India's data management, reporting are under scrutiny
In the last two years, there have been two corrections on important economic indicators tracked by the Central Statistics Office (CSO). Delays in reporting by agencies and the pressure to release data as early as possible are testing the limits of the CSO's structural checks.
Answering investors' questions about data errors is embarrassing, says Samiran Chakraborty, Regional Head of Research (India), Standard Chartered Bank. He crunches numbers and puts out estimates on key economic indicators. While drawing up a forecast for the factory output for February, he found himself in a dilemma. "I wondered if I should factor in the error in the January data for industrial production," he says. "I was sure there was an error, but I didn't know when it would be rectified. And it affects the accuracy of our forecast for the next month."
The Central Statistics Office (CSO), part of the Ministry of Statistics and Programme Implementation, revised factory output growth in January down to 1.1 per cent from the earlier estimate of 6.8 per cent. A CSO release said the error occurred because the sugar output was wrongly reported at 13.41 million tonnes.
In fact, this was the cumulative figure for three months, not just January. Actual sugar production in January was 5.81 million tonnes. Economists are baffled that CSO officials failed to notice the unusually high figure.
Pranob Sen, former Chief Statistician of India, says checks are built into the system to detect unusual data movement. "But if the last step of a human check is not done, those checks are not effective," he says.
In the last two years, there have been two corrections on important economic indicators tracked by CSO. GDP growth for the first quarter of 2010/11 was wrongly reported as 3.7 per cent and later revised to 10 per cent. And now the January figure for the Index of Industrial Production (IIP) has been revised downwards.
Chief Statistician of India T.C.A. Anant defends CSO's role in the IIP correction, saying it followed protocol after it noticed the surge in sugar production. He adds that the food ministry, which collates sugar data, even confirmed the unusually high number. Anant says CSO depends on ministries for the data it uses to calculate indices, and faces pressure from policy makers, who want information quickly.
"The risk of error is higher in certain series," says Anant. For example, the volatility of production data makes IIP more susceptible to error than, say, the Wholesale Price Index. CSO gets data from 15 agencies. There is an agreed time frame for reporting, but most agencies are slipping on their schedules. "The error occurred because of the slippages," says Anant. "The department did cross-check data with the food ministry. It does structural checks, but how well you do the task depends on how much time you have."
Delays in reporting by agencies and the pressure to release data as early as possible are testing the limits of CSO's structural checks. But this explanation does not satisfy policy makers when the data in question is one of the most crucial indicators of the Indian economy.
"A delay of one or two days is lesser a crime than releasing the wrong data," says C. Rangarajan, Chairman of the Prime Minister's Economic Advisory Council. But a delay would mean violating an advance release calendar for data issued by CSO.
The errors suggest that data collection and computation methods have not kept pace with economic growth, and there is a need for reform. "If they don't reform the system, errors will crop up," says economist Sudipto Mundle.
"The IIP, which was very good in the 1980s, is not very good today," says Anant. He adds that reforming it and making it more dynamic with a changing composition would be an institutional challenge. Rangarajan has set up a committee to analyse the structure of IIP, underscoring the need to examine the methodology for computing data. The committee, headed by Planning Commission member Saumitra Chaudhuri, will consider new commodities for inclusion in the index, and revisit the weightages. The data collection procedure has been an issue of concern for several years. The multiplicity of sources cannot be avoided, but greater responsibility could be fixed in each department for the collection and release of data.
While the misreporting of GDP growth was a computation error, the IIP mistake was caused by a flub on the part of the data reporting agency. Rangarajan of the Economic Advisory Council says: "It should have struck somebody when the sugar production shot up. Questions should have been asked."
CSO'S chief says that the spike in sugar production was spotted and verified with the food ministry, but that could not prevent erroneous data from being reported. Economist Mundle says that since IIP is dependent on multiple agencies and departments which "have zero professionalism", data collection is "very messy". Greater accountability and capacity expansion at the reporting agency level may help reduce such errors. But human error, which is responsible for many data reporting errors, is the most challenging to tackle. Data management and reporting experts say a structural change in the way IIP is computed will not necessarily minimise human error. "I don't think there is any way to minimise it," says former chief statistician Sen.
For example, when the exports figure was overstated by $9.4 billion for the period between April and October 2011, the problem was identified at the data entry level. Typically such data is entered by customs officials and collated from ports. The impact of the series of data corrections is being felt widely, and data miners like Standard Chartered's Chakraborty are increasingly depending on data from the Reserve Bank of India, companies and surveys.
In fact, Chakraborty is bracing himself for another revision. "I think that the second- and third-quarter GDP data will be revised substantially, too," he says. "Look at the consumption data - private consumption growth slipped from 5.9 per cent to 2.9 per cent, and back to 6.2 per cent in three quarters last year. Such a zigzag generally does not happen with consumption growth, and the 2.9 per cent number does not gel with other indicators of consumption."
The Central Statistics Office (CSO), part of the Ministry of Statistics and Programme Implementation, revised factory output growth in January down to 1.1 per cent from the earlier estimate of 6.8 per cent. A CSO release said the error occurred because the sugar output was wrongly reported at 13.41 million tonnes.
The statistics department does a set of structural checks, but how well you do the task depends on how much time you have: T.C.A. Anant
Pranob Sen, former Chief Statistician of India, says checks are built into the system to detect unusual data movement. "But if the last step of a human check is not done, those checks are not effective," he says.
In the last two years, there have been two corrections on important economic indicators tracked by CSO. GDP growth for the first quarter of 2010/11 was wrongly reported as 3.7 per cent and later revised to 10 per cent. And now the January figure for the Index of Industrial Production (IIP) has been revised downwards.
Chief Statistician of India T.C.A. Anant defends CSO's role in the IIP correction, saying it followed protocol after it noticed the surge in sugar production. He adds that the food ministry, which collates sugar data, even confirmed the unusually high number. Anant says CSO depends on ministries for the data it uses to calculate indices, and faces pressure from policy makers, who want information quickly.
Delays in reporting by agencies and the pressure to release data as early as possible are testing the limits of CSO's structural checks. But this explanation does not satisfy policy makers when the data in question is one of the most crucial indicators of the Indian economy.
"A delay of one or two days is lesser a crime than releasing the wrong data," says C. Rangarajan, Chairman of the Prime Minister's Economic Advisory Council. But a delay would mean violating an advance release calendar for data issued by CSO.
The errors suggest that data collection and computation methods have not kept pace with economic growth, and there is a need for reform. "If they don't reform the system, errors will crop up," says economist Sudipto Mundle.
"The IIP, which was very good in the 1980s, is not very good today," says Anant. He adds that reforming it and making it more dynamic with a changing composition would be an institutional challenge. Rangarajan has set up a committee to analyse the structure of IIP, underscoring the need to examine the methodology for computing data. The committee, headed by Planning Commission member Saumitra Chaudhuri, will consider new commodities for inclusion in the index, and revisit the weightages. The data collection procedure has been an issue of concern for several years. The multiplicity of sources cannot be avoided, but greater responsibility could be fixed in each department for the collection and release of data.
A delay of two days is a lesser a crime than releasing the wrong data: C. Rangarajan
CSO'S chief says that the spike in sugar production was spotted and verified with the food ministry, but that could not prevent erroneous data from being reported. Economist Mundle says that since IIP is dependent on multiple agencies and departments which "have zero professionalism", data collection is "very messy". Greater accountability and capacity expansion at the reporting agency level may help reduce such errors. But human error, which is responsible for many data reporting errors, is the most challenging to tackle. Data management and reporting experts say a structural change in the way IIP is computed will not necessarily minimise human error. "I don't think there is any way to minimise it," says former chief statistician Sen.
For example, when the exports figure was overstated by $9.4 billion for the period between April and October 2011, the problem was identified at the data entry level. Typically such data is entered by customs officials and collated from ports. The impact of the series of data corrections is being felt widely, and data miners like Standard Chartered's Chakraborty are increasingly depending on data from the Reserve Bank of India, companies and surveys.
In fact, Chakraborty is bracing himself for another revision. "I think that the second- and third-quarter GDP data will be revised substantially, too," he says. "Look at the consumption data - private consumption growth slipped from 5.9 per cent to 2.9 per cent, and back to 6.2 per cent in three quarters last year. Such a zigzag generally does not happen with consumption growth, and the 2.9 per cent number does not gel with other indicators of consumption."
