Has economic liberalisation failed to reduce poverty? Data from the National Sample Survey (NSS) of households suggest that the proportion of people below the poverty line has remained more or less unchanged around 36 per cent in the 1990s. So several critics have jumped to the conclusion that liberalisation and fast growth have bypassed the poor.
In fact the data are even more dramatic; they show that liberalisation has bypassed the middle class, the rich and everybody else! They suggest that consumption levels in all groups have improved very little, so inequalities remain more or less unchanged.
Now, had the NSS shown that the rich were appropriating the lion’s share of rapid economic growth in the 1990s, that might have sounded plausible. In fact the data suggest stagnation without rising inequality. They suggest that the rich have stagnated along with the poor in the 1990s. Not believable, I’m afraid.
The national income figures of the Central Statistical Organisation (CSO) tell a totally different story. They show record economic growth of over 6 per cent annually since 1992. This leaves readers with two choices. You can believe the NSS story of economic stagnation for all groups. Or you can believe the CSO’s story of rapid growth, and throw all NSS conclusions (including those on poverty) in the dust bin.
I simply do not believe that consumption has stagnated in the 1990s. If so, how come the purchase of cars has shot up from 100,000 to 600,000 per year?
How come the purchase of TV sets has more than doubled from 4 million to 9 million sets per year? How come Hindustan Lever, Nestle, Proctor and Gamble and all companies making fast-moving consumer goods have enjoyed record sales and profits in the 1990s? They all say the rural market is booming, while the NSS claims stagnation.
Some leftists believe what the NSS says about the poor while disbelieving what it says about the rich. They argue that the rich want to hide their black money and so understate their consumption to surveyors. Very true. But for similar reasons the rich would have understated their consumption in the 1980s too. So the reported consumption of the rich should have increased in the 1990s rich unless the extent of their understatement increased this decade. That seems implausible: income tax rates fell steeply in the 1990s and the havala exchange rate crashed, so the incentive to hide consumption diminished greatly.
The NSS and CSO use different methods of data collection, so differences in figures are to be expected. But if methodology alone was the issue, the difference should have remained unchanged over time. In fact it is rising fast.
Back in 1971, NSS data on consumption amounted to 76 per cent of CSO estimates.The proportion is now down to 65 per cent. This explains why one source claims an economic boom while the other claims stagnation.
Consider cereal consumption. CSO data show that per capita availability production plus net imports) has increased from around 150 kilos per year to 170 kilos. But NSS data show this falling from around 164 kilos to 145 kilos per capita in the same period. Here again, I find the CSO data more credible. I think it ridiculous to imply, as the NSS data do, that that the green revolution has meant less grain per person.
What do other sources say about poverty trends? We have data on wage rates for unskilled agricultural workers, adjusted for inflation. These rose sharply by an average of 4.6 per cent annually in the 1980s, one reason for the big drop in poverty that decade. They continued rising in the 1990s, though at a more modest average pace of 2.4 per cent annually. That is not consistent with stagnation in poverty. It could, however, mean a slower rate of poverty reduction.
Poverty depends on the casual wage rate and number of days worked. Typically, the wage rate rises when the total amount of work increases. If both wages and the number of days worked have risen, then surely poverty should have fallen.
Finally, the NCAER also conducts household surveys, using a low-income yardstick that differs somewhat from the government?s. The NCAER finds that the proportion of low-income households in rural areas declined substantially from 58.8 per cent in 1988-89 to 48.9 per cent in 1995-96. It also shows increased ownership of consumer durables in every income group. This supports the CSO story and contradicts the NSS version.
Statisticians have long debated how to tackle differences between CSO and NSS figures. NSS data are based on relatively small samples, sometimes very thin samples. They show huge year-to-variations, far more so than CSO data. So for many decades the government adjusted NSS data to be consistent with CSO estimates.
Critics complained that this was wrong. Some pointed out that there was little difference between NSS and CSO estimates of goods typically consumed by the poor, so the NSS data should not be adjusted for the purpose of estimating poverty. This was strongly recommended by the Lakdawala Committee, and accepted by the government in the 1990s.
This decision now looks like a mistake. The difference between NSS and CSO estimates of cereals is now so huge that it is absurd to assert that NSS data are plausible in regard to goods consumed by the poor. We must overhaul our statistical systems to generate more credible data. Till then, there is a strong case for returning to the methodology of past decades. With all its flaws, it looks a bit more credible.
That old methodology suggests that poverty has decreased in the 1990s, though more slowly than in the 1980s. This is consistent with trends in agricultural wages, and so seems plausible.
Why has poverty alleviation slowed down despite faster economic growth in the 1990s? Look out for the answers in next week’s column.