Potential Bias of the Offi cial Estimates

К оглавлению
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 
136 137 138 139 140 141 142 

Criticisms of offi cial poverty estimates focus on the methodology used by

the NBS to calculate the offi cial poverty line (Park and Wang, 2001). The

calculation may be subject to a number of potential biases that may work in

opposite directions, although mainly towards underestimating poverty. First,

an arguably unrealistic food bundle was adopted, in which consumption

items viewed to be non-necessities were excluded (for example alcohol and

sweets). As a result grain accounted for 88 per cent of expenditures, even

though grain comprises only around 70 per cent of actual food expenditures

by poor households. Over-weighting of grain in the standard bundle leads

to the under-pricing of calories, since grain is a relatively cheap source of

calories. This gives a downward bias to the poverty line and leads to an

underestimation of poverty. Second, as we have seen, planned prices rather

than market prices were used to value own-produced consumption goods

before 1990. Insofar as market prices were higher and insofar as the rural

poor had to purchase their marginal needs on open markets, this will bias

downward estimates of food consumption. Third, the reduction in the share

of non-food expenditure in the poverty line estimate was controversial and

appeared to contradict known expenditure patterns.14 Fourth, some have

criticized the NBS sample for under-representing households in remote and

minority areas, and those with illiterate household heads. Fifth, despite the

availability of expenditure data, NSB has always calculated poverty rates

using income data, even though for poverty calculations expenditures are

considered to be a better measure of both current and long-term welfare,

and since individuals try to smooth consumption over time, expenditures

tend to vary less from year to year than incomes. In PRC average incomes

are 10 per cent to 20 per cent higher than average expenditures, so that

using income data results in lower poverty rates.15 Finally, and perhaps

most critically, the implicit infl ation rates evident in the offi cial poverty

lines appear much lower than the actual change in the rural consumer price

index and do not allow for regional price differences.

There has been considerable debate on this latter point. For example,

Khan and Riskin (2001) point out that even the rural consumer price index

is likely to underestimate the growth in living costs of the poor, because

their budget shares for food are higher than the average and food prices

have grown faster than other prices.16 In the initial years the offi cial poverty

lines are consistent with the rural retail price index, and in the fi nal years

of Table A.4.2 they are consistent with the rural consumer price index.

However in the intervening years, there are large discrepancies. Most notably

the poverty line increases only modestly during the high infl ation years of

1988 and 1989, and there is a sharp increase in the poverty line in 1997

that is far in excess of infl ation. This helps explain why offi cial statistics

show a steady reduction in poverty in the late 1980s, while other estimates

show little change. If the 1985 line is infl ated by the rural consumer price

index, the 2000 poverty line reaches 721 yuan compared with the offi cial

line of 625 yuan. This suggests that poverty reduction over time may be

exaggerated signifi cantly.

Like poverty lines in many countries the NBS national poverty line does

not allow for regional price differences in calculating the required food

expenditures, nor does it allow for regional differences in the food bundle.

Food prices vary greatly between different provinces due to transport costs

and imperfect market integration. For example, Chen and Ravallion (1996)

estimate that the cost of purchasing the NBS food bundle was 23 per cent

higher in Guangdong than in Guangxi in the late 1980s. Similarly in 1992

and 1995, the local NBS of Jiangxi province calculated provincial poverty

lines of 400 and 750 yuan using local prices and the national food bundle,

which were well above the offi cial national poverty lines of 320 and 530

yuan, respectively (see Table A.4.2). Failure to account for regional price

differences may exaggerate the concentration of poverty in poor regions.