One study claims the US government's economic forecasts show an average difference of 1.7 per cent between the preliminary GDP estimates and the final GDP estimates. The US economy has, like all First World economies, a low trend growth rate. GDP growth averages 1.5 per cent per annum since 2003.
So, this is a huge error. Since it's an average, errors at key turning points may be more. The preliminary data for October-December 2008 quarter, for instance, indicated the US economy contracted by about 3 per cent. The revised data a year later indicated the contraction was close to 9 per cent.
Markets bet on preliminary numbers, released with a lag of a month or two. What is more, policy decisions are driven by preliminary numbers. It is likely that the US Quantitative Easing of late 2008 and early 2009 would have been scaled up if policy makers had realised the depth of the recession.
The Indian economy has much higher growth rates than the US, off a base about one-seventh in size. But errors in Indian estimates are also high. What is more, the errors aren't only in GDP calculations and associated numbers like the Fiscal Deficit, Revenue Deficit, and so on. There are also large changes in inflation data, in the Index of Industrial Production (IIP), in trade data and therefore, in the Current Account.
Errors in economic data and estimates are common everywhere. National accounts require the reconciliation of vast amounts of data with errors guaranteed. Humorous articles recently highlighted that aggregated national data showed that the world ran a trade surplus, with tongue-in-cheek speculations about exporting to aliens.
The errors in Indian data are however, egregious, even allowing for inherent issues. Revisions in GDP are understandable. GDP is hard to calculate. Revenue and tax data must be netted off and reconciled, versus subsidies and rebates, etc.
Services in particular, are notoriously difficult to estimate and services contribute the most. IIP calculations also depend on tardy, unreliable compilation. So one can allow for big changes in both IIP and Services GDP estimates.
But there are also frequent revisions in agricultural data. The government is the single largest buyer and seller of food, with effective monopolies in many items. That makes large revisions in the Agro-GDP numbers and in food inflation harder to digest.
Also, given a plethora of controls, the government should have a good idea of imports and exports. Yet trade data revisions see big changes. In Q4, 2011-12, for instance, the preliminary estimates claimed exports exceeded imports by Rs 48,000 crore, for a positive trade balance of about $9.5 billion. Revised estimates a year later say that imports exceeded exports by Rs 130,000 crore, for a deficit of $25 billion. That's a swing of $35 billion - almost 2 per cent of GDP.
Given such large errors, should traders care about preliminary estimates? Only because other traders do. Since the market reacts to this news flow, you must track it. Also, since policy is influenced by this data, it is worth tracking. But you shouldn't blindly believe in preliminary estimates. The track record makes large changes highly probable. Unfortunately, revised estimates come with such major lags that they tend to be ignored.
Cross-checks exist. Many industry associations release monthly or quarterly estimates of sales of items like automobiles, two/ three wheelers, white goods, cement, steel, etc. Those estimates are usually accurate. Financial players, like HDFC, also release details of housing mortgages. There are private institutions tracking agro-commodities.
The RBI releases aggregate credit offtake and deposit growth data from the banking system while all banks release non-performing assets and recast loan figures. (The central bank is also by and large, more accurate in its GDP estimates than the government at large).
In aggregate, all this gives a reasonable picture of changes in consumption patterns, of private investment and of overall economic health. If the trends are broadly in sync with the government numbers, the government numbers are more or less trustworthy.
If the picture offered by the private data flow is at variance with the preliminary government statistics, the chances are very high that the private data presents a picture closer to reality. So in those cases, assume the government data is off-kilter.
Once in a while, this may help you to take a contrarian stance. I suspect this is true of Q4, 2012-13. The preliminary GDP and IIP numbers suggest bottoming out. Private data such as falling car and bike sales, flat mortgages, weak cement and steel numbers, etc. suggest the downtrend continues.
