For the past five years or so I have been questioning the credibility and sustainability of the surge in India's services sector since the mid-1990s, as reported in the official national accounts data.

A recent joint paper by two Brookings Institution scholars (Bosworth and Collins) and a Planning Commission adviser (Virmani) has catalysed my returning to this subject. Before turning to this paper (by BCV), let me reiterate my major concerns and embellish them graphically.

The rate of services sector growth has accelerated noticeably after 1996/97, averaging 8.5 per cent a year in these most recent nine years, compared to an average growth rate of under 6 per cent in the previous three decades. Second, and perhaps even more remarkably, the trajectory of services parted company from that of industry (including construction) after 1996/97.

In one-day cricket parlance, the services "worm", which had till then hugged the industry "worm", surges up and away after 1996/97. Until then the growth of these two sectors tended to be quite closely correlated, except in a couple of years of industrial recession when the growth in industry slumped much more steeply than did the growth of services.

After 1996/97 (and especially during the Ninth Plan period of 1997-2002) the growth of these two sectors diverged markedly. Services retained its buoyancy despite a fairly prolonged period of slow industrial growth.

On average, services grew by 8.5 per cent and industry by only 6 per cent between 1996/97 and 2005/06. (The divergence would be greater if the "construction" sub-sector were grouped with services.)

These divergent trajectories had at least two significant consequences: the share of services in GDP rose quickly to 54 per cent by 2005/06 (compared to 41 per cent in 1990/91 and 50 per cent in 2000/01); and India's services sector accounted for a remarkable (and unprecedented) 67 per cent of all the economic growth recorded between 1996/97 and 2005/06.

Nor can this exceptionally high growth contribution of the services sector be explained adequately by the indubitably rapid growth of the "modern" segments such as IT, ITES, telecom and finance.

These certainly grew fast but from low bases. Even more important (if less explicable) was the strong growth of "traditional" segments such as wholesale and retail trade (though Wal-Mart had yet to come!), transport, and community and personal services, which account for the preponderant share of India's services sector.

Elsewhere, I have also pointed out that the post-1996/97 pattern of services growing significantly faster than industry is one, which finds little resonance in the growth experience of other developing countries.

Specifically, in the 35 years, 1965-99, not one of the seven fastest-growing countries (India does not figure in the shortlist) demonstrated faster growth of services over industry. India's post-1996/97 growth pattern really is quite peculiar and puzzling.

In their paper on the "Sources of Growth in the Indian Economy" (presented at an NCAER-Brookings conference in Delhi a few months ago), BCV are similarly perplexed by the high rates of growth of output and productivity they find in the services sector from the mid-1990s onwards.

They point out that high rates of productivity growth in "traditional" segments of the services sector are neither expected nor found elsewhere in the world. They go on to speculate that price trends in services may have been underestimated in the official Indian data, leading to overestimation of real growth.

How is one to deal with this situation? Obviously, attempting serious alternative estimates of the country's national income estimates is way beyond the capacity of individuals or even think tanks.

Ideally, the CSO should publish a serious discussion paper on the sources and methods of value-added estimates in the services sector, which marshals the evidence in support of these estimates and allays the kinds of concerns expressed here.

That is what I had hoped for when I wrote three years ago "the methodology of service sector estimates changed significantly in the mid-90s at the time of the transition to the new base year, 1993/94.

There is need for serious enquiry to reject the hypothesis that this change in methods may have imparted systematic upward bias to Gross Value Added estimates in key service sectors".

In the meantime, let me indulge in a small thought experiment. Let's assume that the official estimates of growth in agriculture and industry are accurate. Let's further assume that after 1996/97 the services sector has grown not at the officially recorded rate but at a rate 1 per cent higher than whatever is the recorded growth rate for industry in a given year.

This assumption (and it is no more than that) draws support from the earlier history of close correlation between the growth of services and industry and also allows for the indubitable dynamism of some "modern" segments of services such as IT and telecom over the past decade. Based on this assumption we can recompute a fresh trajectory for the growth of GDP (hypothetical).

A couple of points are noteworthy. First, average GDP growth in the past nine years in this hypothetical series is 5.3 per cent, instead of 6.2 per cent as shown by the official series.

Second, in the last three years of high growth, the average "hypothetical" GDP growth of 7.5 per cent is reassuringly high and only a little below the officially estimated average of 8.1 per cent. (This is because industrial growth in these years was high; so the "correction" implicit in the hypothetical series for services is small).

As expected, the main divergence between the official and hypothetical growth series occurs during the Ninth Plan period.

Thought experiments are all very well but what do I really believe? Well...on a balance of considerations (as they are fond of saying in government), I think the hypothetical series for GDP growth in the past nine years may be a little closer to reality than the official series.

Other than the arguments offered here I can advance no "proof" for this. It's certainly not a "slam-dunk" case. But I hope it encourages (provokes?) the official national income estimators to make public a well-reasoned case for their estimates of the level and growth in services value added and thus dispel the heretical doubts raised here.

The author is Honorary Professor at ICRIER and former Chief Economic Adviser to the Government of India. The views expressed are personal