What the Consumption Gap Measures

India’s 2022–23 and 2023–24 household surveys against one national-accounts vintage

In current rupees, rural consumption per person grew slightly faster than the nearest national-accounts measure. Ordinary rural monthly consumption expenditure per person rose 9.25% between the August 2022–July 2023 and August 2023–July 2024 household surveys; private final consumption expenditure (PFCE) grew 8.89% per person between the April–March financial years 2022–23 and 2023–24. After each series’ official price adjustment, the ordering reverses: 3.54% for the rural survey against 5.28% for PFCE, a difference of 1.75 percentage points. [1, 2]

That reversal changes how we should read a claim that national growth failed to reach households. GDP measures the value of final production across the economy, including output that households do not consume. PFCE is a closer consumption benchmark, though it includes nonprofit institutions serving households and values some services differently from the Household Consumption Expenditure Survey (HCES). The survey also measures a different population and set of items. We can compare their published growth rates; those rates alone cannot tell us how many families became better off.

The adjacent HCES rounds both ran from August to July and used broadly the same three-visit design and questionnaires, subject to stated changes in the household questionnaire. That offers a closer comparison than a long leap across the 2022–23 survey redesign. The rounds are repeated samples rather than a panel following the same families over time. The national figures come from one edition, the August 2026 National Accounts Statistics with base year 2022–23. Its April–March years still give the two series different endpoints. [1, 2]

Scroll horizontally to see all columns.

Sources: MoSPI National Accounts Statistics 2026, Statement 1.2 and HCES Report 592, Table 3.3. NAS levels are rounded for display; HCES levels are already published in whole rupees. Growth rates are therefore approximate, and percentage-point differences use calculations before rounding the displayed rates. PFCE and HCES differ in period, population, coverage and price adjustment. The table does not compare their rupee levels as equivalent welfare measures.

Real GDP per person grew 6.36%, against 3.54% for rural and 3.48% for urban survey consumption. GDP includes investment, government spending and net exports. Using PFCE instead reduces those growth differences to 1.75 and 1.81 points. National net income per person grew 6.68% in real terms, but it, too, is no measure of the survey households’ income. [1]

Which consumption is being counted?

“Survey spending” is an easy phrase with a misleading edge. Ordinary HCES monthly per capita consumption expenditure, or MPCE, already includes values assigned to home-produced goods, gifts and certain other receipts in kind. It is neither cash spending nor a complete measure of welfare. The report also publishes MPCE with the value of selected free welfare items added. Under that definition, real rural MPCE rises 4.28% and real urban MPCE 3.70%. The PFCE differences then narrow to 1.00 and 1.59 points. The extra-item series does not additionally impute the value of free health and education services. [2]

The extra-imputed series captures specified free items that ordinary MPCE leaves out. It still differs from PFCE’s wider collection of market and imputed consumption. Neither variant wins simply because it gives a smaller or larger gap. The rural growth comparison moves appreciably with the survey definition; that sensitivity is part of the finding.

The national accounts also have an institutional boundary absent from a simple “household bill” reading. PFCE includes nonprofit institutions serving households as well as households. It includes owner-occupied housing services and a measured service component of financial intermediation. HCES has its own valuation rules; the appearance of an imputed-rent question on its questionnaire does not establish how much of that item enters headline MPCE. These categories cannot be subtracted from one another using the published headline means. The current MoSPI sources and methods show that PFCE is compiled from several kinds of data, including survey information, production estimates and price measures. [3]

HCES inputs help compile parts of PFCE, including specified consumption rates, inter-survey growth, some prices and seasonal allocation. Agreement between the series is therefore not independent confirmation of household recovery; disagreement does not establish a survey error. PFCE also combines rural and urban populations. A change in their shares can alter national per-person growth even if neither group’s own growth rate changes. The published headline rates do not reveal how much, if any, of this particular difference comes from composition. Nor do they supply an item-by-item bridge between the accounts and survey.

Why the real ordering reverses

The nominal comparison puts rural ordinary MPCE growth at 9.25%, urban MPCE at 8.31% and PFCE per person at 8.89%. Rural survey growth leads PFCE by 0.36 point; urban survey growth trails it by 0.58 point. Under their respective official price adjustments, both survey domains trail PFCE. The implied price changes explain the switch in ordering: about 5.52% for rural MPCE, 4.68% for urban MPCE and 3.43% for PFCE. [1, 2]

HCES uses rural and urban consumer price indices to report real MPCE. The national accounts use several deflators and volume methods for PFCE. Different base years, 2011–12 for HCES and 2022–23 for PFCE, do not themselves explain the growth contrast; the baskets and their price movements between observations matter. [2, 3]

Apply PFCE’s implied price change to the nominal growth in survey MPCE as a diagnostic. Rural survey growth becomes 5.63%, above PFCE’s 5.28%; urban survey growth becomes 4.72%, below PFCE by 0.56 point. Giving the series a common price change preserves their nominal-growth ranking by arithmetic. It does not create a corrected HCES estimate: rural and urban consumers need not face PFCE’s basket or price movement, and PFCE has a different institutional scope. The exercise shows why the real household comparison depends on the price measure; it does not show that either official deflator is wrong.

There is also a calendar problem. The survey’s August–July fieldwork years overlap two fiscal years each. Using the August 2026 quarterly PFCE figures, weighting quarters by the fraction of their months in each survey window, and interpolating population gives roughly 5.11% real PFCE growth per person. This is 0.18 point below the fiscal-year figure. It leaves a difference of about 1.57 points against ordinary rural MPCE and 1.63 against ordinary urban MPCE. The calculation assumes an even spending flow within each quarter; it is not an official August–July PFCE series. HCES items also use recall periods of 7, 30 or 365 days, so survey fieldwork dates do not define one clean consumption window. The timing adjustment and the imposed common price change are separate probes, not pieces of a decomposition that add up to an explained gap. [1, 2]

What the remaining difference can say

The official point estimates do show a difference. The nearby surveys used broadly comparable procedures; real MPCE rose in both rural and urban India, but by less than real PFCE per person. Adding the selected free items narrows both contrasts without eliminating them. The 1.75- and 1.81-point differences deserve an explanation.

The rival interpretation begins with the nominal rates: rural survey growth already edges above PFCE. A common-price diagnostic restores that ordering, and an approximate calendar shift trims the PFCE benchmark. Coverage, population and measurement error remain unresolved; the two systems also share some inputs. A properly harmonised comparison could still find a material gap. The published difference cannot be counted as consumption that households failed to receive.

Which households improved is a separate question. Both rural and urban real MPCE means rose, so a literal claim that the survey’s per-person mean did not grow conflicts with these point estimates. The means do not show what happened below the average, across regions or to people facing particular prices. PFCE reports an aggregate rather than the household distribution, and the survey rounds do not form a panel following the same families. HCES reports design-based uncertainty for the 2023–24 means, but that alone gives no confidence interval for the difference in growth across systems and years. Two decimal places do not establish statistical significance. [2]

Earlier research has warned against treating these measures as interchangeable. Deaton and Kozel examined India’s survey design and national-accounts concepts; Prydz, Jolliffe and Serajuddin showed how assumptions about unrecorded consumption can alter poverty and inequality conclusions across countries. Kumar and Seth recently examined India’s PFCE–HCES level and category gap. This comparison is not new. The contribution here is to put an adjacent pair of survey waves beside one national-accounts vintage, then change the benchmark, denominator, imputation, price convention and approximate calendar window in view of the reader. None of that earlier work supplies a ready-made adjustment for these years. [4, 5, 6]

To test for a material growth gap on comparable terms, a future study would need a category and valuation bridge, aligned populations and time windows, a defended price comparison, and uncertainty for the resulting growth difference. That formal test would first need uncertainty for the previous survey round, then account for covariance across rounds and other measurement errors. A claim that particular households were left behind would need distributional or longitudinal evidence of its own.

India’s published point estimates show real consumption growth in both rural and urban HCES, slower than PFCE per person under the official calculations. The rural ordering changes under the common-price diagnostic, while the calendar exercise modestly changes the size of both contrasts. Until prices, populations, coverage and uncertainty are reconciled, the difference between these systems cannot be assigned as a loss to households. The next measurement bridge could find a material gap. These headline rates alone cannot tell us its size.

Sources

[1] MoSPI, National Accounts Statistics 2026, released 31 August 2026, base 2022–23: Statement 1.2, Statement 1.14, and quarterly Statement 8.18.1.

[2] MoSPI, HCES 2023–24, Report 592, especially Tables 3.3 and 3.18 and Appendix C.

[3] MoSPI, Sources and Methods for Compilation of National Accounts Statistics, 21 September 2026, chapter 21 and Annexes II and VII.

[4] Angus Deaton and Valerie Kozel, “Data and Dogma: The Great Indian Poverty Debate”, World Bank Research Observer 20(2), 2005, pp. 177–199.

[5] Espen Beer Prydz, Dean Jolliffe and Umar Serajuddin, Mind the Gap, World Bank Policy Research Working Paper 9779, 2021.

[6] Ashish Kumar and Payal Seth, “Bridging the Consumption Gap in India’s Statistics”, Economic & Political Weekly 61(5), 31 January 2026, pp. 22–25. Cited for prior-art scope; its numerical tables are not used here.