ECB survey: households expect bigger bills

Households expect spending to rise by 3.7% and income by 1.1%. The ECB survey released on 23 October points to pressure on budgets and a darker economic outlook as energy makes bills heavier. Three- and five-year inflation expectations levelled off in September after rising in August. The consequences depend on the money households have available and the purchases they may put off.

A larger bill can conceal a smaller purchase. A household pays more while buying fewer items, switches to a cheaper product or leaves something else for another month. For a business, higher nominal sales therefore need not mean higher volumes. For the customer, spending and living standards can move in opposite directions.

September’s Consumer Expectations Survey (CES), gives that distinction immediate relevance. Respondents expect nominal spending to rise by 3.7% over the next twelve months, against income growth of 1.1%. The difference between the two indicators is 2.6 percentage points. It was also 2.6 points in August: each measure has increased by one tenth. The mismatch persists at the same size as in August. S01

That matters for the interpretation. A survey can reveal concern, identify constraints or illuminate a channel through which the economy may change. On its own, it cannot establish a realised loss of purchasing power. This investigation follows both the reported evidence and the choices it might encourage, keeping the two separate.

A small move in the most visible number

Median inflation expectations one year ahead rose from 3.0% to 3.1%. The three- and five-year medians stayed at 2.9% and 2.5% respectively. These are expected annual rates at different horizons. The five-year figure concerns price growth during the fifth year, rather than the cumulative increase over five years. S01 S04

The median locates the middle of a distribution and is resistant to extreme values. The ECB calculates a weighted, interpolated median: interpolation reduces the effects of rounded responses. The result summarises the expectations of the consumers surveyed. The release supplies no confidence interval with which to declare the one-tenth move statistically significant. S03 S06

Method and scope

Common 0–4% scale. Three- and five-year horizons refer to months 24–36 and 48–60, not cumulative inflation. The 2% reference is the ECB’s HICP target, not a target for the CES median. The release provides no confidence interval.

In September, the two longer-term medians levelled off while one-year expectations edged up. September’s stability follows an increase in August, when the three- and five-year medians rose from 2.7% to 2.9% and from 2.4% to 2.5%. Assessing anchoring takes a longer view. S18 An unchanged median could still coexist with a wider distribution or more very high answers, however. Assessing whether expectations are “anchored”, meaning resistant to short-term news at longer horizons, also requires evidence about that distribution and its sensitivity. S12

The ECB’s 2% target is a useful reference, with an important qualification. It applies to medium-term inflation measured by the Harmonised Index of Consumer Prices (HICP). It is not defined as a target for the median response to a survey. A difference between the two invites investigation; it does not mechanically measure lost credibility or determine the size of the next interest-rate decision. S13

Energy turns concern into an immediate expense

Observed prices provide a concrete backdrop. Eurostat reports annual inflation of 3.8% in September, up from 3.2% in August, for the 21-country euro area. Energy contributes 1.71 percentage points to that total. It accounts for roughly 9% of the basket’s weight, yet about 45% of the published annual inflation rate, using the rounded headline as the denominator. Energy prices themselves rose by 18.9% over the year. Weight, contribution and price change describe three different aspects of the same pressure. S02

The economic importance of energy extends beyond a conspicuous price. An expense that is difficult to postpone absorbs income previously available for something else. A necessary commute cannot be adjusted as readily as an optional outing, and a home cannot be insulated the moment a higher bill arrives. The ECB’s work on energy shocks describes this real-income channel and the unequal scope households have to adapt. S08

Method and scope

Energy weight: 90.3 per mille / 10 = 9.03%. Descriptive inflation share: 1.71 / 3.8 × 100 = 45%. Rounded contributions sum to 3.77 points, against the rounded 3.8% headline. The two 100% bars use different, explicitly labelled denominators.

This is a persuasive economic context, not an identification of the sole cause of the extra tenth in expected inflation. September’s survey fieldwork ran from 27 August to 28 September. Final September inflation data were published on 16 October, followed by the CES release on 23 October. Respondents therefore had different information available when answering and could not have known the final price index for the month. S01 S02

The levels require care too. The survey’s 3.6% median perception of inflation over the previous twelve months and Eurostat’s observed 3.8% differ in population, weighting and observation window. The former pools responses from eleven countries, while the latter measures prices across the 21-country euro area. Their numerical proximity does not demonstrate that consumers underestimate inflation by 0.2 percentage points. S01 S02 S03

Follow the cash left in the budget

Income and spending expectations are expressed in current euros. Spending growth may reflect prices, quantities, a different basket or some combination of these. A household expecting to spend more is not necessarily promising retailers more purchases in volume terms.

To calculate these averages, the ECB caps the most extreme responses. This treatment, known as winsorisation, limits their influence without removing the respondents concerned. S03

Method and scope

Population-weighted means of individual growth rates. August gap: 3.6 − 1.0 = 2.6 points; September: 3.7 − 1.1 = 2.6. This does not measure growth in aggregate totals or a decline in the saving rate or purchasing power.

A constructed example makes the arithmetic visible without attributing false precision to the survey. Consider a household with €2,500 in monthly net income and €2,250 in spending. Apply September’s two rates as illustrative assumptions: income would rise to €2,527.50 and spending to €2,333.25. Its monthly surplus would fall from €250 to €194.25.

Now keep the same income but start with €2,475 in monthly spending. The initial surplus is only €25. After the same changes, it becomes a shortfall of approximately €39.08. Neither household is a real respondent or an estimate of the average European household. The example isolates a mechanism: the starting buffer determines how much room remains after the same pair of growth rates.

Method and scope

Two fictional budgets, each with initial income of €2,500. Initial spending: €2,250 and €2,475. Illustrative application of +1.1% to income and +3.7% to spending, without an assumed monthly path. Balance = income minus spending. Exact changes are calculated before rounding to cents; common −€50 to €300 axis. No population share or average outcome is inferred.

The second household would have to adjust elsewhere: cut another expense, earn more, draw on a reserve or borrow where credit is available. The path of prices alone does not determine which option is feasible. Liquid assets, existing commitments and the ability to change income all matter. Earlier ECB research documents substitutions between products, shopping around, using savings and changing work patterns. Those findings concern earlier episodes; they establish plausible channels without measuring the decisions made in September 2026. S08 S14

The example must not be mistaken for an estimate of national saving. The 1.1% and 3.7% figures are weighted averages of individual growth rates with extreme values capped. They are not growth rates for two aggregate accounting totals that can simply be subtracted. Likewise, deducting the 3.1% inflation median from the 1.1% average income expectation would mix statistics and spending baskets. Estimating a real-income loss requires consistent household-level observations or compatible aggregates. S03 S04

That distinction is economically important, rather than a statistical nicety. Suppose a high-income household has most of the additional saving while several lower-income households borrow or cut purchases. The overall saving total could look comfortable while a substantial group experiences pressure. Establishing whether that is happening in this survey requires the underlying joint distribution; the published averages cannot settle it.

Income shapes the outlook

The bottom of the income distribution expects slightly faster nominal income growth than the top: 1.5%, compared with 1.2%. In isolation, this might suggest a catch-up. But a higher percentage applied to a smaller income can still produce fewer additional euros, while the rest of the survey reveals a more anxious assessment of economic risk. S01

Expected unemployment twelve months ahead is 13.6% among respondents in the bottom quintile and 9.7% in the top. Expected mortgage rates are 5.7% and 4.4%, respectively. These are beliefs about market conditions, not unemployment rates among the respondents themselves or mortgage offers they have actually received. The differences document socially uneven expectations; they do not establish discriminatory bank pricing. S01 S04

Method and scope

Household net disposable-income quintiles are country-, wave- and module-specific. Three separate scales, each starting at zero. Unemployment and mortgage rates are market expectations, not personal unemployment or received loan offers. Growth-rate differences are not euro-income differences.

Income quintiles rank respondents by household net disposable income using thresholds specific to each country, wave and survey module. The bottom group therefore has no single Europe-wide income cut-off, and the classification measures income rather than wealth. A household may have a low current income and own a home, or a high income and substantial payment commitments. The grouping is useful precisely when its scope remains clear. S03

Economic anxiety also concerns the position people expect to occupy in the future. A darker national outlook may change a purchase decision even while personal income remains intact. The next analytical step is to find out whether concern turns into postponement, substitution or additional saving, rather than infer actions automatically from opinions.

National growth expectations fall to −1.5%

Expected economic growth fell from −1.2% to −1.5%. A tempting headline would present this as consumers forecasting a European recession of that size. The questionnaire instead asks about the respondent’s country of residence over the next twelve months. The release then aggregates those answers. A population-weighted average of national expectations is not a forecast of euro area GDP, whose aggregation follows different rules. S01 S03 S04

Unemployment presents another interpretative trap. The 11.3% expected in a year should also be considered alongside the 10.6% currently perceived, reported in the same results. The forward difference between these two survey indicators is 0.7 points. The release describes a broadly stable labour-market outlook. That institutional wording does not erase the pessimism; it reveals the comparison being made. Setting 11.3% directly against an official unemployment rate without distinguishing perception from observation would produce a different account. S01

The direction still matters. A weaker view of the economy can make a large commitment less appealing, particularly where little cash is left over each month. The point estimate, however, is neither a recession probability nor the respondent’s personal probability of losing a job. Those quantities require different questions and analysis.

A purchase postponed is an economic channel

September’s results do not yet measure the eventual effect on consumption. A separate ECB study published in 2026, using November 2025 responses, shows how financial uncertainty and spending intentions can interact. Among households finding their financial future difficult to predict, 53% planned to reduce usual spending, compared with 42% among the more certain group. The proportions planning to delay or cancel major purchases were 37% and 26%. S07

Method and scope

Groups compared in November 2025, published in Economic Bulletin 1/2026. Shares are conditional on group; multiple responses are possible. Eleven-point gaps are descriptive, not causal effects, September 2026 observations or mutually exclusive categories.

These percentages compare different groups and allow multiple responses. Their eleven-point differences do not establish that uncertainty alone caused eleven points of additional restraint. Nor does the study track the eventual spending of the September 2026 respondents. It contributes historical context, which should remain visibly separate from the new release.

A different study offers stronger causal evidence. In an experiment conducted in September 2020, researchers randomly assigned information about macroeconomic prospects to shift beliefs and uncertainty separately. After accounting for expected-growth changes, the authors estimate that higher induced uncertainty reduced spending reported in follow-up surveys. This design is more informative than a simple correlation, but its setting and information treatments rule out importing a coefficient unchanged into today’s circumstances. S09

There is also an opposing response. Expecting a higher price can encourage an earlier purchase when a household has the money and feels secure about its income. A research review published by the ECB emphasises this dependence on circumstances. The overall outcome reflects the appeal of buying before prices rise, the risk of inadequate future income and the practical ability to pay. The same increase in inflation expectations can therefore accompany very different spending decisions. S15

Method and scope

Analytical diagram based on research mechanisms, not a causal reconstruction of September 2026. Dashed arrows show conditional possibilities. Areas do not encode money, respondent counts or population shares.

Household budgets and the limits of the average basket

It would be too convenient to dismiss unease as a perceptual mistake. An ECB analysis published in 2025 compares self-reported labour incomes, deflated using national consumer-price indices, with perceptions of real income. It finds discrepancies and studies their relationship with consumption. That construction supplies a statistical benchmark. It does not reconstruct an individual housing bill, all the household’s commitments or its bank statement. S11

Several mechanisms could produce a gap. One respondent remembers prices before the inflation surge; another faces an expense rising faster than the national index; a third wants to rebuild a depleted reserve. Lower inflation leaves the price level already reached in place as long as prices continue rising. An improvement in the rate of change therefore need not make an expensive item affordable again. These mechanisms do not allocate September’s responses into precise categories, but they make a purely dismissive interpretation untenable.

On 20 October, ECB economists also stressed that slower food inflation leaves the higher price level in place, with energy pressures still passing through supply chains. These are the authors’ analyses, rather than a policy decision. S19

Attention to prices changes with the economic environment too. Experiments reported in Econometrica in January 2025 find that households and firms become more attentive to available information when inflation rises. An ECB blog analysis published in May 2026 examines how memories of earlier shocks interact with geopolitical conflict. This research supports the possibility of persistent vigilance without identifying the precise cause of September’s extra tenth. S16 S10

The distinction matters for public communication. Explaining an average accurately remains valuable, but households must still reconcile the explanation with their own constraints. A message about disinflation cannot by itself restore a depleted cash buffer. Equally, an individual experience does not replace a consistent price index. Both observations can be useful without asking either to answer a question it was not designed to measure.

A signal to test against subsequent behaviour

The CES reveals something national accounts cannot: how people see the months ahead before every decision has been taken. Its online panel comprises approximately 19,000 adults across eleven countries. Recruitment combines different approaches; internet non-users are excluded and people over 70 are under-represented. Weights correct observed compositional imbalances without guaranteeing that all bias disappears. The approximate panel size is not the exact respondent count behind every statistic. S01 S05

The strongest conclusion preserves both the risk and its limits. Expected spending continues to outpace expected income. Energy makes a substantial contribution to observed inflation, and opinions about the economy have weakened. Longer-term inflation expectations levelled off over the month. The scale of any adjustment in actual consumption volumes remains to be measured.

For the ECB, the question is whether the shock fades, spreads persistently to other prices or primarily weakens demand. For businesses, it is whether larger receipts reflect more purchases or higher prices. The next useful test is to compare intentions with consumption volumes and income outcomes, then check whether subsequent survey waves sustain the shift. The next CES release is scheduled for 27 November 2026. S01

September’s survey thus leaves a specific warning: households can expect to pay more while becoming less willing to buy. The strength of that signal will emerge in the choices households actually make and how long those choices persist.

Further reading

The distribution of household losses also shapes the consumption risks in France’s 2027 budget. Our investigation into oil, diesel inventories and cash needs examines the physical and financial constraints behind energy supply. The comparison with the ECB’s 2011 tightening provides historical context for monetary policy during an oil shock.

Sources and research method

Verification cut-off: 26 October 2026. Values were transcribed from primary publications and checked against their definitions. Figures distinguish September 2026, historical observations and the illustrative simulation. No interviews were conducted; microdata were neither downloaded nor re-estimated. The aggregate release cannot identify a causal effect specific to September. ECB research papers and analytical notes express their authors’ analyses and are distinct from monetary-policy decisions. S17

1. [S01] Banque centrale européenne / European Central Bank. ECB Consumer Expectations Survey results – September 2026. 2026-10-23.
   Source document
   Primary release read in full. Fieldwork: 27 August–28 September. Inflation medians; otherwise 2% winsorised means unless stated. No individual causal identification or quantified sampling precision in the release.
2. [S02] Eurostat. Annual inflation up to 3.8% in the euro area. 2026-10-16.
   Source document
   September HICP, 21-country euro area. Contributions are percentage points; weights are per mille. Rounded components sum to 3.77 points, against the reported 3.8% headline.
3. [S03] BCE / ECB. CES: Guide to the computation of aggregate statistics. 2026-03.
   Source document
   Interpolated medians, 2nd/98th percentile winsorisation, weights and country-specific quintiles. Text and selected PDF pages inspected. No re-estimation from microdata.
4. [S04] BCE / ECB. CES: Microdata guide and questionnaire. 2026, version 10.
   Source document
   Variable anchors: C4010/C4020 (national growth), C4030/C4031 (unemployment), C3220 (net income), C6110/C6120 (spending), C1210/C1220 and E2010/E2020 (inflation horizons). Rendered pagination varies; use variable IDs.
5. [S05] BCE / ECB. CES: Methodological guide. 2026-01.
   Source document
   Online panel; probability and non-probability recruitment varies by country. Internet non-users excluded and over-70s under-represented. Weighting does not guarantee removal of all bias.
6. [S06] BCE / ECB. Consumer Expectations Survey: frequently asked questions. Undated page.
   Source document
   Accessible documentation of the median, panel and timing. Wave-specific September dates take precedence over generic timing.
7. [S07] Maria Dimou, Maarten Dossche, Teresa Hütten, Georgi Kocharkov / BCE. Consumption and saving amid uncertainty: recent insights from the CES. 2026, Economic Bulletin 1.
   Source document
   53/42 and 37/26 are conditional respondent shares by financial uncertainty in November 2025. Multiple-response intentions; observational differences, not identified causal effects.
8. [S08] BCE / ECB. Energy prices and private consumption: what are the channels?. 2022, Economic Bulletin 3.
   Source document
   Analysis of real-income, precautionary and substitution channels. Used for mechanisms; no 2022 estimate extrapolated to September 2026.
9. [S09] Olivier Coibion, Dimitris Georgarakos, Yuriy Gorodnichenko, Geoff Kenny, Michael Weber. The effect of macroeconomic uncertainty on household spending. 2021-05, ECB Working Paper 2557.
   Source document
   Randomised information experiment in September 2020 with spending follow-up. Working paper reviewed; American Economic Review publication, 114(3), March 2024, 645–677: abstract and metadata checked. Follow-up spending is self-reported. No coefficient transplanted to 2026.
10. [S10] Coibion, Georgarakos, Gorodnichenko, Kenny, Meyer, Pairan / ECB Blog. Geopolitical risk and scarring effects on consumer expectations: insights from the wars in Ukraine and Iran. 2026-05-29.
    Source document
    Analysis of March published in May. Attributed to its authors, not a Governing Council decision. Does not identify the cause of September’s movement.
11. [S11] Baumann, Caprari, Kocharkov, Kouvavas / BCE. Are real incomes increasing or not? Household perceptions and their role for consumption. 2025, Economic Bulletin 1.
    Source document
    Self-reported labour income deflated by national HICP, compared with perceptions. This benchmark does not reconstruct each household’s basket or audit bank accounts.
12. [S12] BCE / ECB. Long-term inflation expectations of consumers: an overview. 2025, Economic Bulletin 3.
    Source document
    Definition and distribution of five-year expectations. Used to interpret anchoring; historical values do not replace September 2026.
13. [S13] BCE / ECB. Our price stability objective. Undated page.
    Source document
    2% medium-term HICP inflation target. The target is not defined over the CES response median.
14. [S14] BCE / ECB. How have households adjusted their spending and saving behaviour to cope with high inflation?. 2024, Economic Bulletin 2.
    Source document
    January 2024 responses about past adjustments. Used qualitatively for consumption, saving and income choices; not a September 2026 measure.
15. [S15] D’Acunto, Charalambakis, Georgarakos, Kenny, Meyer, Weber / BCE. Household inflation expectations: an overview of recent insights for monetary policy. 2024, Discussion Paper 24.
    Source document
    Consumption responses are heterogeneous and context-dependent. Research review, not an estimate of the September-wave effect.
16. [S16] Michael Weber et al., Econometrica. Tell Me Something I Don’t Already Know: Learning in Low- and High-Inflation Settings. 2025-01, 93(1), 229–264.
    Source document
    Original-paper abstract and metadata reviewed. State-dependent attention to inflation; no replication or full reading of paywalled text.
17. [S17] BCE / ECB. CES: Data and methodological information. Accessed 26 October 2026.
    Source document
    Access points for aggregates and microdata. Values used here come from the release and public tables; the guides explain how the statistics are calculated. Microdata not downloaded or re-estimated.
18. [S18] European Central Bank. ECB Consumer Expectations Survey results – August 2026. 18 September 2026.
    Source document
    Three- and five-year inflation medians rose in August from 2.7% to 2.9% and from 2.4% to 2.5%. Historical context for September’s unchanged readings.
19. [S19] Elena Bobeica, Gerrit Koester, Friderike Kuik and Christiane Nickel / ECB Blog. Has food inflation lost its bite?. 20 October 2026.
    Source document
    Analysis of persistent price levels and energy-cost transmission through food supply chains. The authors’ views are distinct from monetary-policy decisions.