ECONOMY

Questions of statistical validity

How valid are questions raised by the Centre for Monitoring Indian Economy regarding growth rate projections released by the Central Statistical Organisation?

V. SRIDHAR

WHEN the Central Statistical Organisation (CSO) released its Advance Estimates of Gross Domestic Product (GDP) for 1998-99 on February 9 projecting a 5.8 per cent increase, the figures were greeted with surprise. It was generally believed that Indian industry was going through a trough for the third successive year, but the CSO projected a growth rate of 5.7 per cent for the manufacturing sector, 4.7 for industry as a whole and 6.7 per cent for the services sector. Although the agricultural sector was expected to recover only marginally (it had declined by 1.5 per cent in 1997-98), the CSO projected a 5.3 per cent growth.

The scepticism was heightened by the fact that the CSO had changed the base year for computing its estimates of real GDP - the value of all goods and services produced - to 1993-94 from the earlier base of 1980-81. The fact that the new base, which resulted in a higher level of GDP, was used by the Finance Minister in the Union Budget for 1999-2000 in order to show that fiscal deficit as a percentage of GDP had declined fuelled further doubts about the CSO's projections. Was the CSO seeking to aid the Government by providing results that favoured it?

The Centre for Monitoring Indian Economy (CMIE), a private think-tank which makes independent forecasts of some basic macroeconomic aggregates, disputed the CSO projections as "overestimates" in its Monthly Review for February. The CMIE said that GDP would grow only by 4.5 per cent in 1998-99. It argued that since foodgrain production was set to increase by only 1.5 per cent, unless the rest of the agricultural sector grew by 10.5 per cent in the remaining three months, 5.3 per cent growth would not be possible. Such a growth rate was unprecedented, it argued.

The CMIE also forecast a growth rate of 4.2 per cent in the industrial sector and of 6.2 per cent in the services sector. It argued that the Index of Industrial Production (IIP) had grown by only 3.5 per cent in the nine months of 1998-99. Industrial growth would need to be in excess of 7 per cent in the remaining three months in order to cause a higher rate. "This is unlikely," it observed. (When the CSO released its projections, the IIP figures were available only for the April-November period. The latest IIP data indicate that industrial growth has been 3.32 per cent and that manufacturing output has increased by 3.53 per cent between April 1998 and January 1999.)

The CSO maintains that the IIP reflects the value of output and not the value addition in industry, which is the relevant data for computing the GDP emanating from the industrial sector. Officials of the CSO explained that although the IIP was projected to grow by 4.6 per cent, industrial growth in terms of value addition would be 5.7 per cent. This, they said, was because the weighting diagram in the IIP was substantially different from the weighting diagram for value addition determined by using data from the Annual Survey of Industry (ASI), which is based on returns filed by all factories covered by the Factories Act. For instance, transport equipment, which has a weight of 5 per cent in the IIP, contributes about 9 per cent of the value addition in the manufacturing sector as reflected in the ASI data. The CSO projects that this sector will grow by 24 per cent. Since this sector has a higher weightage, it will lend buoyancy to the growth in value addition in industry as a whole. Sources in the CSO told Frontline that given the latest IIP data, the CSO's forecasts for industry may not prove correct. However, they pointed out that this in no way invalidated their earlier estimate.

HOW did the CMIE arrive at its forecast? Its executive director, Mahesh Vyas, told Frontline that the CMIE made a "study of the long-term relationship (15 years) between value addition and output." He explained that in the time series greater weightage was given to "more recent years" and less to earlier years. Vyas admitted that this was an "arbitrary procedure" but argued that this was the only way to capture the effects of "structural changes in the Indian economy in the last few years." "We do not use econometric modelling techniques for forecasting, unlike organisations such as the CSO and the National Council for Applied Economic Research (NCAER), because those kinds of models do not allow structural changes to be captured by their models," Mahesh Vyas said. He said that "structural changes have ensured that the economy is no longer linked to historical data." He also confirmed that the CMIE data did not "capture value addition in unregistered manufacturing". However, many economists would construe that the CMIE's methods take substantial liberties with economic statistics.

The CSO's estimates that the growth in the agricultural sector may be 5.3 per cent are more likely to be accurate. Recent reports indicate that the rabi crop may be the best ever. In fact, there are reports that the good performance of the agricultural sector may even result in a GDP growth of 6.4 per cent.

Mahesh Vyas initially told Frontline that foodgrains accounted for about 55 per cent of the weightage in the CMIE's calculations of the GDP emanating from agriculture. However, when it was pointed out that foodgrains account for only 29 per cent of the weightage in the CSO's method, he admitted that the CMIE's figures would need to be reviewed. The CSO estimates reckon that foodgrain output will increase by 1.5 per cent - by 2.5 per cent in terms of value. A senior CSO official told Frontline that much of the increase, in value terms, would be because of the 7 per cent growth in the value of output from commercial crops.

S.L. Shetty, director of the Economic and Political Weekly Research Foundation, told Frontline that he reckoned that the CMIE "jumped to its conclusion" because its statistical coverage of the agricultural sector may not be as good as that of the CSO. He does not think that the CSO's figures reveal any bias.

Regarding data on industry, Shetty said that his own analysis of industrial data on output and value addition showed that value addition had been occurring faster than the value of output in recent years. He said: "There is a complete change in the 1990s. I suppose it means that liberalisation has meant that productivity has increased in industry; we probably have to accept it."

The CSO has made improvements or included anew 17 economic activities while shifting to a new base using 1993-94 prices. Shetty is of the opinion that the change in the base year can only be good, as it would enable better capturing of the composition of the economy's constituents, which have changed over time.

THE focus on the CSO and its methods of collecting, analysing and disseminating data have also raised questions about the weaknesses of the data system of economic statistics in India. It is not as if the CSO itself is happy with the statistics that it gathers or uses. For many years now, CSO officials and economists have pointed out that there has been a steady deterioration in both the quality and quantity of data gathered in India. There is also a perception that the institutional arrangements for the data gathering system - in both agriculture and industry - have weakened during the last two decades.

The CSO is dependent on the Union Agriculture Ministry for statistics on crop output. The CSO has repeatedly said that crop-cutting estimates, which form the basis for determining yields, are being done in an unscientific manner. The Patwari system in Tamil Nadu, Andhra Pradesh and Karnataka, under which village officials were responsible for collecting statistics on crops grown three times in a year, had collapsed and "the gap was never filled", according to CSO sources. In States such as West Bengal, Kerala and Orissa - the "permanently settled States" - the concept of a village official who would register the crop area, the area sown and the output never evolved. (In Kerala a "cross-reporting system" has evolved, which, they say, has been "reasonably good".)

Earlier, the National Sample Survey Organisation (NSSO) conducted surveys on the yield and these were used to "cross-check" the data available to the CSO. Sources in the CSO say that there is a need to reintroduce such independent estimates. There is also a perception that data on agriculture needs to be freed from the control of the Ministry so that more unbiased estimates may be arrived at. "There is uneasiness among data compilers and there is need for the CSO to be able to coordinate directly with the bureaus of statistics in the States," Shetty says. The CSO has said that about 40 per cent of the geographical area of India is outside the purview of any scientific estimation techniques on crop yields and output. "We are helpless, all the estimates are coming from the Ministry of Agriculture," says a CSO official. Shetty welcomes the CSO's reported intention to change the base year every five years, coinciding with the quinquennial surveys conducted by the NSSO. He says that this would enable the capture of the changes in the composition of the economy.

A source in the CSO pointed out that it was not as if the inadequacy of accurate data on agriculture was merely a concern for academics; the problem, he said, could be far more serious. He pointed out that in 1995-96, wheat was offloaded in the open market by the Food Corporation of India on the basis of optimistic forecasts by the Agriculture Ministry. However, the wheat crop fell short by 10 million tonnes necessitating substantial imports. The reverse happened in 1996-97 when the output was substantially higher than the estimates. Imports were contracted, based on the estimate.

REGARDING industrial statistics, experts feel that economic processes outside the control of the CSO are at work. There is a growing perception that economic liberalisation has resulted in a deterioration in the quality of economic statistics - particularly data on industry - because economic units no longer report data that they reported earlier under a regulated environment. Under the regulated regime, the Directorate-General of Technical Development (DGTD) not only was responsible for issuing licences, but was empowered to gather industrial data from companies. Its role as a regulator had a powerful effect on compliance by companies and industrial units. A senior CSO official told Frontline that "only 50 per cent of the returns are filed by industrial units." Under-reporting is not confined to small units. Shetty says that many of the bigger companies producing ice-cream are not reporting because the sector is reserved for small-scale units.

Data on industry are now gathered by the Department of Industrial Policy and Promotion in the Industry Ministry. "The Ministry," Shetty said, "has lost its rationale to gather data after the abolition of the DGTD." He said: "Much of the control has disappeared and these functions ought to be transferred to the Department of Statistics." Sources in the CSO say that after the abolition of the DGTD, data from new enterprises are not captured, nor do they accurately reflect those which have exited."

The CSO plans a new IIP based on returns filed by "200 plus companies", each employing more than 200 persons. Shetty suggests the formation of a national statistical authority that would govern the Department of Statistics, the NSSO, the CSO and other Central statistical bodies gathering economic and social development statistics. He says that regional bodies should have "work carved out for them" to prevent the possibility of excessive centralisation. "It is very important," says Shetty, "that the authority should have powers to command information from regional and private bodies to cross-check data." A CSO official agreed with this suggestion.