📈 economics

China's Consumption Geography: A County-level Estimation of Daily Consumption Using Big Data

Using Baidu big data to overcome previous spatial limitations, this study reveals that China's county-level daily consumption exhibits significant east–west disparities driven by administrative levels and urbanization, where consumption markets reduce low-level consumption while amenities expand high-level consumption.

Original authors: Yuqingyang Wang, Tao Liu

Published 2026-07-14

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Yuqingyang Wang, Tao Liu

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine China as a giant, bustling video game map. For a long time, players trying to understand how much money people spend on their daily lives (like snacks, clothes, and bus tickets) had to guess based on blurry, low-resolution screenshots. They either looked at huge, fuzzy regions or zoomed in on just one tiny village. They couldn't see the whole picture clearly.

But now, researchers Yuqingyang Wang and Tao Liu from Peking University have unlocked a "God Mode" view using a massive digital lens called Baidu Population Profiling. This isn't a survey where people fill out forms; it's like watching the digital footprints of over one billion people as they scroll, shop, and move around on their phones. Using this data, they mapped out the "Daily Consumption" (RDC) of almost every single county in China in 2021.

Here is what their map reveals, told through the lens of a digital explorer.

The Great East-West Divide

If you look at the map, the difference between the East and the West is as stark as night and day. The East (the coastal side) is glowing with high-energy "spending zones," especially around the big coastal city clusters. These areas are like VIP lounges where people have more money to burn on daily treats.

In contrast, the West is more like a quiet, rugged wilderness. Here, the "low-consumption" population is concentrated, meaning people are spending less on their daily routines. The middle ground, the Central region, is mostly filled with the "medium-consumption" crowd. It's a clear step-by-step gradient: the further you go west, the more the spending power drops.

The "City Level" Ladder

The researchers found that where you live on the administrative ladder matters a lot. Think of it like climbing a ladder:

Rural Counties (The Bottom Rung): These areas have the highest share of people in the "low-consumption" group.
County-Level Cities (The Middle Rung): As you move up to these slightly bigger towns, the number of low-spenders drops by about 5 percentage points.
Municipal Districts (The Top Rung): These are the big, fully urbanized city centers. Here, the "high-consumption" crowd is the most crowded.

The data shows that as you climb from a rural county to a city center, the population shifts from "spending little" to "spending a lot." It's not just that cities have more people; it's that the type of spending changes.

The Urbanization Engine

There is a strong link between how "urban" a place is and how much its residents spend. The authors suggest that as the percentage of urban residents goes up, the spending structure upgrades.

Early Urbanization: When a place starts getting more urban, the "low-spenders" disappear first, replaced by "medium-spenders."
Late Urbanization: As the city gets fully developed, the "medium-spenders" start turning into "high-spenders."

It's like a video game character leveling up: first, you stop buying the basic starter gear (low consumption), then you buy standard gear (medium), and finally, you unlock the legendary, high-end items (high consumption). The data shows a correlation coefficient of 0.43 between urban population share and average weekly spending, meaning they move together pretty closely.

Two Different Engines: The Market vs. The Amenities

The paper argues that two different things drive this spending, and they work in different ways. Imagine a car with two engines:

The Consumption Market (The Engine of Income & Jobs):
This is driven by how rich the area is (Per Capita GDP) and how many people live there. The data suggests this engine is great at kicking out the low-spenders. When the economy is strong, people stop struggling to buy basics and move up to the middle class. However, the researchers found a weird twist: having a huge number of employees in a region actually correlates with lower local spending. They suspect this is because many of those workers are migrants who earn money there but send it home to their families in other towns to spend. So, a big workforce doesn't always mean big local spending.
The Consumption Amenity (The Engine of Stuff & Roads):
This is about how easy it is to buy things and get around. It includes the number of shopping facilities (like malls and stores) and the density of the road network.
- Shopping Facilities: Having more stores doesn't just help everyone spend a little more; it specifically boosts the high-spenders. It's like having a luxury department store: it attracts the people who are ready to buy the expensive stuff.
- Roads: Better roads seem to create a "polarization." They help both the low-spenders (who can now get to the market) and the high-spenders (who can travel easily), but they sometimes squeeze out the middle group.

What the Paper Says It Doesn't Know

The authors are careful to tell us what their map doesn't show.

No Big Purchases: Their data tracks "routine" weekly spending (like food and transport) but excludes big-ticket items like buying a house or a car. They can't tell you how much people spend on a new apartment.
No "Why" for Specific Categories: They can't separate "survival" spending from "fun" spending.
The Smartphone Bias: Since the data comes from phone users, it might miss people who don't have smartphones or don't use them much. The authors suggest this could skew the results, especially in poorer areas where phone ownership might be lower.
The Migrant Mystery: Because the data tracks individuals, not families, it can't perfectly capture the "work here, spend there" situation of migrant workers.

The Bottom Line

The study suggests that China's daily spending is a story of geography and urbanization. The East and big cities are the spending powerhouses, driven by a mix of high incomes and great access to shops and roads. While a booming economy helps people move out of the "low-spend" category, having a dense network of shops and roads is what really pushes people into the "high-spend" category.

It's a snapshot of 2021 using 2,827 county-level units (merged into 2,125 for the final math), showing us that in the game of daily life, where you live and what's available nearby are the biggest cheat codes for how much you spend.

Technical Summary: China's Consumption Geography: A County-level Estimation of Daily Consumption Using Big Data

Problem Statement
Residents' daily consumption (RDC) is a critical indicator of regional development, yet existing research has been constrained by data availability. Previous studies have largely relied on coarse spatial scales or localized case studies, failing to capture fine-grained spatial disparities in consumption. Furthermore, most literature focuses on specific consumption types (e.g., medical or tourism) or individual behaviors, leaving a gap in understanding the comprehensive, spatial distribution of daily consumption levels across China's vast county-level units. The rapid urbanization of China, characterized by production-led growth and significant urban-rural gaps, necessitates a detailed examination of consumption geography to inform policies aimed at "expanding domestic demand."

Methodology
This study utilizes a novel big data approach to estimate RDC at the county level across 2,827 county-level units in China (aggregated to 2,125 units for regression analysis).

Data Source: The primary data is the 2021 Baidu Population Profiling dataset, derived from machine learning models integrating user behavioral records, device information, and location data from over one billion individuals. This dataset classifies individuals into three consumption tiers based on average weekly expenditures:
- Low: < CNY 110
- Medium: CNY 110–506
- High: > CNY 506
- Note: The data covers routine expenditures (clothing, food, transport) but excludes major purchases and housing rent.
Indicator Construction: A per capita weekly RDC indicator ( $WRDC_i$ ) is constructed for each county $i$ using the population sizes of the three consumption groups as weights. To handle the open-ended high-consumption interval, the study assumes geometric progression for midpoints, estimating representative weekly expenditures as CNY 55 (Low), CNY 308 (Medium), and CNY 1,724.8 (High).
Analytical Framework: The study posits that RDC is driven by two main factors:
1. Consumption Market: Administrative level, urbanization rate, per capita GDP, and employment size.
2. Consumption Amenity: Density of retail facilities (shopping POIs) and road network density (accessibility).
Statistical Approach: Multiple linear regression models (OLS) are employed.
- Model 1: Dependent variable is the per capita weekly RDC level.
- Models 2–4: Dependent variables are the population shares of low-, medium-, and high-level consumption groups, respectively.
- Controls: Age composition, educational attainment, and spatial indicators (region, city size, urban agglomeration status).

Key Results

Spatial Patterns: RDC levels exhibit a pronounced east–west gradient. High-consumption populations are heavily concentrated in coastal urban agglomerations and provincial capitals, while low-consumption populations are clustered in western China, particularly the southwest. Medium-consumption populations are dominant in central China.
Administrative and Urbanization Effects: There is a clear hierarchy in consumption levels corresponding to administrative status: Municipal districts > County-level cities > Counties. Urbanization is positively correlated with RDC levels (correlation coefficient of 0.43). As urbanization increases, the share of low-consumption populations declines, while the high-consumption share rises; the medium-consumption share follows an inverted U-shaped trajectory.
Determinants of Consumption Structure:
- Per Capita GDP: Strongly positively associated with overall consumption levels and the upgrading of consumption structures.
- Employment Size: Surprisingly, shows a significant negative relationship with local consumption levels. The authors attribute this to migrant workers who earn income in destination regions but consume in their places of origin (remittances).
- Consumption Amenities: Both road network density and the number of shopping facilities are positively correlated with consumption levels.
- Mechanism Differentiation: The study distinguishes the roles of market factors versus amenity factors. Improvements in the consumption market (e.g., urbanization, GDP) primarily drive consumption upgrading by reducing the proportion of the low-level consumption population. In contrast, better consumption amenities (e.g., retail density, road access) are more strongly associated with expanding the high-level consumption population.

Significance and Claims
The paper claims to provide a more comprehensive and fine-grained understanding of the spatial distribution of RDC in China, overcoming the limitations of traditional statistical surveys. By leveraging big data, the study reveals the close association between consumption levels and the urbanization process, highlighting that highly urbanized areas benefit from both larger consumer markets and superior consumption amenities.

The authors modestly frame this work as an "initial exploration" of regional consumption levels. They acknowledge limitations, including:

The metric captures routine daily spending rather than major purchases or distinct consumption categories (survival vs. enjoyment).
Potential selection biases inherent in big data, such as reliance on smartphone penetration and multi-device usage.
The complexity of migrant consumption patterns, suggesting that future research should adopt the household as the primary unit of analysis to better address the spatial separation of work and consumption.

Ultimately, the study contributes to the literature by empirically validating the roles of "consumption markets" and "consumption amenities" in shaping regional disparities, offering a data-driven basis for policies aimed at stimulating domestic demand and improving the quality of urbanization.