Dissecting Individual Expectations on the Macroeconomy: Household Responses to Unexpected Income Shocks

Randomized hypothetical vignette experiment conducted within a survey. Respondents are randomly assigned to a hypothetical income-shock scenario and asked how they would respond. Randomization determines only the hypothetical scenario presented; no actual income shock occurs, no real payment is made, and no factual information about actual economic conditions or policies is provided.

Eight between-subject arms in a fully crossed 2 × 2 × 2 design varying sign (positive vs negative), size (EUR 1,500 vs EUR 5,000), and source (idiosyncratic household-specific shock vs aggregate fiscal shock affecting most households).

- A: idiosyncratic positive shock of EUR 1,500.
- B: idiosyncratic negative shock of EUR 1,500.
- C: aggregate/fiscal positive shock of EUR 1,500.
- D: aggregate/fiscal negative shock of EUR 1,500.
- E: idiosyncratic positive shock of EUR 5,000.
- F: idiosyncratic negative shock of EUR 5,000.
- G: aggregate/fiscal positive shock of EUR 5,000.
- H: aggregate/fiscal negative shock of EUR 5,000.

2026-10-05

2026-10-27

MPC: Marginal propensity to consume: share of the shock allocated to increases in essential and non-essential spending in positive arms, or absorbed through reductions in essential and non-essential spending in negative arms.
MPS: Marginal propensity to save/dissave: share allocated to saving or investment in positive arms, or absorbed through reduced saving, use of savings, or sale of investments in negative arms.
MPD: Marginal propensity to repay/borrow: share allocated to debt repayment in positive arms, or absorbed through reduced debt repayment or increased borrowing in negative arms.
MPT: Marginal propensity to transfer: share allocated to gifts or transfers to others in positive arms, or absorbed through reduced gifts or transfers in negative arms.

Construction: All primary propensities refer to the three months following the shock, the horizon stated in the allocation question. Each primary outcome equals the euro amount assigned to the relevant margin divided by the absolute size of the assigned shock (EUR 1,500 or EUR 5,000). The allocation task presents seven mutually exclusive categories whose amounts are constrained to sum to the absolute shock amount, so every share lies in [0,1]. The four primary propensities do not themselves sum to one: MPC, MPS, MPD and MPT, together with the two residual margins (receiving fewer transfers from others in positive arms or receiving money from others in negative arms, and other uses or adjustments), exhaust the shock by construction.

Sign convention: In positive arms each propensity is the share of the windfall allocated to an increase in the corresponding margin. In negative arms each propensity is the share of the loss absorbed by the corresponding margin, that is, reduced spending, reduced saving, drawing down savings or selling investments, reduced debt repayment or additional borrowing, and reduced transfers. All propensities are therefore reported as non-negative shares. Comparisons by shock sign therefore measure asymmetry in the magnitude of the response to a loss relative to a gain of the same size and source.
Multiple hypothesis testing: MPC, MPS, MPD and MPT form a single family of four primary outcomes. The primary objects of interest are point estimates and confidence intervals. For each pre-specified treatment contrast, we report unadjusted p-values and Romano-Wolf stepdown adjusted p-values across the four primary propensities, using a resampling procedure consistent with household-level clustering.

Planned durable purchases. Shock-induced change in planned durable purchases over the next 12 months (elicited outside the adding-up allocation task, and therefore not additive with the primary propensities). The question is asked of all respondents in positive arms; in negative arms it is asked only of respondents who planned a major purchase before the vignette, and the reduction may not exceed the planned amount. Respondents in negative arms who planned no major purchase are coded zero.
Essential vs non-essential consumption. Shares of the shock allocated to essential and to non-essential spending, which sum to the MPC.
Open-ended reasons. Open-ended explanations for respondents’ allocation choices.
Work effort. Expected change in household members’ work effort following the assigned financial shock.
Fiscal-financing beliefs. Beliefs in aggregate/fiscal arms about how the fiscal measure would be financed or how additional public revenues would be used.
Price-level expectations. Expectations in aggregate/fiscal arms about the general price level over the next 12 months relative to a situation without the fiscal measure.
Medium-horizon allocation intentions. Intentions for the period between 3 and 12 months concerning money initially saved out of a positive shock, or the part of a negative shock initially absorbed through reduced debt repayment, additional borrowing, reduced saving, use of savings, or sale of investments.
Cumulative 12-month propensities. Measures combining the three-month allocation with the medium-horizon allocation intentions described above. For example, the 12-month MPC in positive arms equals the amount allocated to spending plus the share of the saved amount that respondents intend to spend between 3 and 12 months, divided by the absolute size of the shock. Other cumulative allocation measures are constructed from the corresponding follow-up responses. Where debt and saving adjustments cannot be separately identified, they are reported jointly.

Design: Individual-level, between-subject 2 × 2 × 2 randomized design embedded in a single LISS survey wave. Each respondent sees exactly one hypothetical vignette. No real payment is made, and no factual information about actual economic conditions or policies is provided.

Not available

Randomization is performed by LISS/Centerdata when a respondent opens the questionnaire. Assignment is evenly randomized in rounds of eight respondents: within each round, each of the eight treatment arms is assigned exactly once, in random order. For each new round, the eight treatment arms are randomized again.

The randomization unit is the individual respondent. The treatment is not clustered. The research team does not implement or influence the randomization procedure; treatment assignment is recorded in the survey data for analysis.

No

Clusters are at the household level, since multiple members of the same household may respond. We expect approximately 3,070 households for 4,300 individual respondents. The number of households is not known in advance, as it depends on which panel members respond; the figure is an estimate based on the average number of respondents per household in the LISS panel (approximately 1.3–1.5). The actual number of household clusters will be reported after data collection.

LISS expects a net response of approximately 4,300–4,500 respondents. Assignment is balanced within each complete round; the number of usable responses may differ across arms because of survey breakoff or non-response. The LISS sample is designed to be representative of the Dutch population. Where possible, and while maintaining representativeness and the targeted net sample, LISS oversamples panel members who have not objected to linkage with CBS microdata. Because more than one member of the same household may take part, standard errors are clustered at the household level.

Total N = 4,300 individual respondents, randomly assigned with equal probability to 8 arms (approximately 538 respondents per arm)

Sample size is determined by the LISS fieldwork target rather than by ex ante power targeting. LISS expects a net response of approximately 4,300–4,500 respondents. We compute minimum detectable effects using the lower end of this range, N=4,300, corresponding to approximately 538 respondents per arm under balanced assignment. With 4,300 respondents, we can detect differences of approximately 0.085 standard deviations when comparing positive versus negative shocks, large versus small shocks, or fiscal versus idiosyncratic shocks, pooling across the other two randomized vignette dimensions. For a share outcome with a standard deviation of 0.30, this corresponds to approximately 2.6 percentage points. For a comparison between two individual arms, the minimum detectable effect is approximately 0.171 standard deviations, or approximately 5.1 percentage points; the same benchmark applies to a two-way interaction between randomized vignette dimensions, averaged over the third dimension. Heterogeneity analyses on the CBS-linked subsample will have larger minimum detectable effects because of the smaller available sample. All calculations assume two-sided tests, a 5 per cent significance level and 80 per cent power, and do not account for household-level clustering, precision gains from covariates, or the multiple-testing adjustment described in Section 4 of the additional document.