How households’ inflation expectations affect their spending is a central question for researchers and policymakers alike. Yet empirical evidence remains mixed. This column shows that the consumption question researchers choose can change the sign and strength of the estimated relationship. Questions about spending attitudes and plans yield negative associations with expected inflation, while questions about past spending yield weakly positive or insignificant associations. Differences in question type matter much more than reference unit or recall horizon. Treating these measures as interchangeable can lead to misleading conclusions about the expectations channel of monetary policy.

Do households spend more when they expect inflation to rise? This question is fundamental to how monetary policy affects the real economy. Standard macroeconomic models typically predict that higher expected inflation should encourage households to bring consumption forward by lowering the real interest rate, thereby stimulating aggregate demand. Yet the empirical evidence remains inconclusive: some studies find that higher expected inflation boosts spending, while others find no effect or even the opposite effect (e.g. Andrade et al. 2023, Bachmann et al. 2015, Dräger and Nghiem 2021, Gorodnichenko et al. 2021, Hoang et al. 2015, Weber et al. 2022). What explains these diverging results?

A closer look at the literature reveals an important source of heterogeneity. Consumption questions vary widely across studies — by time horizon, reference unit, and whether they capture realised spending, intentions, or attitudes, such as whether now is a good time for major purchases. In a recent paper (Assenza et al. 2026), we investigate whether these differences in survey design help explain the divergent empirical results on the relationship between inflation expectations and consumption.

We implement a unified experimental framework in the online survey on consumer expectations of the Deutsche Bundesbank (Bundesbank Online Panel Households – BOP-HH). Focusing on spending on durable goods, where the intertemporal substitution logic of standard models applies most strongly, we vary question framing along three key dimensions: reference unit (individual versus household spending), time horizon (spending over the past one, three, or 12 months), and question type (attitudinal measures, spending plans, and qualitative and quantitative reports of past spending). Comparisons within the same respondents and across randomly assigned survey groups allow us to attribute any differences in the estimated relationship between inflation expectations and spending to question framing.

Reference unit and recall horizon have limited influence

We find no evidence that asking about individual rather than household spending systematically affects the estimated relationship between inflation expectations and reported consumption. For recall horizon, a comparison across respondents suggests a stronger association over shorter recall periods, but this may reflect differences in purchase histories. Comparing different recall periods within the same respondents yields no significant difference.

Question-type framing is key

Prior studies use different types of consumption questions. Some ask whether now is a good time to make major purchases, others ask about planned spending, and others ask respondents to recall actual spending over a previous period. These measures are often treated as interchangeable proxies for consumption, but our results show that they capture distinct aspects of spending.

Standard macroeconomic models predict that, holding nominal interest rates constant, higher expected inflation encourages households to shift consumption towards the present and away from the future. Consistent with this mechanism, we find a strong and statistically significant negative association between inflation expectations and planned future spending.

Strikingly, the attitudinal question — whether now is a good time to make major purchases — also yields a significant negative association. This appears at odds with the theoretical prediction when the question is interpreted as a proxy for current spending. Yet the attitudinal measure closely mirrors future spending plans rather than reports of realised purchases. It therefore appears to capture a forward-looking assessment of households’ willingness or readiness to spend: an indicator of consumer sentiment rather than current expenditure.

Recall-based questions, by contrast, yield weakly positive or statistically insignificant associations with expected inflation. This applies to both qualitative questions, which ask whether spending increased or decreased, and quantitative questions, which ask respondents to report the amount spent over a given period. Realised purchases depend not only on willingness to spend but also on households’ income, liquidity, affordability, and immediate needs. A household may be reluctant to spend but still need to replace a broken refrigerator; another may be willing to buy but unable to afford it. The main divide is therefore between measures of sentiment and intentions and measures of realised spending behaviour.

Three validation exercises support our interpretation. First, the correlation structure supports this distinction. Qualitative and quantitative recall-based measures are more closely correlated with each other than with the attitudinal question, which is more closely aligned with planned spending. Second, our quantitative past-month spending question closely tracks realised aggregate consumption in external data, much more so than the attitudinal measure. Third, applying the specifications of Dräger and Nghiem (2021), Andrade et al. (2023), and Ichiue and Nishiguchi (2015) within our unified survey framework reproduces several divergent findings reported in the literature. Unlike the original studies, our setting holds country, institutional environment, time period, sample, and survey constant. These exercises show how differences in question design help explain why previous studies reach different conclusions. The point is not that every small wording change matters, but that different questions often measure different aspects of spending behaviour.

Realised spending reflects necessity and affordability more than inflation expectations

To understand why realised spending reacts only weakly to expected inflation, we asked respondents directly what motivates their durable purchases (see Figure 1). The most frequently reported reason was the need to replace a broken device (93.1%), followed by a desire for innovative products (71.9%) and attractive offers (51.3%). Expected short-term price increases were cited by 37.5% and anticipated income changes by 24.3%. These responses help explain why recall-based measures of realised spending are only weakly associated with inflation expectations: actual purchases reflect immediate needs and households’ circumstances, including liquidity, income, and budget constraints. Attitudinal and planned-spending measures, by contrast, capture consumer sentiment and willingness to spend and are therefore more responsive to inflation expectations.

Figure 1 When and why do people make major purchases?

Note: The figure shows the share of individuals who at least somewhat agree with various statements about when and why they buy durable goods, including necessity, expected price increases (short- and long-term), special offers, optimism about the economy, access to cheap credit, replacement needs, adherence to a budget plan, and expected pay raises.

Consequences for researchers and policymakers

Our findings show that survey-based evidence on the expectations channel is more measurement-sensitive than often assumed. Planned future spending is negatively associated with inflation expectations, consistent with households shifting consumption towards the present. Attitude shows a similarly negative association. If this question measured current spending, the result would contradict the theoretical prediction that higher expected inflation raises current consumption. The attitudinal measure is, however, more closely aligned with future spending plans, suggesting that it captures forward-looking willingness to spend rather than current expenditure.

The related literature commonly interprets retrospective spending reports as measures of current consumption and uses their relationship with inflation expectations to assess intertemporal substitution. Under this interpretation, the weak positive association we find offers only limited support for the predicted increase in current spending. Retrospective questions nevertheless relate past purchases to expectations elicited today, so they do not directly establish whether expectations held before those purchases caused spending to change. This timing qualification matters for how these results should be interpreted.

At the same time, not every survey-design choice matters equally. We find no systematic difference between asking about individual and household spending, and changing the recall horizon from one to three to 12 months does not produce a robust change in the estimated relationship. The main distinction is therefore between questions that measure sentiment and intentions and those that measure realised spending.

Central banks should use these measures as complementary indicators. Attitudinal questions reveal perceived buying conditions and consumer sentiment, spending plans describe intentions, and recall-based questions approximate realised spending. Treating them as interchangeable risks mistaking changes in sentiment or intentions for changes in expenditure. The choice of consumption question therefore shapes the policy conclusions researchers draw about the expectations channel.

Source : VOXeu