Households' economic response to their perceived income uncertainty

The first question starts by collecting respondents' current household income allocation decisions. Afterwards, everyone received a guaranteed income as the control group. The third question randomly split the whole sample into four groups. Two groups received a risk preference question and another two groups received an uncertainty ambiguity preference question. From the fourth question to the fifth question, whole sample will be randomized into two treatment groups (4 sub-groups).

Treatment group 1: Perceived income risk treatment
- Respondents are shown hypothetical income distributions with identical expected values but different variances (low vs. high).
- Treatment assignment randomizes the order and presentation of these distributions within subjects.
Treamnet group 2: Perceived uncertainty ambiguity treatment
- Respondents face scenarios without explicit probabilities for income outcomes, requiring them to rely on their own subjective beliefs.
- This condition identifies behavioral responses to uncertainty under ambiguity.

The intervention is implemented as a series of structured survey modules embedded in the Bundesbank Online Household Panel (BOP-HH). The design employs a within-subject control combined with between-subject treatment assignments to identify behavioral responses to different forms of income uncertainty. The intervention consists of five sequential questions as described below.

Question 1 – Baseline Allocation Question
Respondents begin by reporting their current household income allocation across three categories:
1. Consumption,
2. Saving, and
3. Debt repayment.

Question 2 – Guaranteed Income Scenario (Control Condition)
All respondents are then presented with a guaranteed income scenario over the next 12 months.
• The scenario specifies a fixed, certain level of household income (i.e., no risk or uncertainty).
• Respondents are asked to indicate how they would allocate that income among consumption, saving, and debt repayment.
This question functions as the within-individual control condition, enabling comparison of each respondent’s decision under certainty versus their decisions under subsequent risk or uncertainty treatments.

Question 3 – Preference Elicitation (Risk vs. Ambiguity)
The full sample is randomly divided into four groups to elicit baseline attitudes toward income risk and ambiguity:
• Two groups receive risk preference questions.
• Two groups receive uncertainty (ambiguity) preference questions.
Each respondent is shown a set of hypothetical future income scenarios that differ either in variance (for risk) or in the degree of ambiguity (for uncertainty). Respondents are asked to rank the scenarios from most to least preferred.
A neutral option (“no preference”) is also included. Respondents selecting this option are not required to rank the remaining scenarios, ensuring that those with indifference or low engagement do not contribute noise to preference ordering data.
This module captures heterogeneity in baseline preference structures prior to the behavioral treatment tasks.

Question 4 – Randomized Behavioral Treatment Assignment
After the preference elicitation, two groups will be risk treament group and another two groups will be risk uncertainty group.
• Treatment Group 1: Perceived Income Risk Treatment
o Respondents are presented with hypothetical future income distributions (vignette question) that all share the same expected income but differ in variance:
 Low variance condition: ±20% perceived income fluctuation.
 High variance condition: ±40% perceived income fluctuation.
o The order and presentation of these distributions are randomized within respondents to mitigate order effects.
o After reviewing each distribution, respondents allocate their hypothetical income among consumption, saving, and debt repayment.
• Treatment Group 2: Perceived Uncertainty (Ambiguity) Treatment
o Respondents are shown income scenarios without explicit probability distributions.
o They are asked to make the same allocation decisions (consumption, saving, debt repayment) based on their own subjective beliefs about how likely each income outcome is.
o This design isolates behavioral responses to uncertainty, holding expected income constant but removing known probabilities.

Question 5 – Within-Individual Comparison and Behavioral Measurement
Each respondent thus provides allocation decisions under three key conditions:
1. Certainty (control)
2. Known risk (variance-based treatment) or
3. Ambiguity (subjective uncertainty treatment)

By comparing allocations across these conditions within the same respondent, the design identifies how individuals adjust spending, saving, and debt repayment in response to different forms of perceived income uncertainty.

The resulting measures quantify both the precautionary saving motive and the marginal propensity adjustments implied by subjective risk and ambiguity perceptions.

2025-11-17

2026-01-29

Each outcome variable is constructed as the self-reported percentage share of total household income (normalized to 100) allocated to consumption, saving, and debt repayment. For each respondent, we compute differences in these allocation shares across the certainty, risk, and ambiguity conditions. These within-individual changes are interpreted as behavioral responses to (i) known income risk (variance-based uncertainty) and (ii) ambiguous income uncertainty (unknown probabilities). The key constructed measures are:
• ΔSaving (risk) = Saving under risk – Saving under certainty
• ΔSaving (ambiguity) = Saving under ambiguity – Saving under certainty
Similar difference measures are constructed for consumption and debt repayment. These indicators serve as behavioral proxies for the precautionary saving motive and ambiguity aversion.

The primary outcomes are theoretically grounded in models of intertemporal consumption under income uncertainty, where individuals allocate income across consumption, saving, and debt repayment to smooth utility over time (Carroll 1997; Gollier 2001). Within this framework, precautionary saving and ambiguity aversion emerge as behavioral responses to income risk and uncertainty, respectively.
Empirically, each respondent reports income allocation shares (summing to 100) under three conditions: (i) certain income, (ii) income with known variance (risk), and (iii) income with unknown probabilities (ambiguity). From these allocations, we construct within-individual differences to capture behavioral responses to uncertainty:
• Precautionary saving response (risk):
 ΔSavingᵣ = Saving under risk – Saving under certainty
• Ambiguity response:
 ΔSavingₐ = Saving under ambiguity – Saving under certainty
Parallel difference measures are derived for consumption and debt repayment.
Theoretically, increases in saving (or reductions in consumption) under risk reflect the strength of the precautionary motive, consistent with higher curvature of the utility function with respect to income. Increases in saving under ambiguity reflect ambiguity aversion—a preference for avoiding ill-defined uncertainty as in multiple-prior or smooth ambiguity models (Ellsberg 1961).
These constructed outcomes thus operationalize key behavioral parameters of interest: (1) the marginal propensity to adjust saving in response to income risk and (2) the marginal propensity to adjust saving in response to income ambiguity. The within-subject design ensures that these outcomes reflect individual behavioral responses rather than between-person heterogeneity.

This study is a randomized survey experiment. The study aims to understand how households’ consumption, saving, debt repayment resposnese under income risk and uncertainty, seperately. The experiment is implemented within a national representative household online panel. Each respondent completes a series of income allocation under three conditions: a guaranteed-income (certainty) scenario, a known-risk scenario with small and big income shock, and an ambiguous-income scenario.

At the start of the survey, respondents are randomly assigned at the individual level to four groups. Two groups are risk groups in which income over the next 12 months follow mean-preserving spread with equal probability. For example, respondents are told that their household income will be x% higher or lower than their current household income, with equal probabilty for both scenorios. Between these two risk groups, one group will received smaller income shock another with will received bigger shock.
Another two groups are ambiguity groups, in which probabilities are not disclosed. Respondents only know the size of their income variace but need to make decision based on the distribution in their mind. The uncertainty questions follow after knowing equal probability for anchoring people firstly under the certainty situation. All respondents first report baseline allocations under prior and certainty. Meanwhile, they can all the time see what they answered under certainty situation. Afterwards, respodents will face the risk and ambiguity preference questions.

The control condition corresponds to the certain-income scenario, providing a within-respondent benchmark. Behavioral responses are captured as the change in the share of income allocated to saving, consumption, and debt repayment between the certain and uncertain conditions. The main hypotheses are: (i) households increase saving (and reduce consumption) under known income risk relative to certainty, consistent with a precautionary saving motive; (ii) the same or stronger response occurs under ambiguity, consistent with ambiguity aversion; and (iii) the adjustment under ambiguity is at least as large as under risk.

The experiment is implemented as a structured survey module within the Bundesbank Online Household Panel (BOP-HH). The design combines a within-subject control and a between-subject randomization to identify behavioral responses to income risk and uncertainty. Each respondent completes a sequence of five tasks that together generate individual-level measures of allocation behavior, risk preference, and ambiguity preference.

Experimental structure and sequence.
The survey begins with a baseline allocation question, where respondents report their current allocation of household income across three categories: consumption, saving, and debt repayment. Next, all respondents complete a certainty (control) scenario, in which they allocate monthly household income across the same three categories under fully certain scenario. This scenario serves as the within-individual benchmark for subsequent comparisons.

After these baseline questions, respondents complete a preference elicitation question, which measures individual differences in attitudes toward risk or ambiguity. At this point, respondents are randomly assigned at the individual level into one of four treatment groups. Two groups belong to the risk condition, in which income over the next twelve months follows a mean-preserving spread around current income with known and equal probabilities. One group faces a low-variance condition (20% income become uncertainty), and the other faces a high-variance condition (40% income become uncertainty). The remaining two groups belong to the ambiguity condition, where the size of potential income variation is disclosed (20% or 40% depends on which group peole belongs to), but probabilities are not specified. Respondents must therefore form subjective beliefs about the likelihood of outcomes. The subjective belief of likelihood is captured from another existed question from BOP-HH (density question of expected household income over the next 12 months).

Within each assigned treatment group, respondents complete behavioral allocation tasks in which they decide now how to distribute hypothetical future income under their assigned uncertainty scenarios. The order of scenario presentation is randomized within individuals to minimize order and anchoring effects. In both the risk and ambiguity groups, respondents can refer back to their certainty allocation throughout the task, ensuring consistent scaling across conditions.

Timing of randomization and treatment exposure.
Randomization occurs automatically by computer at the beginning of the preference elicitation module, after all respondents have completed the certainty baseline. This ensures that treatment assignment does not affect initial responses. The randomization algorithm assigns each respondent with equal probability to one of the four treatment groups, with no stratification or blocking.

Outcome measurement and hypotheses.
The primary behavioral outcomes are the shares of household income allocated to consumption, saving, and debt repayment under each scenario. The control condition corresponds to the guaranteed-income scenario. Behavioral responses are calculated as within-individual changes in allocation shares between the certainty and uncertainty conditions.
The main hypotheses are:
(i) Relative to certainty, households increase saving (and reduce consumption) under known income risk, consistent with a precautionary saving motive. May some respondents show priorities for firstly repay their debt and then consider to save if they already have certain level of debt but don’t have very high level of wealth as buffer.
(ii) A similar or stronger response occurs under ambiguity, consistent with ambiguity aversion.
(iii) The adjustment under ambiguity is based on the subjective distribution within each individual’s mind. The subjective distribution should based on the existed question uncer density question of household income. For households who are ambiguity-aversion, they may response even more to save.

The randomization is performed automatically within the survey platform implemented by the professional survey company Forsa. Respondents are randomly assigned to one of four treatment groups (risk-low variance, risk-high variance, ambiguity-low variance, ambiguity-high variance) at the start of the preference elicitation module. The assignment will be done by Forsa to ensure balanced group sizes across condition.

Individual respondents

No

NA

The Bundesbank wil recurit around 3,000 respondents. However, the respondents can choose to answer or not to answer questions.

Group 1 (750 respondents): small income variance
Group 2 (750 respondents): big income variance
Group 3 (750 respondents): small income variance + uncertainty
Group 4 (750 respondents): large income variance + uncertainty