India's data collection needs to go local to capture finer details

As the economy becomes more regionally differentiated, national and state averages are no longer enough to track growth

India’s gross domestic product (GDP) numbers are back in the debate — over methods, revisions, and what different indicators are really telling us about the economy.

 

Such debates are not unusual. GDP has never been a neat, singular truth, but a constructed measure pieced together from partial and sometimes conflicting signals. At the national level, these fragments can still be assembled into a coherent picture. At finer geographic levels, they begin to fray.

 

Yet this is where attention is shifting. Growth in India is uneven, clustered, and concentrated in specific corridors — technology hubs such as Bengaluru and Hyderabad, manufacturing centres in Tamil Nadu and western India, and expanding urban regions around Delhi. A handful of regions account for a disproportionate share of new investment and formal job creation. Recent Budget statements have explicitly framed cities as “economic regions”, signalling a stronger policy focus on local economies.

 

But while the framing has moved forward, the ability to observe these dynamics has not kept pace. Part of this difficulty is structural in nature. In more formal economies, local activity shows up in dense administrative records — tax filings, payroll data, firm accounts — that allow for regular and granular estimation. In India, however, a large share of output and employment remains informal, with enterprises that do not report earnings systematically. This increases reliance on surveys and approximations, which are not designed for consistent local measurement.

 

As a result, India’s statistical architecture is strongest at the level of national and state accounts, where these limitations can be smoothed over. At finer spatial levels, however, it runs into constraints and often falls back on proxy-based allocations and top-down assumptions. District GDP estimates, for instance, are typically derived from state-level aggregates using simplifying assumptions rather than direct measurement. These methods produce neat numbers, but often miss underlying spatial dynamics.

 

Efforts are underway to improve this. The new GDP series incorporates richer survey and administrative data and is expected to strengthen regional estimates. This is an important step forward. But even as data improves, the challenge shifts to how these different sources — surveys, administrative systems, and newer datasets — are interpreted and combined to reflect local economic activity.

 

Large surveys such as the Periodic Labour Force Survey, the Annual Survey of Industries, and the Annual Survey of Unincorporated Enterprises provide insights into employment and enterprise activity. But they are not designed for local analysis. As one moves to districts or cities, sample sizes become too small and estimates lose reliability. Entire segments, particularly services and construction, remain only partially captured, often with time lags.

 

Some states are beginning to expand survey efforts, for instance through state-level adaptations of surveys like the PLFS, but the challenge persists. As one moves to more granular levels, the sample sizes required for reliable estimates increase sharply, placing pressure on already constrained surveying capacity in terms of enumerators, funding, and implementation. Surveys cannot keep scaling indefinitely, making it necessary to complement them with more robust statistical techniques to estimate local economic activity even with limited samples.

 

Administrative data is increasingly being used to fill these gaps. India generates vast amounts of it through goods and services tax (GST) filings, payroll systems, electricity distribution companies, and transport networks. However, much of this data is not readily accessible or usable for economic analysis at the state level, making it harder to use consistently for local estimation.

 

But here the challenge is different. It is not too little data, but too much without a clear way of interpreting it. These datasets are collected for taxation, compliance, or programme delivery, not economic measurement. GST captures transactions, not necessarily where value is created. Payroll reflects formal employment. Electricity data reflects a mix of industrial and household demand shaped by pricing and behaviour.

 

What these datasets offer are not measures of the economy, but traces of it. This has led to growing interest in alternative data such as digital payments and satellite imagery. These are granular and high-frequency, and seem to offer a way of observing the economy as it evolves.

 

But there is a risk of mistaking visibility for understanding. Night-time lights, for instance, indicate intensity but do not map cleanly to value added, and tend to plateau in developed areas. Used uncritically, they can create a false sense of precision. The risk is that what is easiest to measure begins to stand in for what actually matters.

 

What is emerging is, therefore, not simply a data gap, but a problem of interpretation across multiple sources. The task then is not to measure everything, but to systematically interpret and combine these fragments into a coherent view of the local economy. This means investing in better ways of linking signals, using structured estimation to infer where and how economic activity is evolving.

 

The aim is not perfect precision. It is to generate estimates that are good enough to guide decisions, capturing where activity is intensifying, where it is shifting, and which regions are emerging as centres of growth.

 

This matters because policy is increasingly being shaped at this level. And in India, what appears “granular” is anything but small. A district such as Pune has a population comparable to that of Switzerland, while Nashik is similar in scale to Denmark. These are large economic regions, yet remain only partially visible in current data systems.

 

Without a clearer view of local economies, the shift towards place-based policy risks remaining incomplete. As the economy becomes more regionally differentiated, national and state averages will tell us less about where growth is taking place. The choice is not between precision and approximation, but between developing a more grounded understanding of local economic change or continuing to plan with partial visibility.

 

The author is with Artha Global