Can an economy become statistically more visible without becoming better understood?
It is a question India should be asking as it prepares to modernise the machinery through which it measures economic activity. The Union Cabinet’s approval of the Rs 4,371.10-crore Viksit Bharat Initiative for Statistical Transformation, Analytics and Research (VISTAR) on November 4 is being presented as an important step towards a more sophisticated statistical system. It is. But there is a larger issue buried within this technological transformation, one that deserves considerably more attention.
The Government intends to bring artificial intelligence and machine learning into statistical processing, while using alternative sources of information such as mobile-phone records, retail scanner data and e-commerce transactions alongside conventional surveys. It also plans new economic indices and more granular estimates of activity across districts and cities.
The ambition is understandable. A country of India’s scale cannot depend indefinitely on statistical systems that struggle to capture the speed and complexity of a changing economy. But the arrival of better instruments does not necessarily mean a better understanding of what those instruments are measuring.
Consider a simple distinction. A transaction conducted through a digital payment platform leaves a record. A similar transaction conducted in cash may leave almost none. A delivery worker whose activity is tracked by an application generates a continuous stream of information. A person earning through several informal occupations may remain largely invisible to the same system. Both participate in the economy. Only one leaves a convenient digital footprint.
This creates a peculiar danger. The economy that generates the most data can gradually become the economy that policymakers understand best, even when it does not represent the whole.
The distinction is particularly consequential for India. Its economic activity is spread across formal companies, family-owned establishments, small manufacturers, street vendors, agricultural workers and countless informal enterprises. Their contribution cannot always be inferred from the behaviour of businesses that maintain formal accounts or operate through organised supply chains.
In fact, this is not merely a theoretical concern. Research published by the International Growth Centre in March 2026 examined a longstanding problem in India’s national accounts: the estimation of informal-sector output using the growth of the formal sector as a proxy. The assumption that smaller, unincorporated businesses move in tandem with organised enterprises can become particularly unreliable when economic conditions change unevenly. If one part of the economy expands while another contracts, a proxy can conceal the difference.
The problem is not that statistical estimation is inherently unreliable. Every economy requires estimates, classifications and assumptions. The problem arises when an assumption gradually acquires the authority of an observed fact.
There is an even more uncomfortable implication. Measurements do not always remain passive descriptions of reality. Once institutions begin making decisions around particular indicators, people and organisations adapt their behaviour to those indicators. Economist Charles Goodhart’s observation, commonly known as Goodhart’s law, captures this difficulty: when a measure becomes a target, it often ceases to be a good measure. A performance indicator designed to describe an institution can eventually influence what that institution chooses to perform.
Imagine a hypothetical credit programme that evaluates small enterprises largely through their recorded digital transactions. Businesses with extensive digital records would appear more economically legible than equally productive businesses operating predominantly in cash. If access to institutional support increasingly depended on such indicators, enterprises could face pressure to optimise their recorded activity rather than simply improve their productivity.
The statistical system would then do something more consequential than misdescribe the economy. It would begin influencing the incentives within it.
This is where India’s proposed technological transformation deserves scrutiny. Artificial intelligence can identify patterns in enormous datasets, detect inconsistencies and reduce the time required to process information. But it cannot independently establish whether the information supplied to it represents the economic reality it is supposed to describe. A sophisticated model can produce an extraordinarily precise answer to a poorly framed question.
The issue is not whether machines can count faster than human beings. They plainly can. It is whether the categories, assumptions and underlying data used by those machines are sufficiently representative. There is a useful lesson here from the philosophy of scientific measurement. A measuring instrument does not offer an unrestricted view of the world. It records particular properties through a defined method, with specific limitations and uncertainties. Scientists therefore distinguish between the precision of an instrument and the accuracy of the conclusion drawn from its readings. Economic statistics deserve the same intellectual discipline. More observations do not automatically eliminate uncertainty. More granular datasets do not necessarily produce more representative conclusions. And greater computational power cannot compensate for a flawed understanding of the phenomenon being measured.
The implications extend beyond GDP. Consider employment. A single headline figure can conceal differences between secure salaried work, temporary contracts, unpaid family labour, gig work and multiple forms of self-employment. Similarly, rising consumption expenditure does not necessarily tell us whether households are becoming financially secure or merely spending more to maintain their standard of living.
An economy is not a single number. It is a collection of relationships between production, consumption, employment, capital, technology and human welfare. Reducing these relationships to a handful of indicators may be necessary for policymaking, but treating those indicators as the economy itself is a mistake.
There is also a question of accountability. As Governments increasingly use administrative and alternative datasets, the public must be able to understand how economic indicators are constructed. Definitions, estimation methods, revisions and sources of uncertainty should not disappear behind technological complexity. Statistical independence and methodological transparency are not obstacles to modernisation; they are what make modernisation credible.
VISTAR could provide India with an opportunity to build precisely this kind of statistical infrastructure. Its proposed expansion of surveys, development of district-level estimates and use of alternative data sources could address longstanding gaps. But its success should not be judged simply by how much information the government collects, how quickly algorithms process it or how many new indicators are introduced.
The more important test is whether the system can distinguish between what is visible and what is representative, between a useful proxy and an actual observation, and between a number that describes economic activity and one that merely rewards a particular kind of activity.
India’s development ambitions will increasingly depend on its ability to make decisions using vast quantities of information. That makes the integrity of its measurement systems a question of economic governance, not simply statistical administration. The country does not need an economy that looks better in its databases. It needs a statistical system capable of seeing the economy as it actually is, including the parts that do not leave a digital trail.
Because the greatest danger in an age of artificial intelligence may not be that machines miscalculate the economy. It may be that they calculate it perfectly according to assumptions that were wrong from the beginning.
The author is a physicist at the University of North Carolina at Chapel Hill and an opinion contributor to The Wall Street Journal; Views presented are personal.