Policy making is a tricky exercise
When data cannot be fully trusted, policy making becomes a tricky exercise, says Samiran Chakraborty.
The professional life of a market economist revolves around data. Every new data point is, in principle, an opportunity to reassess economic outlook and react accordingly. So economists demand wide data coverage which is timely, transparent and trustworthy. Recently, Reserve Bank of India Governor D. Subbarao acknowledged issues relating to data that have been bothering economists for a long time.
With India's relative economic prosperity, financial market development, and policy making complexity, the demand curve for data has shifted upwards. Often, the supply of data has not kept pace with the demand. Even as analysts and policy makers debate the growth versus inflation trade-off, timely data for these variables is sketchy. On growth, the gross domestic product, or GDP, and Index of Industrial Production, or IIP, data come with a lag of two months. There is scope to reduce this lag, but a more immediate concern is inadequate data on GDP sub-components.
For example, India has no data on consumer spending, which other countries collate from retail sales. Data on employment and capacity utilisation is also non-existent. We, therefore, have very little sense of the slack in the economy at any point. Similarly, for inflation, the debate over whether the wholesale price index, or WPI, or the consumer price index, or CPI, should be the preferred metric remains unresolved.
The search for a harmonised CPI remains elusive. In its absence, WPI has become the choice of the market. But it is mostly an indicator of supply-side price pressures and does not capture price movements in the services sector at all.
Frequent and substantial data revisions are the second sore point. GDP growth numbers undergo at least four revisions over a two-year period. The initial estimate is extrapolated from just a few indicators, while the final number is often based on completely different sources of data. For example, as a proxy for manufacturing activity within GDP, IIP is replaced by data from the Annual Survey of Industries; we have observed that the IIP data has consistently underrepresented industrial growth. WPI data, which is revised two months after its initial release, is also prone to large revisions, often in excess of 100 basis points.
Even export and import data undergo revisions throughout the year. We sometimes wonder whether we should be forecasting based on the initial release (which has an immediate market impact) or the revised data, which is likely to better reflect the state of the economy. Then, there are large revisions to data when the base year is changed. So the economy often looks different in the rear-view mirror. Changes in base years also bring in different methodologies for estimating data. Time-series data prior to the new base year becomes non-comparable and irrelevant.
The third big challenge is data comparability. Data on production of capital goods is often taken as a proxy for investment in the economy. However, in the recent past, investment growth as part of the GDP data has been declining consistently, but capital goods production within the revised IIP series has held up well. Explaining these anomalies can be quite onerous.
The fourth factor relates to reliability of data. In the provisional IIP and WPI data, the non-response rate among manufacturers to the questionnaires is very high - often in excess of 50 per cent. While this improves in the revised series, more needs to be done to improve its reliability from the start. Several commodities in the WPI basket undergo price revisions very infrequently.
For example, the price of petroleum was unchanged from November 2010 to June 2011, even though international crude prices moved considerably during this period. With close to 90 per cent of the economy in the unorganised sector, we are often unsure whether the IIP or WPI data - which are collected mostly from the organised sector - are an accurate reflection of what is happening in the economy. Data on the rural economy is also sparse, often failing to reflect India's macroeconomic transformation.
As a consequence of these data challenges, even constructing a coherent story of what happened in the past becomes tricky, let alone forecasting what will happen in the future. Policy making is conducted in a world of imperfect information that could potentially lead to sub-optimal responses. Economists cannot avoid looking into the crystal ball, but better quality of data would help us to do so more accurately.
The author is India Head of Research at Standard Chartered Bank
With India's relative economic prosperity, financial market development, and policy making complexity, the demand curve for data has shifted upwards. Often, the supply of data has not kept pace with the demand. Even as analysts and policy makers debate the growth versus inflation trade-off, timely data for these variables is sketchy. On growth, the gross domestic product, or GDP, and Index of Industrial Production, or IIP, data come with a lag of two months. There is scope to reduce this lag, but a more immediate concern is inadequate data on GDP sub-components.
Chakraborty's take
|
The search for a harmonised CPI remains elusive. In its absence, WPI has become the choice of the market. But it is mostly an indicator of supply-side price pressures and does not capture price movements in the services sector at all.
Frequent and substantial data revisions are the second sore point. GDP growth numbers undergo at least four revisions over a two-year period. The initial estimate is extrapolated from just a few indicators, while the final number is often based on completely different sources of data. For example, as a proxy for manufacturing activity within GDP, IIP is replaced by data from the Annual Survey of Industries; we have observed that the IIP data has consistently underrepresented industrial growth. WPI data, which is revised two months after its initial release, is also prone to large revisions, often in excess of 100 basis points.
Even export and import data undergo revisions throughout the year. We sometimes wonder whether we should be forecasting based on the initial release (which has an immediate market impact) or the revised data, which is likely to better reflect the state of the economy. Then, there are large revisions to data when the base year is changed. So the economy often looks different in the rear-view mirror. Changes in base years also bring in different methodologies for estimating data. Time-series data prior to the new base year becomes non-comparable and irrelevant.
The third big challenge is data comparability. Data on production of capital goods is often taken as a proxy for investment in the economy. However, in the recent past, investment growth as part of the GDP data has been declining consistently, but capital goods production within the revised IIP series has held up well. Explaining these anomalies can be quite onerous.
The fourth factor relates to reliability of data. In the provisional IIP and WPI data, the non-response rate among manufacturers to the questionnaires is very high - often in excess of 50 per cent. While this improves in the revised series, more needs to be done to improve its reliability from the start. Several commodities in the WPI basket undergo price revisions very infrequently.
For example, the price of petroleum was unchanged from November 2010 to June 2011, even though international crude prices moved considerably during this period. With close to 90 per cent of the economy in the unorganised sector, we are often unsure whether the IIP or WPI data - which are collected mostly from the organised sector - are an accurate reflection of what is happening in the economy. Data on the rural economy is also sparse, often failing to reflect India's macroeconomic transformation.
As a consequence of these data challenges, even constructing a coherent story of what happened in the past becomes tricky, let alone forecasting what will happen in the future. Policy making is conducted in a world of imperfect information that could potentially lead to sub-optimal responses. Economists cannot avoid looking into the crystal ball, but better quality of data would help us to do so more accurately.
The author is India Head of Research at Standard Chartered Bank
