Households in the middle of Pakistan’s consumption distribution spent 5.3 per cent less per person in real terms in 2024–25 than their counterparts did in 2018–19. Real spending on food fell by 12 per cent, on clothing by 18, on schooling by 11 and on recreation by 36. Housing and utilities provided the largest offset to these declines, up 19 per cent on the broad housing price index and 27 per cent when its components are deflated separately. Neither figure is firm. The two survey rounds record rent differently, for reasons the survey documentation does not explain, and the price index weights housing differently from the survey.
“Middle” here means the third fifth of households when every household is ranked by consumption per person, not a “middle class” defined by income, occupation or assets. These are two snapshots of the middle of the distribution, not the same households followed over time. The analysis uses the Pakistan Bureau of Statistics (PBS) Household Integrated Economic Survey (HIES) for each year.
The average middle-consumption household in 2024–25 consisted of 6.1 members, down from 6.4 in 2018–19. Its monthly consumption spending was Rs 63,000, up from Rs 30,531. In nominal terms, spending roughly doubled, but after adjusting for inflation, it fell by 9.3 per cent per household. Allowing for the smaller household size, the fall per person was 5.3 per cent. The lead story in this issue, Who pays for the state?, follows the taxes carried on the same purchases.
In 2018–19, most of the household budget already went to core necessities. For this analysis, necessities are food, housing and utilities, and health. By 2024–25 they took 2.2 percentage points more of the budget than six years earlier. The question is how a smaller real budget was divided.
What gave way
Figure 1 shows the budget shares in each round. Housing and utilities, including fuel, saw the largest increase in budget share, up 1.6 percentage points from 20.5 per cent in 2018–19 to 22.1 per cent in 2024–25. Food, which already claimed 41.8 per cent of the budget in 2018–19, rose by half a point to 42.3 per cent. Together, these categories take up roughly two-thirds of the household budget for middle-consumption households, so small changes in their shares move a lot of rupees.
Health showed a more modest increase of 0.1 percentage point. Restaurants, tobacco and miscellaneous goods also gained a little share, though real spending on all three fell. At the same time, clothing’s share of spending declined by 1.4 percentage points, education’s by 0.7 and recreation’s by 0.4. Together, the other spending categories lost the 2.2 points that necessities gained.
Shares measure only how the budget is divided. They are measured in the rupees of the day, so a category’s share can rise even when households buy less of it, if its price rises faster than the overall basket. Deflating each category by its own price index shows how real spending changed (Figure 2). Real spending figures here are per person unless otherwise stated.
Food, already the largest category, gained half a point of share even as real spending on it fell by 12 per cent. Housing and utilities behaved differently: the category had the largest increase in real spending of any category shown, 19 per cent, alongside its 1.6-point gain in share. That figure should be read with care, as HIES combines rent, estimated rent for owner-occupied homes and utility bills differently from the consumer price index (CPI) for housing used to adjust those costs for inflation. The increase therefore should not be read as households consuming 19 per cent more housing or energy. Deflating the housing components separately leaves the increase positive but changes its size, and the two rounds also differ in how rent is reported, for reasons the survey documentation does not explain. Real spending on health rose by about 5 per cent, with a 0.1-point gain in share.
Clothing, education and recreation are a different story, declining both in real and in relative terms. Real spending on clothing fell by 18 per cent, and its budget share by 1.4 points. Real spending on education fell by 11 per cent and its share by 0.7 points. Total education spending divided by the number of children aged 5 to 16 fell by 9.4 per cent in real terms, so the decline is not simply the result of fewer children per household. This measure includes all education spending, not just spending on children aged 5 to 16. The figures do not establish that a larger budget share devoted to necessities caused households to reduce education expenditure. But the education decline sits alongside a broader shift in which essentials absorbed more of the household budget.
Recreation’s share fell by only 0.4 points, but it had the largest fall in real spending of any category, 36 per cent. No standard errors are reported for these real changes, so the size should be read as approximate, though the fall is between 35 and 37 per cent under every alternative definition of the middle.
Figure 3 puts the categories together. The three necessities gained 2.2 points of the budget while real spending on them fell by 2.1 per cent, and the other categories lost those 2.2 points while real spending on them fell by 11.7 per cent. The smaller fall in necessities comes from housing, while food and health together fell almost as far as the other categories. Food and health together fell by 11.2 per cent, and real spending per person on everything except housing and utilities fell by 11.4 per cent.
The housing figure also carries the largest uncertainty in the analysis. Deflating its components separately raises its real increase to 27 per cent, turns the 2.1 per cent fall in necessities into a 0.3 per cent rise, and shrinks the fall in total consumption from 5.3 to 3.7 per cent. The fall in living standards survives that check. However, the claim that real spending on necessities fell does not. The difference in rent reporting between the two rounds is present in the released survey data rather than introduced in processing, and the survey documentation does not explain it, so both housing estimates should be treated with caution.
How comparable are the surveys?
The shift toward necessities was almost identical in urban and rural third-quintile households: 2.1 points in urban households and 2.2 in rural. The fall in real spending differed more. Under the main housing adjustment, real spending on necessities fell by 2.9 per cent in urban households and 1.7 per cent in rural households. The point estimates for the necessities share also rose in all four provinces, from around 0.6 percentage points in Balochistan to 2.8 in Punjab, though the intervals for Balochistan and Khyber Pakhtunkhwa include zero. The rise in the necessities share is present across the rural-urban divide and, on the point estimates, across provinces.
The two HIES surveys are not perfectly comparable. HIES 2024–25 was based on a sampling frame redrawn after the 2023 Population Census, while HIES 2018–19 used an older frame. That matters because some apparent changes can arise from changes in which households and places are represented by the survey. One visible example is the urban share of the third quintile, which increased from 30.4 per cent in 2018–19 to 37.5 per cent in 2024–25. This could reflect growth in the urban population, the new frame’s reclassification of places as urban, or a shift in where urban households sit in the national distribution. The two surveys cannot tell these apart or show particular households moving down.
The rise in the necessities share is present within both urban and rural households, so it cannot be explained solely by the changing balance between the two groups. That does not rule out changes in who is counted within each group, or effects of the new frame.
There is a second complication in comparing the two survey years. HIES 2018–19 was collected on paper, while HIES 2024–25 was the first fully digital round. This change in the data collection method can affect item reporting, interviewer errors, and recording of small expenditures. Any reporting effect created by the new collection mode cannot be separated cleanly from genuine changes in household spending between the two periods.
Does this change the argument?
The definition of the middle can be tested, even if the collection mode cannot. Ranking households by adult-equivalent rather than per-person consumption, widening the group to the middle 60 per cent, restricting the analysis to the items most directly comparable across the two rounds, or dropping households close to the quintile cut-offs all leave the necessities share up by 2.2 to 2.4 points. With the main housing deflator, real spending on necessities falls by between 1.4 and 2.3 per cent under these checks. The figures are in the method note.
None of these tests can remove the break between the two survey rounds, particularly any reporting effect of digital collection. They do make it hard to attribute the result to one definition of the middle, to borderline households or to a looser set of expenditure matches. The housing deflator changes the conclusion about real spending on necessities: using separate price indices for its components turns a small fall into a small rise. Total real spending still falls under both calculations.
The pressure on households in the middle is not visible in their rupee spending, which roughly doubled. Measured in what those rupees bought, they had between 3.7 and 5.3 per cent less per person, depending on how housing is deflated. Real food spending fell by 12 per cent even as food’s share of the budget rose, because food prices rose faster than the overall basket. Clothing and schooling fell with it. Housing and utilities provided the largest offset among the categories shown, and whether that is more housing consumed, higher bills or a mismatch between the survey aggregate and the price index used to deflate it, the survey documentation cannot establish. The size of the decline is uncertain. Its reach across the budget, food included, is clear in these estimates.
This analysis compares HIES 2018–19 and 2024–25 for households in the national third consumption quintile, ranked separately in each round by per-capita consumption using population weights. The third-quintile samples contain 5,008 households in 2018–19 and 5,884 in 2024–25.
Expenditure items were harmonised across surveys and grouped into common categories. Budget shares are calculated from weighted aggregate spending. Real expenditure is deflated using official PBS category-specific CPI, matched by interview month and urban/rural residence, and expressed in 2015–16 prices. Real figures are per capita, unless otherwise stated. Total real per-capita consumption, calculated by summing expenditure deflated separately for each category, fell by 5.3 per cent between the two rounds. Communication is excluded from Figure 2 because the size of its real-spending increase depends on which expenditure items are matched, 60.1 per cent under the broad specification and 109.8 per cent under the strict one. Both are positive, though its budget-share change switches sign, from −0.11 to +0.06 percentage points. Education spending per school-age child is total third-quintile education expenditure divided by the number of children aged 5 to 16; it does not isolate spending on those children.
Comparability is imperfect because 2024–25 used a new post-2023 census sampling frame and digital data collection, while 2018–19 used the previous frame and paper questionnaires. Results are therefore descriptive, not causal.
Robustness checks on the definition of the middle: ranking by adult-equivalent consumption, necessities share +2.4 points and real spending −1.9 per cent; the middle 60 per cent of the distribution, +2.4 and −1.4; the strict common-item expenditure universe, +2.4 and −2.3; the third quintile with households near its cut-offs removed (retaining about 93 per cent of the weighted sample), +2.2 and −2.1. For housing and utilities, alternative component-level deflation checks also leave the measured real-spending increase positive, at 26.8 per cent compared with 19.0 per cent under the headline specification, although the two rounds contain an unresolved difference in owner-occupier rent reporting, and the composition of third-quintile housing expenditure does not exactly match the CPI housing weights. The rent difference is present in the released HIES data rather than introduced in processing, and the survey documentation does not explain it. Carried through to the totals, the alternative housing deflator turns the 2.1 per cent fall in real spending on necessities into a rise of 0.3 per cent and reduces the fall in total real consumption from 5.3 to 3.7 per cent. The overall decline survives; the direction of the necessities change depends on the housing deflator.
Ghufran Khalid is an economics graduate from NUST and an applied researcher working across economic, financial and policy analysis. He works in equity research and writes independently on Pakistan's economy, with a particular interest in industrialisation, structural change and data-driven economic research. LinkedIn.