What we count when we count research
| AUTHOR | Arindam Goswami |
| DATE | October 29, 2026 |
| CATEGORIES | High Tech Geopolitics Research and Development |
Parliament’s Standing Committee on Petroleum and Natural Gas tabled a report on 12 August on research spending by India’s oil PSUs. It asks the petroleum ministry to set a floor, a minimum share of profit before tax each must spend on research. A version of that is already law. But the case weakens once one sees how these eleven companies spend and patent, because capabilities differ.
The Department of Public Enterprises revised its CPSE performance guidelines on 6 May, adding two items to the scorecard each company signs with its ministry: R&D spending as a share of the previous three years’ average profit before tax, and patents admitted that year. Each carries two marks out of a hundred, feeding the rating that governs performance related pay for senior executives, so what the scorecard counts enters the calculations of those who run these companies.
India’s gross expenditure on research has hovered near 0.7% of GDP for years, and the political instinct is to push it up. CPSEs spent about Rs. 9700 crore on research in 2024-25, oil and gas accounting for roughly a third, second only to defence production.
The trouble is the range. Spending across the eleven runs from 0.58% of profit at one refinery to 4.17% at another, a sevenfold spread in one sector under one owner in one year. The large firms are not the laggards: spending tracks profit almost proportionally, and size does not predict the share devoted to research. What differs is what each has built, its laboratories, and whether it sells technical services or only refines crude.
A floor corrects one failure - a profitable firm underinvesting relative to its means. It also overrides differences in what firms can do with money. Set at 3% of profit, it would require the sector to spend about 47% more, and six of the eleven would fall short. One standalone refinery would have to multiply its research budget more than fivefold, without the laboratories or scientists for it. The report does not record its research headcount. Money a company cannot deploy gets relabelled, existing engineering and testing spending counted towards it.
So the ministry should establish who can absorb research money before mandating that they spend it. It would require a circular asking each PSU for its research staff and laboratory capacity, and targets can then be set firm by firm against its own three-year average.
DPE deserves credit for the second item, which its guidelines frame as a shift from inputs to outputs. The diagnosis is right, but not the count. One company records nearly eight granted patents per technology developed over five years, against one to just over two for the others, which makes it look four times as inventive. It is not. Seven in ten of its grants came from outside India, and one invention filed in five offices produces five grants. Adjust for the foreign share of each portfolio and the gap closes. The objection is not to foreign filing, often the better commercial decision, but to counting one invention five times.
Variation in patent counts is therefore variation in filing budgets. The guidelines do not say whether foreign admissions count. If they do, a CPSE can raise its score by filing existing inventions in three or four more offices, for a few tens of lakhs and no new research.
Patent statisticians solved this long ago by counting families rather than grants. A family groups every patent worldwide protecting one invention, so five offices give one family. The European Patent Office publishes family data free. DPE should count families, or Indian admissions only, which works almost as well. This is not to discourage filing of patents in foreign jurisdictions; just to club same filings in the scorecard.
A further finding complicates this. These companies convert 57 to 81% of the technologies they develop into commercial use, so deployment is not the failure, but licensing to outsiders is. The largest spender recovered under 3% of a year’s research budget in royalties and licence fees. But the same company saved Rs. 33 crore in a year by designing its own engineering packages instead of buying licences, and used 19000 tonnes of self-developed catalyst instead of imports. That should get counted as royalty income.
A target counting only external receipts would push these companies away from where they create most value, substituting for imports and avoiding licence fees. The ministry should require avoided costs to be disclosed alongside receipts, audited and comparable across firms.
One last reason the metric matters: research spending here is not a committed programme. It moves by twenty to thirty per cent a year in both directions. A number management can move at will is exactly the kind that responds to a scorecard.
The lesson is not confined to petroleum. Any target built on a countable proxy is met by whatever is cheapest to count. India is running the same experiment at far larger scale through the Research, Development and Innovation Scheme, Rs 1 lakh crore over six years with about Rs. 2000 crore sanctioned. Getting the measure right while the sums are small is cheaper than fixing it later.