Warning that the revamp may be costly, statistically unnecessary and potentially harmful to the credibility of inflation measurement, a technical report released by IIT Bombay has raised serious concerns over the Union government’s proposed changes to the way housing services are measured in India’s consumer price index (CPI).
The report, "Consumer Price Index: Methodological Aspects of Measuring Housing Services", authored by Praggya Das, former adviser-in-charge (monetary policy department) at the Reserve Bank of India (RBI), and Prof Ashish Das of IIT Bombay’s department of mathematics, critically examines proposals put forward by the Union ministry of statistics and programme implementation (MoSPI) as part of the CPI base revision from 2012 to 2024.
Housing is among the most consequential components of the CPI basket, carrying a weight of 10.07% in the all-India index and a much higher 21.67% in the urban CPI. Given its direct influence on measured inflation and, by extension, RBI’s monetary policy decisions, MoSPI had issued a separate discussion paper proposing major changes to how housing services, largely house rents, are captured.
While the IIT Bombay report acknowledges several positives in the broader base revision exercise, it strongly challenges the premise that the existing housing index methodology is flawed.
At the heart of MoSPI’s concern is the current panel-based system of rent collection, under which rental data are collected monthly from one-sixth of the sampled dwellings, with each dwelling revisited every six months. MoSPI has attributed periodic six-monthly dips in housing inflation to this chain-based panel approach, describing these movements as ‘unexplainable’.
The IIT Bombay report firmly rejects this assertion, demonstrating through formal statistical proof that the existing chain-based index is transitive and mathematically equivalent to a fixed-base index. In simple terms, the methodology itself cannot generate rhythmic distortions. If such dips are observed, the authors argue, they are more likely the result of data handling issues, computational errors or one-time implementation decisions, rather than any inherent flaw in the method.
The report highlights that the sharp spike in housing inflation observed in FY13–14, which continues to influence perceptions of the housing index, resulted from an administrative decision to maintain the index at 100 for the first five months of the CPI series, despite partial rent data being available. This, the authors note, was an avoidable implementation choice rather than a methodological failure. Had the available data been used progressively, the spike and the subsequent inflation surge could have been moderated.
Against this background, the report is sharply critical of MoSPI’s proposal to collect rent data every month from the entire housing sample of more than 25,000 dwellings across urban and rural India. Such an approach, it argues, ignores the fundamental characteristics of housing markets, where rents typically change infrequently and are often governed by annual contracts. Surveying every dwelling every month would significantly increase costs, lead to respondent fatigue, overburden field staff, and risk undermining data quality, particularly if physical verification is replaced by proxy reporting. Crucially, the authors find no evidence that full-sample monthly collection would deliver any meaningful informational advantage over the existing rolling panel system.
Rather than dismantling a statistically robust framework, the authors suggest that if MoSPI faces constraints due to the limited availability of rented dwellings in smaller towns or rural areas, it could consider a reduced-panel approach, such as three panels instead of six. This, they argue, would continue to meet CPI’s monthly data requirements while materially reducing cost and administrative burden.
Where the report aligns strongly with MoSPI is in the treatment of employer-provided housing. Under the current CPI, rents for government and public sector accommodation are imputed using house rent allowance foregone plus nominal licence fees, a method that does not reflect market realities. This distortion became particularly evident during the implementation of the 7th Central Pay Commission, when housing inflation spiked sharply in states with a high concentration of government housing. Excluding employer-provided dwellings from direct rent measurement, while implicitly accounting for them through Census-based weights, is described as a necessary and welcome correction.
The IIT Bombay report also cautions against MoSPI’s proposal to replace geometric means with arithmetic means at the elementary index level for housing. Given the wide heterogeneity of housing markets, which span differences in dwelling size, location, and tenure, arithmetic averaging risks overstating inflation by giving disproportionate weight to extreme rent increases in a small segment of the market. International best practice, including guidance from the IMF-backed CPI Manual, favours geometric means for aggregating price relatives at the elementary level, with arithmetic means reserved for higher levels of aggregation.
In conclusion, the authors argue that while expanding housing coverage to rural areas and correcting distortions arising from employer-provided accommodation are sound reforms, abandoning a statistically proven panel-based system is not. The housing index, they stress, is too important to be reshaped based on perception rather than empirical evidence. As the report bluntly puts it, “If it ain’t broke, don’t fix it.”
With inflation data shaping interest rates, borrowing costs and household finances across the economy, the report serves as a timely reminder that sound policy-making depends not on collecting more data at any cost, but on applying rigorous statistical judgement to what is already available.
