Consumption Responses to Income Expectations

How to cite

Yin, Xiao; Blundell, Richard. 2025. "Consumption Responses to Income Expectations." AEA RCT Registry. AEARCTR-0016320. https://doi.org/10.1257/rct.16320-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Xiao Yin* , Richard Blundell

* Primary investigator

Registry ID
AEARCTR-0016320
DOI
https://doi.org/10.1257/rct.16320-1.0
First registered
Last updated
Status
In development
Trial dates
2025-07-15 – 2026-07-31
Countries
China ()
Registry record
http://www.socialscienceregistry.org/trials/16320

Abstract

Consumption-saving decisions lie at the heart of household behavior and macroeconomic dynamics. These decisions depend critically on households' expectations about future income, which guide how they allocate resources intertemporally. While a large literature has estimated the marginal propensity to consume (MPC) out of realized income shocks, much less is known about how consumption responds to expected income—particularly across different future horizons. This project fills that gap by studying the MPC to income expectations at multiple horizons, using experimentally induced belief variation. By distinguishing short- from long-run expectation shifts, it aims to clarify how forward-looking beliefs shape consumption dynamics.

Intervention

We have four treatment groups and one control group. For each individual in the treatment groups, we show them the following information

Based on an AI model trained on the financial and employment data of over 10 million urban Chinese households, we have generated income growth forecasts for households like yours.

According to the model, households with similar characteristics to yours are expected to experience an average annual income growth rate of approximately X% over the next k years. This growth reflects general life-cycle income trend, local labor markets, and the macroeconomic situation.

where k∈{0.25,1,3,5} respectively for the four treatment groups. For T1 where k=0.25, we provide information about quarterly earnings. So the bolded part of the information treatment is

Experimental design

We expect to select around 40,000 consumers from 40 cities in China. Then

1. we send out surveys to elicit basic information and income expectations
2. provide information treatments
3. elicit income expectations again
4. send follow-up surveys in 3 and 6 months.

Design details

Randomization method
a random variable generator using Python
Randomization unit
randomized at individual level
Clusters
we expect to have 25,000 individuals in the end
Observations
we expect to have 25,000 individuals in the end
Arms
we expect to have 5,000 for each group (one control and four treatments)
Primary outcomes
Income expectations and spending

Cited by

No citation data available for this registration yet.

Keywords