadequate? Using consumer price indices over long periods of time has led to absurd results by way of grossly underestimating the required expenditure for maintaining even a modest
consumption standard. The particular levels of money expenditure in 1973-74 which allowed official nutrition norms to be met, when updated by price indices to 2004-05 permit only much lower nutritional levels, sharply lower in many states by 600 to 1,000 calories daily . The questions precisely are, how “real” are these authors’ estimates of “real per capita expenditure” calculated over two decades using conventional indices, and to what extent is their “given level of per capita expenditure” actually given? Economists generally tend to treat the consumer price index as equivalent to a cost-of-living index, and term as “real expenditure”, spending levels deflated by the consumer price index. It is clear by now that such an identification is untenable for time periods in excess of five years or so, for we have been getting seriously misleading results by treating price indices and cost of living indices as the same. Money incomes and expenditure should be rising enough for working people to permit both the maintenance of the minimum nutritional standards as well as reasonable spending on health and utilities, and thus should be rising in tandem with the cost of living in order to maintain real purchasing power. But clearly that is not happening, money expenditures are not growing in tandem with the cost of living and this is reflected in absolute decline in the consumption of necessities like foodgrains and cloth. Consumer price indices are increasingly understating actual change in the cost of living, since the nominal spending level which allowed a modest nutrition norm of 2,100 calories per capita to be accessed three decades ago, when consumer price index adjusted, now allows only 1,400 to 1,700 calorie intake in most states. The actual nutrition-invariant cost of living is more than double the price-index adjusted poverty lines. For urban employees, the consumer price index alone has never served as the sole basis for determining the cost of living for fixing salaries. The author joined her university as an associate professor in 1973, the base year for official poverty estimates, at a gross pre-tax salary of about Rs 1,000 per month. The Consumer
february 6, 2010 vol xlv no 6 EPW Economic & Political Weekly
Price Index for Urban Non-manual Employees rose 10.55 times by 2004-05, so applying this index the salary for an entrant into the same post should be Rs 10,550 per month. But no urban middle class professional could possibly maintain herself on such a sum, which the University Grants Commission considers adequate only for the senior research fellowship awarded to young doctoral scholars. The actual entrylevel salary for the same post in New Delhi by 2005 was more than double the CPIUNME index adjusted base year salary and close to the multiplier value of 2.6 yielded by taking the ratio between the 2,100 calories poverty line and the official poverty line in 2004-05.
2 Alternative Deflators to Calculate Real Expenditure Since the consumer price indices have not been capturing the actual change in the cost of living, and deflating expenditure series over a decade or more by these indices no longer gives a correct approximation to “real expenditure”, we need alternative deflators to convert nominal spending to real spending. The official poverty lines (OPL) measure the cost of a fixed base-year basket updated by consumer price indices, so one set of implicit deflators is given by the change in the OPLs at successive points of time, representing the change in the consumer price indices. An alternative set of implicit deflators is given by taking the nutrition-invariant or direct poverty lines (DPLs) at successive points of time at the norm and at lower levels. “Real expenditure” is found to be highly sensitive to the specific deflator used, which not merely affects the values but also alters the direction of change. The interesting results of applying the alternative deflators to average nominal MPCE during 1983 to 2004-05, normalising for 1993-94, are shown in Table 1. Applying the OPL implicit deflator gives a rise in average “real expenditure” by 25% in rural India from Rs 259 to Rs 323 and by 40% in urban India from Rs 397 to Rs 557. This is the same finding as Deaton and Dreze (2009). But applying the implicit deflator from the DPLs at norm levels, alters the direction of change to the opposite: rural real spending declines by 25.5% from Rs 306 to Rs 228 and
declines by 5.6% in urban India from Rs 441 to Rs 416. Table 1: Nominal Average Monthly Per Capita Expenditure Deflated Using Alternative Implicit Deflators from Official and Direct Poverty Lines Implicit Deflators
1 Rural a < 2,400DPL (norm)
b < 2,200 DPL
c < 1,800 DPL
2 Urban a < 2,100 DPL (norm)
b < 2,000 DPL
c < 1,800 DPL
3 Nominal mean MPCE, Rs Rural 112.5 281.4 558.8 Urban
164.0 458.3 1052.4
4 Real mean MPCE, Rs (base 1993-94) Rural a Applying < 2,400 DPL deflator 305.7 281.4 227.7 b Applying <2,200 DPL deflator 300.8 281.4 254.6 c Applying<1,800 deflator
282.7 281.4 271.3
d Applying OPL deflator
259.2 281.4 323.4
5 Urban a Applying <2,100 DPL deflator 440.9 458.3 416.0 b Applying<2,000 DPL deflator 429.3 458.3 444.4 c Applying <1,800 DPL deflator 403.0 458.3 488.6 d Applying OPL deflator
397.1 458.3 556.8
Source: Patnaik (2007) gives the official and the direct poverty lines at both norm levels and at 1,800 calories level. Direct poverty lines for other nutritional levels are from author’s charts of the g relation.
Implicit deflators have been obtained also for two additional, below norm nutritional levels. Over the two decades from 1983 to 2004-05 there is a decline in rural real spending with implicit deflators for every nutritional level. Urban India shows spending decline with the norm level deflator while the lower level deflators give rise in spending. Note that the difference between the OPL deflator and the
1,800 calories DPL deflator is not very large but it is enough to alter the rise in rural real spending to a decline. The performance using the DPL deflators, is very different during the period 19831993-94 compared to the subsequent economic reforms decade. Urban real MPCE shows unambiguous rise in the pre-reform decade for all nutrition level deflators. This improvement in real spending saw a small rise in average calorie intake. Rural real MPCE shows decline by 7.9% and 6.4% using the norm level and next lower deflators, while it remains virtually constant with the lowest level deflator. But over the subsequent economic reforms decade, which saw income deflating macroeconomic policies, urban real spending declines by 9.2% and 3% for the norm and next lower level deflators while only the lowest level deflator gives a rise by 6.6%. Rural India shows much larger decline in real spending by 19.1% and 9.5% for the first two DPL deflators while even the lowest level deflator shows decline. All these results are very different from the OPL deflator showing improvement in real expenditure over both periods and in both rural and urban India. What of the fractile specific behaviour of “real expenditure” over time? As Table 2 shows for urban India, the OPL deflator gives rising real spending for all spending classes over 1993-94 to 2004-05 with the average rise at 21.5%. But with the DPL deflator at norm level, the finding is the opposite. There is a 9.2% average decline in real spending, and 11 of the 12 spending classes containing 95.1% of all persons
Table 2: 2004-05 Expenditure Deflated to 1993-94 Using Different Deflators by MPCE Classes (All-India Urban) MPCE Persons 2004-05 1993-04 Class 1993 2004 Current Current (%) (%) (Rs) (Rs) 1 2 3 4 5
2004-05 DPL Index Deflated 6
2004-05 OPL Index Deflated 7
2 3 4
(Col 6- Col 5) Calorie Intake Change % Change (Rs) 8
For real spending col 6 uses 2,100 calories DPL Index = 253, and col 7 uses OPL Index =189 both with base 1993-94 = 100. The last column gives the relative change in calorie intake in 2004-05 over 1993-94 for normalised calorie intake viz, deducting 1,000 calories from both sets.
Economic & Political Weekly EPW february 6, 2010 vol xlv no 6
show decline while only the top 4.9% of persons show a rise. There is a clear association between the decline in real expenditure and decline in calorie intake for 10 classes. The anomalies are the lowest class which shows a small rise in calorie intake despite fall in real expenditure1 and the richest class which shows fall in calorie intake albeit from very high levels, with a rise in real spending. On average there is a 4.9% calorie intake decline with 9.2% real spending decline. Clearly, “real” per capita spending shows rise without exception for every period in the Deaton and Dreze (2009, Table 2) exercise, and in both rural and urban areas, solely owing to the specific price indices they use, which have been used always for updating official poverty lines, and have cumulatively underestimated the actual rise in the cost of living. The “downward drift” of calorie intake at a “given real expenditure” is the result of using these specific indices to the exclusion of any other. If the alternative deflators from the nutritioninvariant poverty lines are used for comparing what Deaton and Dreze (2009) have called the “calorie Engel curves”,2 the pattern would be very different from the one in their Figure 1. The curves for different years would not drift downwards but would cluster closely together. Thus, there is no puzzle of declining per capita calorie intake with consistently rising per capita “real spending” since such behaviour does not exist in reality: applying reasonable alternative deflators shows decline in real spending, which turns out to be particularly large in the economic reforms decade. This is consistent with macroeconomic policies over this decade which were employment and income reducing in the material productive sectors while favouring tertiary sector growth. Even applying the standard price index deflators, the consumption data show not only falling real spending on food – as Deaton and Dreze have correctly pointed out – but also falling real spending on cloth. The exception was urban India between 1983 and 1993-94 when per capita real expenditure rose using the alternative deflators, and average calorie intake far from declining also registered a marginal rise. Indian consumers are neither abnormal nor irrational, and poor consumers who
are already undernourished and inade- Table 3: Percentages of the Population Living in Households with Per Capita Calorie Consumption quately clothed, do not further cut back on Below 2,100 Urban and 2,400 Rural basic necessities like food and cloth unless (abstracted from Deaton and Dreze 2009 Table 5) Year Round Rural Urban All-India they are obliged to do so by falling pur1983 38 66.1 60.5 64.8 chasing power. The alleged puzzle in con- 1993-94 50 71.1 58.1 67.8 sumer behaviour arises from ignoring every 2004-05 61 79.8 63.9 75.8 bit of adverse evidence such as rising unemployment rates, falling foodgrains (given in briefer form as Table 3 in this availability per capita, stagnant mass in- article) only has a note saying “Authors’ comes in the material productive sectors calculations based on NSS data”. The quesespecially the primary sector, loss of tion is precisely how the authors have calowned land and livestock by the majority culated, which remains unstated, making of peasants – all of which indicate falling their procedure non-transparent. We find purchasing power. Relying exclusively on that most of their numbers for other years increasingly faulty tools, the consumer are also off the mark to some degree but price indices, for deflating Table 4: Distribution of Persons, Average MPCE and Average Nutritional Intake current spending values by MPCE Classes, All-India Rural, 61st Round 2004-05 (U-30) and Distribution of to constant values pro- Persons and Average Expenditure by MPCE Classes 61st to 63rd Round 2006-07 (MRP) duces misleading results. MPCE % of Cumulative % Average Average Cumulative % of Persons Persons of Persons MPCE Rs Calorie Intake (Current Values) My argument is not that Class URP URP URP URP MRP MRP MRP these conventional tools 2004-05 2004-05 2004-05 2004-05 2004-05 2005-06 2006-07 2 3 4 5 6 7 8 should be rejected, but 1 0- 235 4.8 4.8 199.53 1,376 2.8 2.0 1.2 that they should be used 235-270 5.1 9.9 253.80 1,575 6.1 4.5 3.2 critically, never be treated 270-320 9.9 19.8 296.64 1,679 14.4 11.4 7.5 as cost of living indices, 320-365 10.5 30.3 342.40 1,800 23.9 18.7 12.7 applied only over short 365-410 10.2 40.5 387.72 1,885 33.9 27.9 20.8 periods of four years 410-455 9.4 49.9 432.06 1,962 43.6 36.5 29.1 at most between large 455-510 9.9 59.8 481.55 2,042 54.4 46.6 39.0 sample surveys, and sup- 510-580 10.2 70.0 543.25 2,158 65.3 58.9 50.3 80.4 630.40 2,290 77.5 72.3 64.9 plemented by alternative 580-690 10.4 9.8 90.2 775.00 2,380 88.8 86.1 80.3 deflators. However I have 690-890 890-1155 4.9 95.1 999.94 2,568 94.7 93.2 90.6 no illusion that any 1155 ++ 4.9 100 1,956.57 3,018 100.0 100.0 100.0 established academic will 100.0 558.78 2,047 ever think out of the All Note that in Report 508 perhaps owing to rounding, the percentage of persons by MPCE box, and economists in classes totals to 100.2. We have adjusted for this by assuming the last two classes to contain 4.9% of persons each and not 5 as stated. particular tend to be Source: NSS Report 508, Level and Pattern of Consumer Expenditure, 2004-05 Statement 2R p 46, Appendix Tables 1R and 5R. Report 513, Nutritional Intake in India 2004-05, more deeply attached to on Report No 527. Household Consumer Expenditure 2006-07, Table, p 1. their standard concepts and tools however faulty these have the largest divergence from the data is in proved to be than they are to their spouses. their rural poverty estimate for 2004-05, which is 79.8%, 7 percentage points lower 3 Problem of Deaton and than the correct 86.7%. So let us consider Dreze Estimates this, reproducing the relevant information Deaton and Dreze (2009, Table 5) have from NSS Reports for the reader’s ready given estimates of the percentage of per- reference in Table 4. sons below nutrition norms which do not It was earlier shown (Patnaik 2007) that seem to square with the basic NSS data. direct head count rural poverty percentages Their rural estimates for 1993-94 and changed little over the period 1973-74 to 2004-05 are 71.1% and 79.8% of the per- 1993-94. The variation is plus/minus 3% sons falling below 2,400 calories and or so, around 72% below 2,400 calories inurban estimates are 58.1% and 63.9% of take. This relatively stable picture changed persons below 2,100 calories. They have over 1993-94 to 2004-05 with the rural not mentioned the poverty lines at which direct poverty line rising from Rs 326 to they obtain these percentages, and neither Rs 800 and the percentage of persons below the data for the ogives they use, nor for these lines rising by an unprecedented the g relation are given. Their Table 5 12 points to reach an all-time high of 87%. february 6, 2010 vol xlv no 6 EPW Economic & Political Weekly
average plus half the difference of this with the next class average of 2,380 80 calories). Even without plotting the graphs it is crystal 60 clear that reaching 2,400 calories intake requires 40 spending at a level just over half way between 20 Rs 690 and Rs 890, and the 0 poverty percentage has to 200 400 600 800 1000 1200 MPCE, Rs lie above 80.4 and include Source: Table 4 col 1 upper end values and col 3. more than half of the 9.8% of persons falling in this Chart 1b: Daily Per Capita Calorie Intake by MPCE, All-India Rural, 2004-05 URP 3500 class. Even assuming the ogive is linear at this 3000 stretch which it is not. It has to be over 80.4+4.9 = 2500 85.3% and cannot possibly be 79.8%. 2000 From Chart 1b, 2,400 calories can be accessed at 1500 Rs 800, so this is the direct 1000 poverty line. From Chart 1a, 0 500 1000 1500 2000 MPCE, Rs 86.7 or rounding to the Source: Table 4 col 4 and col 5, URP distribution. This is the g relation referred to in nearest whole number, the text. 87% of persons in houseGiven the data of Table 4 from the cited holds, failed to reach this level. Deaton NSS Reports, used for the relations depicted and Dreze’s 79.8% of persons fall below in Charts 1a and 1b, and applying the Rs 685 spending level which allows a nutrition norm of 2,400 calories, there is a calorie intake of 2,325 only, not 2,400 unique value for the rural poverty line and calories. Plus ca change ……..again and the associated poverty percentage. The em- again we find unstated and implicit lowpirical reading can vary at most by plus/ ering of the standard to arrive at under minus Rs 5 for the poverty line and plus/ estimated poverty. Deaton and Dreze’s minus 0.5% for the associated poverty “below 2,400 calories” estimate is actually percentage when we read the graphs, but a “below 2,325 calories” estimate. One it cannot possibly differ by as much as 7 wonders what the point of doing this percentage points. That is why the Deaton might be, and what conceivable differand Dreze estimate of 79.8% below 2400 ence it would make to admit that the becalories is so problematic. low 2,400 calories rural poverty is nearly Inspecting the first five columns of 87% and not just below 80%. Table 4 shows that the Deaton and Dreze How have Deaton and Dreze arrived at calculation of 79.8% under 2,400 calories, their very precise looking but incorrect cannot be correct. 80.4% of all persons estimate of 79.8% of rural persons below spent below Rs 690 per month, with an 2,400 calories intake in 2004-05? Let us try average calorie intake of 2,290 calories to solve the puzzle. If the relation of calorie per day in the class Rs 580-Rs 690. Clearly intake to MPCE levels (which is available 79.8% of persons would be spending a only for the URP distribution, and is shown little below Rs 690. The average calorie in Chart 1b) is used to obtain the direct intake of even the next higher class Rs 690- poverty line, Rs 800 then the only way a Rs 890 is marginally below 2,400 calories lower figure than the correct estimate can and 90.2% of all persons spent less than be obtained is to apply this poverty line to Rs 890. Deaton and Dreze’s 79.8% of per- an ogive which is not the appropriate URP sons could not possibly access anything ogive to which the calorie intake data perbut a little below 2,335 calories (2,290 class tain, but a different ogive lying to the right 1a Ogive, All-India Rural 2004-5 URP Chart 1a: Ogive, All-IndiaChart Rural, 2004-05 (URP) 100 100
Percent of Persons
% of persons
Daily per capita Calorie Intake by MPCE, All-India Rural 2004-5 URP
Daily calorie intake
Economic & Political Weekly EPW february 6, 2010 vol xlv no 6
of the URP one. The mixed-recall period (MRP) for the same year gives precisely such an ogive (col 6) since it records the same per head food expenditure but higher per head non-food expenditure (on account of the 365 day recall period used rather than the 30-day period, for infrequently purchased non-food items). Let us do some detailed statistical detective work to trace the possible steps through which a lower than actual poverty percentage can be claimed, referring to Table 4. Suppose the DPL of Rs 800 from the g relation of Chart 1b is incorrectly applied to the MRP ogive for the same year (col 6) which lies to the right of the URP ogive, we would get 84% of persons below Rs 800. This while shaving off 3 percentage points from the correct figure remains higher than the Deaton and Dreze estimate. Carrying the detective work a step further, if we apply the same poverty line to the MRP ogive for the next year 2005-06 which lies even further to the right (col 7), we get a fraction over 80%, still a tiny bit “too high” (see Chart 2, p 78). If the relevant curved segment of the ogive is treated as linear however the percentage of persons drops a whisker below 80, the poverty estimate claimed by Deaton and Dreze for 2004-05. Chart 2 presents the 2005-06 MRP ogive with grid lines for greater ease of checking the argument. We are not in a position to assert positively that Deaton and Dreze (2009) have arrived at their figure of 79.8% by using this particular procedure, since only they know what they have done and they have not chosen to share their method with their readers. All we are saying is that following the incorrect procedure of taking the URP poverty line for 2004-05, and applying it to the 2005-06 MRP ogive while treating the relevant segment as linear, does give exactly their estimate. The fact that the Planning Commission has been applying its URP price-updated poverty line to the MRP distribution to give a second, lower poverty estimate suggests that the incorrect method has wide acceptance at present among economists working in the area. However widely it is being used, the fact remains that applying a URP poverty line to the MRP distribution is incorrect procedure even if the distributions are for
Ogive, MRP 2005-6 All-India Rural
(Table 5) and this extra spending is recorded on account of non-food items under the 365-day recall for infrequently purchased goods. The average food spending and calorie intake are exactly the same for both, since non-food items cannot be used for obtaining calories; it is only Charlie Chaplin who 1000 1200 1000 1200 on celluloid could boil shoes to make soup, but in real life it does not happen. So it is a matter of common sense that the URP basis expenditure level at which 2,400 calories can be obtained, has to be suitably raised by the amount of extra non-food spending, to give the MRP basis poverty line to preserve access to the same level of nutrition. Differently put, a URP poverty line giving the norm level of calorie intake, will necessarily correspond to a lower than norm calorie intake and hence a lower than norm level poverty ratio, when applied to the MRP curve of the same year. Since any given calorie intake level will be accessible only at a higher MRP spending than URP spending, the g curve pertaining to the MRP distribution for a given year will be a different one and it will always lie below and to the right of the g curve of the URP distribution for the same year. We can see precisely how the average expenditure and calorie intake by MPCE
Chart 2: Ogive, Mixed Recall Period 62nd Round 2005-06, All-India Rural 100 100
% of persons
Percent of Persons
Source: Table 4, Col 7.
600 600 MPCE, Rs MPCE, Rs.
the same year. The average MPCE and calorie intake by classes will not remain the same for the MRP distribution as the URP one when the same money class limits are retained as is the case, and therefore the poverty line itself will be a different and higher one than Rs 800. And it is doubly incorrect to take the MRP distri bution for any later year and use the same poverty line, because the price inflation factor will kick in additionally to make the given money poverty line inapplic able. At the very least the poverty line has to be revised upwards for Inflation. The second point is obvious but the first point needs more elaboration, which is relevant also for our later critique of the Planning Commission practice of giving two separate poverty estimates using the same poverty line. The MRP basis MPCE is on average only Rs 20 higher than the URP basis spending
Table 5: Per Capita Expenditure by MPCE Classes in MRP Distribution Compared to URP Distribution, Both Before and After ‘Adjustment’ of Households, by Unchanged Money Class Limits (All-India Rural 2004-05) MPCE Class Rs 1
Cumulative % of Persons URP 2
Before ‘Adjustment’ Average MPCE, Rs URP Basis MRP Basis 3 4
Cumulative % of Persons MRP 5
After ‘Adjustment’ Average MPCE, Rs URP Basis MRP Basis 6 7
Source: NSS Report No 508, 2004-05 Table 6R for the four columns of average MPCE. The “before adjustment” refers to the average taking the same numbers of households in each MPCE class even with MRP spending, while the “after adjustment” refers to the result of shifting out of households according to whether their MRP spending exceeds the class limit (or in a smaller number of cases falls below the class limit). The MRP values of Col 7 are associated with a lower per capita food spending level and hence lower calorie intake compared to the URP Col 3, values, for the reasons explained.
classes will change for the MRP distribution, by consulting the detailed Tables 6R and 6U in NSS Report 508 which give the cross-tabulation of households under both MRP and URP using the same money class limits. The MPCE for the two distributions are shown in our Table 5, abstracted from the detailed Table 6R of the cross-tabulation. Studying these data the reader can appreciate the point being made here, that the new g relation relevant for the MRP distribution will lie to the right of the URP g relation, just as the MRP ogive lies to the right of the URP one. The two distributions will give exactly the same direct poverty estimate. Consider the poorest class spending below Rs 235 containing 4.8% of persons. The MPCE is Rs 199.5 of which food spending is 68.5%, and mean calorie intake is 1,376. With MRP basis spending, still taking the same set of households (namely before “adjusting” them), the average spending goes up to Rs 225.11, but food spending remains unchanged and is a lower share, 60.7% of this, hence average calorie intake will be unchanged. However, the number of households will not actually remain unchanged in the given class with MRP spending since those mainly lying in the top half of the class will exceed the class limit of Rs 235. Only 57% of households originally in this class continue to remain in it (see Table 6R, Report 508) with adjustment, the remaining 43% move up while some households move into this class from higher spending classes (see fn 5 for the reason). The average MPCE declines to Rs 188.32 on URP basis for the new group of 2.8% of persons. This is lower than before by Rs 11.2 so the food spending falls, entailing at least 30 to 35 calories decline in average intake, given the slope of the g relation in this segment. Or, take the class of 9.8% of persons, spending between Rs 690 and Rs 890, which contains the Rs 800 direct poverty line. Of Rs 775 average spending 53.9% is on food, giving 2,380 calories. The same set of households spend on MRP basis almost Rs 40 more, Rs 814.4. The unchanged food spending declines to 51.3% of this and calorie intake is unchanged. After households are “adjusted” depending on whether their MRP basis spending keeps
february 6, 2010 vol xlv no 6 EPW Economic & Political Weekly
them within the same class limits or moves them up or down,3 60% remain, most move up, some move down and many enter it from lower classes. The URP basis spending for the new group of 11.3% of persons lying within the same class limits, declines by Rs 22 to Rs 753.12 so food spending and hence calorie intake will be lower, by 18 to 20 calories given the slope of the g relation over this segment. To sum up, although overall average calorie intake is the same for the URP and the MRP basis data, in the latter the average calorie intake will be lower in each MPCE class, and a higher value of MRP spending than URP spending, will be required to reach any given nutrition level. (These lower class-wise per capita calorie intakes are getting weighted differently in MRP, by a higher proportion of persons being in higher spending classes, with the overall average calorie intake unchanged.) Thus, for a given year the g relation for MRP will always lie to the right of the g relation for URP. Inspecting Table 5, we see that the col 3 URP values and the col 7 adjusted MRP values are almost identical for all classes but the top one, which is bound to be the case since the money class limits are fixed except for the open-ended top class. But it is important to remember that the MRP values of col 7 have lower per capita food spending and hence lower calorie intake compared to the URP col 3 values, for the reasons explained. We cannot plot the g relation for MRP 2004-05 exactly yet since we have not been given the new calorie intake values for the adjusted classes of the MRP distribution by NSSO. Nevertheless, since the observed MRP spending is about Rs 40 higher than URP spending in the relevant segment, we can safely infer that the required MRP poverty line for accessing 2,400 calories will be about Rs 840 corresponding to the URP poverty line of Rs 800. Applying this Rs 840 to the MRP ogive for the same year, shows nearly 87% of persons falling below it. In short, no one can possibly obtain a below 2,400 calories poverty percentage which is a lower one simply by using a MRP distribution – since this has exactly the same per capita calorie intake – unless the wrong procedure is followed of applying the URP poverty line to it. Those who do this are necessarily making
the absurd assumption that MRP gives a higher calorie intake than URP. In short people not only make soup out of shoes like Charlie Chaplin, they also obtain calorie intake from consuming the result. Statistics do not lie, it is only mishandled and abused statistics that are made to lie, by applying the poverty line relevant for one distribution incorrectly to a different distribution. The recent Tendulkar Committee Report has fallen into the same error when it says that its revised MRP poverty line of Rs 446.7 for rural India, 2004-05 gives 1,999 calories. This stated calorie figure is not correct, it is only consistent with applying this MRP poverty line to the URP g relation of Chart 1b. Unknowingly, the Committee too has assumed that calorie intake is obtainable from the excess recorded spending on shoes, clothing and the like under MRP. Correct computation requires getting the URP value for the stated MRP line, before applying it to the URP g relation. Alternatively, the stated MRP poverty line should only be applied to the new g relation of the MRP distribution. From the ogives for 2004-05, we find that the stated MRP poverty line of Rs 446.7 is lower at Rs 415 on URP basis and it gives 1,930 calories. In Table 6 we have included the Tendulkar Committee’s new poverty estimates using MRP lines, and we have obtained the URP poverty lines corresponding to these, Table 6: Rural Poverty in India Including Revised Estimates by Tendulkar Committee (All-India 1983 to 2004-05) NSS Round
38th 50th 61st 1983 1993-94 2004-05
(1) Direct poverty line (DPL),Rs <2,400 Kcal
(3) a Official poverty line (OPL), Rs 89.5
(2) Direct poverty ratio (DPR), % b Revised OPL Tendulkar Committee
(ROPL) NA NA 446.7 (MRP) (4) a Official poverty ratio (OPR), % 45.6
b Revised official poverty ratio (ROPR)
(5) a Calorie intake @ OPL
b Calorie intake at ROPL
(6) a Deficit of calorie intake at OPL - 340 - 420
b Deficit of calorie intake at ROPL NA
The URP basis poverty lines marked with asterisk are comparable with other official poverty lines in the table. These PLs correspond to the revised poverty percentages for 1993-94 and 2004-05 presented in the Tendulkar Report, and have been obtained by this author from the ogives for the two years. The Report gives new poverty ratios for both years and the MRP poverty line of Rs 446.7 for 2004-05, but not the MRP poverty line for 1993-94. This MRP line is likely to be about Rs 255. Prefix R indicates revised values in Tendullkar Committee Report.
Economic & Political Weekly EPW february 6, 2010 vol xlv no 6
comparable to the earlier official ones, by plotting the ogives. The NSSO should calculate and give us the exact average food expenditure and the average calorie intakes corresponding to the MRP distribution, in its reports on nutrition. Such an explicit statement of calorie intakes for the MRP distribution would go a long way to introduce clarity and would help to check the mishandling and abuse of the data we see at present. The mishandling of data had started with the Planning Commission itself most unfortunately using the MRP distribution in an illegitimate fashion. It has been giving not only its regular poverty estimate applying the official poverty line to the URP distribution but has been applying the very same poverty line to the MRP distribution, without apparently realising that the poverty line itself will necessarily change. This is so even though the official poverty line is an indirect one, since it still retains an umbilical connection as it were, however distant, with a nutrition norm. Rural poverty is 28.3% at a poverty line of Rs 356, which is the price-index updated Rs 49 base year level, and rural poverty is stated to be 21.8, nearly seven points lower, by applying the same poverty line to the MRP distribution. The latter procedure is incorrect, and the figure of 21.8% is meaningless. The fact has been conveniently forgotten that the original Rs 49 poverty line was derived by applying a nutrition norm to calorie intakes for a distribution of rural persons in the base year which was not a MRP distribution, so the priceupdated Rs 49 line cannot be validly applied now to a current MRP distribution to obtain 21.8%, as is being done. The commission has to obtain first the current MRP spending which corresponds to its current URP poverty line, and then this has to be applied to the current MRP distribution. This is readily done using the already mentioned cross-tabulations in NSS Report 508 giving the relation between the two distributions. From Table 5 interpolating from the given series, about Rs 383 on the MRP spending basis corresponds to Rs 356 official poverty line which applied to the MRP ogive, gives 28.3% in poverty – both distributions necessarily give the same result. The economists in the Planning Commission do not seem to have applied their
minds seriously to the question of correct computational procedure using their own definitions. Had they done so they would have realised that presenting two widely different poverty estimates on the basis of the same data set as regards food expenditure is a contradiction in terms. To conclude, research scholars are strongly advised not to be tempted to take short cuts in their dissertations and theses by mechanically quoting the figures of “poverty percentages” produced by any individuals or institutions however wellknown, but to go always to the basic NSS data source and take the trouble to understand the data. Published poverty estimates – including this author’s, needless to say – should be checked for definitional clarity, methodological correctness and internal consistency against the basic data. Genuine mistakes can always occur but they cannot be corrected unless estimators clearly give their method of estimation. Otherwise, we would simply have an endless cycle of incorrect estimates being quoted, particularly given the proneness of our intellectuals to commit the fallacy ad verecundiam which Aristotle warned against (the “appeal to awe”, uncritically accepting incorrect arguments – in this
case, incorrect numbers – merely because persons making them may be public figures or have a respected position in society). It is learnt that NSSO is proposing to dispense altogether with the URP basis data and present the MRP basis data alone – which appears cosmetically “better” in giving a lower percentage of persons below every spending limit. Economists engaged in poverty estimation, with their newfound enthusiasm for the MRP distribution, should remember, however, that abandoning the URP basis data will not make an iota of difference to obtaining the correct direct poverty estimate, because the correct poverty line will necessarily be higher for the MRP distribution and give exactly the same result as the URP distribution. Fortunately, the integrity of the Indian data collection system itself remains unimpaired so far and the NSSO makes its reports available to all online, in the interests of transparency. No measure which undermines the integrity of our data collection system built up by P C Mahalanobis decades ago should ever be permitted. Notes 1 The relevant deflator for the poorest class is not the 2,100 calories implicit deflator; if we apply the
1,800 calories deflator real spending falls little to Rs 130 by 2004-05. 2 Incidentally, I consider it inappropriate to call the g relation a “calorie Engel curve” because Ernst Engel who studied a sample of 153 Belgian households in 1863 discussed the share of food expenditure in the total and not calorie intake. Every variable which depends on expenditure cannot be called an Engel relation. 3 While the majority of households move up with MRP classification, a minority also moves down to lower spending classes. This could well be as the Report 508, pp 12-13 mentions, because these households happen to have made the infrequent bulky purchase in the particular reference 30 days, and this same expenditure reported for 365 days in the case of Mixed Recall Period response, being divided by12.17 gives a greatly lowered 30day average. Report 508, pp 12-13 points out that “many more households move up than down as a result of adjustment”.
References Deaton, A and J Dreze (2009): “Food and Nutrition in India: Facts and Interpretations”, Economic & Political Weekly, Vol XLIV, No 7, 14 February. Ministry of Finance, Government of India: Economic Survey 2008-09. National Sample Survey Organisation: Report No 402, Level and Pattern of Consumer Expenditure 1993-94. – Report No 405, Nutritional Intake in India 1993-94. – Report No 508, Level and Pattern of Consumer Expenditure 2004-05. – Report No 513, Nutritional Intake in India 2004-05. – Report No 527, Household Consumer Expenditure 2006-07. Patnaik, U (2007): “Neo-liberalism and Rural Poverty in India”, Economic & Political Weekly, Vol XL11, No 30, 28 July-3 August. Planning Commission, Government of India (2009): Report of the Expert Group to Review the Methodology for Estimation of Poverty. Reserve Bank of India (2009): Handbook of Statistics on Indian Economy.
SAMEEKSHA TRUST BOOKS
1857 Essays from Economic and Political Weekly A compilation of essays that were first published in the EPW in a special issue in May 2007. Held together with an introduction by Sekhar Bandyopadhyay, the essays – that range in theme and subject from historiography and military engagements, to the dalit viranganas idealised in traditional songs and the “unconventional protagonists” in mutiny novels – converge on one common goal: to enrich the existing national debates on the 1857 Uprising. The volume has 18 essays by well known historians who include Biswamoy Pati, Dipesh Chakrabarty, Peter Robb and Michael Fisher. The articles are grouped under five sections: ‘Then and Now’, ‘Sepoys and Soldiers’, ‘The Margins’, ‘Fictional Representations’ and ‘The Arts and 1857’.
Pp viii + 364 2008 Rs 295 Available from
Orient Blackswan Pvt Ltd Mumbai Chennai New Delhi Kolkata Bangalore Bhubaneshwar Ernakulam Guwahati Jaipur Lucknow Patna Chandigarh Hyderabad Contact: [email protected]
february 6, 2010 vol xlv no 6 EPW Economic & Political Weekly