Section 135 of the Companies Act, 2013 gave India a predictable pool of social capital, and the way companies spend it rewards compliance more than results. Our position is that the pool should be judged by what changes for the people it reaches after three to five years, and that this needs a different design: multi-year commitments, paid-for measurement, funded implementation capacity and projects built to join government programmes. Spending that is visible and finished by March is easy to report. It is rarely what improves a child's reading or a village's water supply.
A good rule with a short horizon
The law is simple. Companies above set thresholds of size and profit must spend two per cent of their average net profit over the preceding three years on activities listed in Schedule VII, which covers areas such as education, health, sanitation, skill development and rural development. A board committee approves a policy and an annual plan, and spending is reported each financial year. Few countries have anything similar, and it has moved large sums to causes that would otherwise have waited for public budgets.
The financial year is the problem. A plan approved in spring is spent by the following March, which favours things that can be built, bought and photographed: a classroom block, a water tank, a set of benches, a health camp. These have value. But a school building does not teach, and a tank without a maintenance arrangement stops working. Visible assets also suit a reporting culture in which the question put to a programme is whether the money was spent and the certificate collected.
What the 2021 changes allowed, and what they did not
The 2021 amendments to the CSR Rules moved in the right direction. They defined ongoing projects that can run beyond one year, with unspent money held in a separate account and used within a set period, and they required impact assessment for larger projects of larger companies. Administrative overheads remain capped at five per cent of CSR spending. That cap, combined with annual grant cycles, shapes the whole sector. An implementation partner allowed to recover so small a share of its costs for management cannot afford to build data systems, train supervisors or keep experienced staff. Funders then complain that NGOs lack capacity, and their own terms produced the shortage.
What outcome-based design needs
We would start with four requirements.
- Multi-year funding agreed at the start, so that a partner can hire and plan beyond a single grant cycle.
- A baseline and a measure that a sceptic would accept, paid for as part of the project and not squeezed out of the overhead cap.
- Implementation capacity treated as the thing being funded, including supervision, data and management time.
- Convergence with a government programme that will continue after the grant ends.
The last is the one most often missed, and education shows why. The learning gaps documented year after year by the ASER survey are not closed by one company's classroom block. The state already pays teachers and runs Samagra Shiksha, and NIPUN Bharat sets a national target for reading and numeracy in the early grades. CSR money is best used to make that spending work better: teacher coaching, remedial instruction inside the school day, assessment data that principals can use, and tested methods that a state can adopt. If a programme succeeds, the test is that the state takes it over. Skilling follows the same logic. Training that ends in a certificate has little value unless employers hire, and unless someone tracks the trainee a year later.
The best use of private social capital is to make public spending work better and then step back.
The objections, and where we land
Boards will say that outcomes are uncertain and that evaluation consumes money that could reach beneficiaries. Both are true, and neither justifies the current default. Spending the entire amount on delivery without learning what works wastes more over a decade than spending a measured share on finding out. Another objection is that companies should stay close to their own communities and operations. That is a fair preference for locality and it suits some programmes. It should not become a reason to scatter funds across many projects too small to measure.
Impact bonds show the logic at its extreme. The development impact bonds in Rajasthan, one on girls' schooling and one on maternal and newborn health, paid investors only when measured results were met. They were costly to design and cannot become the norm. Their lesson is smaller and more portable: write the outcome into the contract, pay for it to be measured and let the implementer adjust midway. A board can do that without a bond.
What we are watching is whether boards begin to approve CSR plans on five-year outcome targets, and whether impact assessments produce results that are published. The open problem is attribution. When a company, a state programme and an NGO all contribute to a gain in reading, someone has to pay for the evaluation that says which part worked, and today nobody is obliged to.