China material stocks and flows account for 2019–2023

Abstract

China is the world’s largest consumer of bulk materials, and the accumulation of material stocks and flows plays a critical role in shaping its socioeconomic metabolism and environmental sustainability. Detailed and up-to-date provincial-level data on material stocks and flows are therefore essential for understanding resource use and supporting regional sustainability transitions. This study presents the most recent update of the Provincial Material Stocks and Flows Database (PMSFD), extending our previous dataset for 1978–2018 to cover the years 1978–2023. The PMSFD provides consistent time-series data for 31 provinces and 13 bulk material categories across five key end-use sectors, compiled using a bottom-up material flow analysis framework that integrates official statistics and standardized estimation methods. Compared with the earlier version, the updated PMSFD captures the recent slowdown in material accumulation and reveals the regional heterogeneity. The data can be applied to studies on material metabolism, resource efficiency, circular economy strategies, and embodied carbon emissions accounting, as well as to inform regional and national sustainability policymaking in China.

Background & Summary

Systematic accounting of material stocks (e.g., material accumulation) and flows (e.g., material demands and end-of-life waste) is essential for quantifying resource extraction, waste generation, and environmental impacts, and for underpinning effective sustainability strategies^1,2. The Global Resources Outlook 2024 published by UNEP warns that without urgent global action, resource extraction could increase by around 60% above 2020 levels by 2060, intensifying the triple planetary crisis of climate change, biodiversity loss, and pollution^3,4. Given China’s position as the world’s largest consumer of raw materials, comprehensive data on material stocks and flows are critical for understanding its socioeconomic metabolism and for assessing environmental consequences at both national and regional scales^5,6.

In advanced economies, material stocks growth has gradually stabilized and in many cases reached saturation, reflecting structural transformation and the declining demand for infrastructure and durable goods^7,8. For instance, the Netherlands, as one of Europe’s most infrastructure-intensive countries, exhibited an S-shaped growth pattern in its concrete stocks between 1950 and 2023, with annual growth decelerating to approximately 1.5% after 1970, coinciding with the country’s infrastructure saturation period⁹. In Japan, building stocks—holding the dominant share of national material stocks—registered accelerating growth during the rapid economic expansion of the 1960s, peaked at 9.5 Gt between 2005 and 2008, and subsequently declined modestly to 9.3 Gt¹⁰. Similarly, in nations with long industrial histories, such as the United States, the United Kingdom, and Germany, stocks have either slowed or plateaued, with per capita in-use steel stocks ranging between 11 and 16 tonnes¹¹. These trends illustrate a broader transition from material-intensive growth to the maintenance and optimization of existing stocks.

By contrast, China has followed a distinct trajectory. Accelerated industrialization, urbanization, and infrastructure expansion have fueled quick growth in material accumulation over the past four decades^5,6,12. Within forty years, China’s material stocks expanded far more rapidly than historical patterns in many other regions⁸. Over the past decades, China’s growing material demand has made the efficiency of resource use and its associated environmental impacts key concerns for national development. This resource-intensive growth has highlighted the importance of understanding material stocks and flows for guiding sustainable development strategies and mitigating environmental pressures^3,12–15.

In recent years, both the magnitude and composition of China’s material stocks have begun to shift. Urbanization is decelerating, infrastructure and construction demand are approaching saturation, and industrial restructuring and the economic impacts of the COVID-19 pandemic have reshaped material demand^16–19. At the same time, China’s economic development has entered the “New Normal”, marked by slower but more sustainable growth, with policies focusing on efficiency and structural upgrading²⁰. The implementation of China’s dual-carbon strategy (carbon peaking and carbon neutrality) has further accelerated efforts to optimize resource use and reduce environmental impacts²¹. These changes are increasingly evident at the regional level, where material stocks are showing signs of relative decoupling from economic growth²².

To analyze these dynamics, robust and harmonized subnational datasets are indispensable. However, systematic and consistent material stocks and flows accounts are lacking not only at the provincial level but also at the national scale. Existing national and provincial estimations provide only partial coverage and do not constitute a comprehensive material flows database. Most studies focus on single materials, individual sectors, or limited time frames—for example, estimating in-use steel stocks in several provinces from 2000 to 2018, or compiling stocks of twelve bulk materials for all provinces only in 2000 and 2020^23,24. The discrepancies in data sources, estimation methodologies, and scope among these studies further hinder integration and comparability. To date, the only province-level database that covers the whole country, multiple sectors, and multiple materials extends only to 2018²⁵, leaving an urgent need for more up-to-date and harmonized accounts.

This study addresses the gap by presenting an updated Provincial Material Stocks and Flows Database (PMSFD) in China, extending coverage from 1978–2018 to 1978–2023. The dataset includes 31 provinces and 13 bulk materials, with annual series of inflows, outflows and stocks across five key end-use sectors: buildings, infrastructure, transportation equipment, machinery, and domestic appliances. Constructed through a standardized bottom-up material flow analysis (MFA) framework and integrated with official statistical data, the PMSFD provides consistent, transparent, and verifiable accounts. All underlying data and calculation codes are openly available. The dataset enables diverse applications and provides a critical empirical foundation for research on China’s resource metabolism, environmental accounting, and sustainability transition. It facilitates cross-regional comparative analyses, supports modeling efforts, and promotes evidence-based decision-making in fields such as circular economy design, climate change mitigation, and urban planning.

Methods

Data sources and data preprocessing

Considering data accessibility and coverage, this study extends our previous dataset (1978–2018) by adding new data for 2019–2023, thereby providing a continuous 45-year time series from 1978 to 2023. The spatial scale is at the provincial level, covering 31 provinces and autonomous regions in Mainland China. In line with previous research²⁵, the data were classified into five end-use sectors: (i) buildings; (ii) infrastructure; (iii) transportation equipment; (iv) machinery; and (v) domestic appliances. A total of 103 representative products were selected from these sectors.

Data prior to 2018 were sourced from previous research²⁵, with some missing values supplemented. Missing values for the pre-2018 period, along with all data for 2018 and beyond, were collected from official channels, such as socioeconomic databases, statistical reports, and yearbooks for each province. Key sources included websites such as National Data, National Bureau of Statistics (https://data.stats.gov.cn) and China Yearbooks Full-text Database (https://navi.cnki.net/knavi/yearbooks/index). Linear interpolation was utilized to estimate missing data points between continuous records. For Gross Domestic Product (GDP), provincial per capita GDP figures were adjusted to constant prices using 1978 as the base year. More details about the data source can be found in the dataset²⁶.

Modeling methodology

To estimate material stocks and flows, multiple models were integrated within a combined framework. The bottom-up accounting method was utilized to calculate material stocks. Subsequently, the dynamics stock-driven model, including the lifetime model and mass balance approach were applied to determine the material inflows and outflows. The models encompassed 13 commonly used metallic and non-metallic materials, namely steel (Fe), aluminum (Al), copper (Cu), rubber, plastic, glass, lime, asphalt, sand, gravel, brick, cement, and wood.

Material stocks

Bottom-up accounting was used to calculate material stocks. Product data containing the target materials were collected for each end-use sector, and their material intensity was assessed, enabling the calculation of sectoral material stocks $S_{\stackrel{\circledR }{C}}\left(t\right)$ (Eq. 1). The total material stock $S\left(t\right)$ was then obtained by aggregating the sectoral stocks (Eq. 2).

$$S_{\stackrel{\circledR }{C}}\left(t\right)={\sum }_i{N}_i\left(t\right)\times I_i\left(t\right)$$ 1

$$S\left(t\right)={\sum }_C{S}_{\stackrel{\circledR }{C}}\left(t\right)$$ 2

where $S_{\stackrel{\circledR }{C}}\left(t\right)$ represents the material stocks for end-use sector at time $t$, $N_i\left(t\right)$ is the quantity of product $i$ at time $t$, and $I_i\left(t\right)$ is the material intensity of product $i$ per unit at time $t$. $S\left(t\right)$ is the total material stocks at time $t$. The material intensity values were sourced from our previous research²⁵.

In addition, specific assumptions were formulated to address the absence of the official data. Given the lack of official statistics on floor area for non-residential buildings (including public and industrial buildings), the ratio of non-residential to residential building observed at the town or township level was assumed to be applicable for urban or rural areas separately at each province. The estimation formulas are as follows (Eqs. 3–4):

$$R\left(t\right)=\left(f_p\right(t)+f_i(t\left)\right)/f_r\left(t\right)$$ 3

$$F_{\mathit{UR}}\left(t\right)={(A}_{\mathit{CR}}\left(t\right)\times P\left(t\right))\times R(t)$$ 4

where $R\left(t\right)$ is the ratio of non-residential to residential buildings in urban (or rural) areas at time $t$, $f_p\left(t\right)$ and $f_i\left(t\right)$ are the floor areas of public and industrial buildings in towns (or townships) of each province at time $t$, and $f_r\left(t\right)$ is the floor area of residential buildings in towns (or towships) of each province at time $t$. F_UR(t) represents the material stocks of non-residential buildings, $A_{\mathit{CR}}\left(t\right)$ is the per capita floor area of residential buildings in urban (or rural) areas, and $P\left(t\right)$ is the urban (or rural) population at time $t$.

Material flows

Following the acquisition of material stocks data, the material flows for the period 1978–2023 would be estimated using the dynamics stock-driven model. In this model, the annual material outflow was determined from the stock based on the life cycle model, and the annual material inflow was determined from the mass balance of the outflow and the stock change, which is (Eq. 5).

$$\mathit{inflow}=\mathit{outflow}+\mathit{stock}\mathit{change}$$ 5

We assume that the product life cycle in each end-use sector is constant, as shown in Tables 1, 2, due to comprehensive data on product lifetime variability and regional heterogeneity remain unavailable^11,25. In addition, to enhance estimation accuracy, we assume that the time when the product was first put on the market was in 1949, starting from the founding of the People’s Republic of China. The dynamics stock-driven model is as follows (Eqs. 6–9).

$$S_p\left(t_0\right)=K_{p1}\times e^{\left({\mathit{Kp}}_2\times t_0\right)}$$ 6

$$\lambda \left(t,t\prime ,L,\sigma \right)=\frac{1}{\sigma \sqrt{2\pi }}\times \mathit{exp}\left(\frac{-(t-t\prime -L)}{2{\sigma }^2}\right)$$ 7

$$F_{\mathit{out}}\left(\mathrm{t}\right)=\sum _{t^{\prime }\le t}{F}_{\mathit{in}}\left(t\prime \right)\times \lambda (t,t\prime ,\tau ,\sigma )$$ 8

$$F_{\mathit{in}}\left(t\right)=S_{\left(t\right)}-S_{\left(t-1\right)}+F_{\mathit{out}}\left(t\right)$$ 9

Table 1.

Mean lifetime and standard deviation for five end-use sectors.

End-use sector	Mean lifetime	Standard deviation
Buildings	50	15.0
Infrastructure	50	15.0
Transportation Equipment	25	7.5
Machinery	30	9.0
Domestic Appliances	15	4.5

Table 2.

Comparisons of stocks estimation/prediction between our results and previous studies.

Region	Sector	Method	Year	Material(s)	Other Study	This study	Difference
Beijing	Buildings, Infrastructure	Bottom-up	2018	Material stocksa	3621 Mt29	3020 Mt	−16.6%
Shanghai	Buildings	Bottom-up	2018	Material stocksb	4744 Mt28	3698 Mt	−22.1%
China	All	Top-down	2018	Aluminum stocks	255 Mt31	221 Mt	−13.3%
China	Machinery	Bottom-up	2019	Steel stocks	377 Mt30	431 Mt	14.3%
Beijing	Buildings, Infrastructure	Top-down	2020	Material stocksa	174 t/cap32	135 t/cap	−22.4%
Shanghai	Buildings	Top-down	2020	Material stocksc	4147 Mt33	4160 Mt	0.3%
China	All	Top-down	2020	Material stocksd	219 Gt23	191.3 Gt	−12.6%

^aMaterials stocks without considering rubber and plastic.

^bMaterials stocks without considering aluminum, copper, rubber, and plastic.

^cMaterials stocks without considering lime, copper, rubber, and plastic.

^dMaterials stocks without considering lime.

The material stocks prior to 1978 were inferred using a power function (Eq. 6)²⁷, where$S_p\left(t_0\right)$ represents the material stocks in each province at time $t_0$ (1949 ≤ $t_0$ ≤ 1977), and $K_{p1}$ and $K_{p2}$ are the coefficients for each province in the fitted model. The lifetime distribution (Eq. 7) indicates the probability of each end-use sector reaching the end of its life cycle at time $t$, where $\lambda \left(t,t\prime ,L,\sigma \right)$ denotes the normally distributed lifetime with $t\prime$ = 1949, $L$ is the dependent mean of the end-use sector, and $\sigma$ is the standard deviation. When $S_{\left(t\right)}$< $S_{\left(t-1\right)}$, we set the inflow $F_{\mathit{in}}\left(t\right)$ to zero to prevent negative values. Due to the dispersed nature of the data, when analyzing the dynamic patterns of material inflows, we employed the Autoregressive Integrated Moving Average (ARIMA) methodology with a 5-year moving average window to analyze the dynamic patterns of material inflows.

Data Records

The dataset titled “Provincial Material Stocks and Flows Database (PMSFD) 2019–2023” is accessible on Figshare²⁶. The dataset comprises a total of 534,040 data records, structured within core tables saved as xlsx files. The file description as follows:

1. 108,893 are the account of products in 31 provinces from 1978 to 2023 [File ‘Input data];

2. 331 are material intensity data for 13 materials from 1978 to 2023 [File ‘Material intensity];

3. 43,200 are national and provincial material stocks inventories (13 materials, from 1978 to 2023) [File ‘Provincial_material_stocks_1978–2023];

4. 43,200 are national and provincial material inflows inventories (13 materials, from 1978 to 2023) [File ‘Provincial_material_inflows_1978–2023];

5. 43,200 are national and provincial material outflows inventories (13 materials, from 1978 to 2023) [File ‘Provincial_material_outflows_1978–2023];

6. 295,216 are supporting data for the Data Overview and Technical Validation section [File ‘Supplementary Information]

Both material stocks and flows inventories are held in matrices (29 columns x 46 rows per spreadsheet) in a consistent format. The 29 columns detail 13 material stocks or flows within 5 end-use sectors, while the 46 rows cover the 46-year timeframe.

Data Overview

In 2023, China’s total material stocks were estimated at 208.70 Gt, slightly below the 209.78 Gt in 2022 (Fig. 1a). Per capita stocks reached 148.05 t/cap, compared with 148.60 t/cap in 2022. From 1978 to 2023, material stocks grew nearly tenfold, from 21.14 Gt to over 200 Gt, with per capita stocks rising from 21.96 t/cap to over 140 t/cap. While the general trend remains upward, the pace of growth has slowed since 2019. Non-metallic materials consistently made up most of the stocks, with a slight decrease from 99% in 1978 to 97% in 2023, while metallic materials increased from 1% to 3%. In 2023, the largest material shares were gravel (37%), sand (33%), cement (14%), and brick (10%).

Fig. 1.

Spatiotemporal Patterns and Socioeconomic Drivers of Material Stocks in China. (a) The temporal patterns of national material stocks during 1978–2023 by 13 material groups at aggregate levels and per-capita levels. The stacked area chart (left axis) depicts the aggregate material stocks of 13 material groups, while the line chart (right axis) shows the per-capita total material stocks. The inset pie chart illustrates the proportion of material stocks for each category in 2023. (b) Per capita material stocks versus per capita GDP at province level. The black dashed line is the fitted curve to estimate trend of material stocks. Data are smoothed using a 5-year moving average. (c) Distribution of total material stocks across provinces by end-use sector in 2023. The inset bar–pie composite chart illustrates the proportional contributions of each end-use sector to China’s total material stocks in 2023. (d) Rank–size distributions of provincial per-capita material stocks at five-year intervals from 1978 to 2023. The inset line plot depicts the temporal evolution of the Gini coefficient. The gray dashed line serves as a background reference line.

At the provincial level, the growth of per capita material stocks follows an S-shaped curve, slowing as per capita GDP exceeds 20,000 yuan (Fig. 1b). Saturation occurs around 40,000 yuan, leading to a partial decoupling of GDP from material stocks. Provinces such as Zhejiang, Jiangsu, and Guangdong have passed their growth peaks, while Shanghai, Tianjin, and Beijing have more room for accumulation.

Spatially, China’s material stocks exhibit clear heterogeneity, with the highest concentrations in the southeastern provinces (Fig. 1c,d). Guangdong, Shandong, and Jiangsu are the top-ranking provinces in total stocks, while Zhejiang leads in per capita stocks. The building sector remains the primary contributor to material stocks, accounting for about 90% at the provincial level. Within the overall decline in provincial demand, changes across different end-use sectors and material categories display broadly consistent patterns among provinces.

Material demand (inflow) reached 9.18 Gt/year in 2019, and scrap (outflow) generation increased to 0.91 Gt/year by 2023 (Fig. 2). Densely populated and economically developed southeastern provinces exhibited higher material flows than the northwest, with Zhejiang, Jiangsu, and Guangdong surpassing 1 Gt/year in scrap generation. Importantly, a marked decline in material demand after 2018 indicates a structural shift in China’s material consumption patterns. Further details are available in the Tables S1-S5 of the file “Supplementary Information.xlsx.” on Figshare²⁶, which document national and provincial material stocks (Table S1), per-capita material stocks (Table S2), per-capita GDP (Table S3), and material inflow (Table S4) and outflow (Table S5) dynamics for China and its 31 provinces from 1978 to 2023.

Fig. 2.

National- and provincial-level analysis of material flows dynamics during 1978–2023. Dashed blue curves represent material inflows, while solid orange curves denote material outflows. Data are smoothed using a 5-year moving average.

Technical Validation

Comparison with existing estimates or predictions

Due to the lack of comprehensive material stocks data for recent years (post-2018), direct comparisons at the provincial or national level remain unfeasible, which means our estimates provide the most up-to-date and extensive dataset on material stocks in China and its provinces, serving as a critical supplement to existing inventories. However, we attempted to conduct comparisons at finer scales. In other words, we compared our results with previous case studies conducted at urban, material, or sectoral scales (Table 2). Overall, our findings align closely with those of other studies, with an average discrepancy within 20%^23,28–33. This study also corroborates previous predictions regarding the saturation of material demand³⁴. Moreover, the observed decoupling between material stocks and GDP in economically developed regions is consistent with earlier research. For instance, previous study reported similar decoupling trends in major cities such as Chongqing and Beijing²².

Previous studies on developed economies, such as the Netherlands, the United States, the United Kingdom, and Germany, have demonstrated that once economic development reaches a certain stage, the accumulation of material stocks tends to slow down or even stagnate^7–11. Studies on China have already identified a declining trend in material demand within the construction sector¹⁷. Projections regarding China’s material stocks trend also estimate a slowdown around 2030^35,36. Overall, our findings align with expectations. In recent years, factors such as China’s dual-carbon strategy (carbon peaking and carbon neutrality), industrial restructuring, and the COVID-19 pandemic may have contributed to an accelerated deceleration in material accumulation, generating impacts across various provinces and industries nationwide^16–19,37. Due to the data limitations, the present analysis could not provide direct causal evidence for these influences. Future research could explicitly examine the mechanisms through which such factors affect material stock trajectories.

Uncertainty

To quantify the uncertainty in stocks estimates arising from material intensity, we performed the Monte Carlo method³⁸. The coefficient of variation (CV) of material intensity was assumed as 10%, consistent with values commonly adopted in previous material stock and flow studies³⁵. Product data obtained from official sources and therefore considered highly reliable were treated as deterministic (CV = 0). The 97.5% uncertainty range is calculated as the 97.5% confidence intervals of the 100,000 estimations. The results indicate that the uncertainty associated with material intensity exhibits a slight increasing trend over time but remains stable within the range of 3.6%–4.3%, exerting a negligible influence on the total material stock (Fig. 3a). Among all materials, steel contributes the largest share of uncertainty, with an average effect of approximately 1.2%, primarily due to its broad coverage across the products (Fig. 3b). In addition, the distributions of material stock uncertainty across provinces exhibit a high degree of overlap, and the overall uncertainty levels are low (<6%), suggesting that, under the current data availability and perturbation settings, the use of non–province-specific material intensity assumptions do not introduce significant systematic bias (Fig. 3c). Further details regarding provincial-level material intensity impacts (Table S6) and uncertainties (Table S7) over the period 1978–2023 are available in the file “Supplementary Information.xlsx” on Figshare²⁶.

Fig. 3.

Uncertainty of Material Intensity derived from Monte Carlo simulation. (a) Annual average uncertainty of material intensity over the study period. (b) Average contribution of uncertainty from 13 materials to total material stock. (c) Interannual distribution of material intensity uncertainty across provinces.

Limitations and future work

Our dataset has three main limitations. (1) Average material intensities are presently used due to data accessibility constraints, which do not fully capture spatial and temporal heterogeneity arising from differences in product size, function, and production technology^39–41. Future efforts are recommended to refine material intensities by province and shorter time intervals. (2) Lifetimes are assumed to be uniform across all provinces. Future studies should consider lifetimes differences influenced by market conditions, policies, and production technologies^42,43. (3) The stocks calculation for non-residential buildings based on simplifying assumptions. Additional data sources are needed to improve accuracy. We aim to address these limitations in future work by refining model parameters and expanding the scope of the dataset to include more products and indicators, thereby enhancing the accuracy of provincial-level material stocks accounting in China.

Usage Notes

The dataset comprises continuous records of material stocks and flows over a span of 45 years, from 1978 to 2023, suitable for tracking and analyzing material metabolism at the provincial level and above. The data are recorded in xlsx files to facilitate data analysis and modeling in environments such as MATLAB, R, and Python. The dataset comprehensively covers five end-use sectors—buildings, infrastructure, transportation equipment, machinery, and domestic appliances—and 13 material types, including steel (Fe), aluminum (Al), copper (Cu), rubber, plastic, glass, lime, asphalt, sand, gravel, brick, cement, and wood.

Given significant temporal fluctuations, the Autoregressive Integrated Moving Average (ARIMA) methodology is recommended for analyzing dynamic inflow patterns. Furthermore, the computational models utilized in this study can be adapted to estimate the inventory of other materials or the stocks of the same material across different end-use sector classifications.

Author contributions

J.H. collected and assembled the data, run the model, and prepared the manuscript. G.H. and L.S. collected and assembled the data. L.S. revised the manuscript.W.-Q.C. and L.S. led the project.

Data availability

The dataset produced in this work, namely product data, material intensity data, material stocks and flows inventories, has been deposited in the Figshare repository and is publicly available at 10.6084/m9.figshare.30334780.

Code availability

The code used for data processing and calculation of material stocks and flows (code_calculation.m), are also stored on Figshare²⁶ for transparency and verifiability. All data processing and computations were conducted using Matlab R2021a.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41597-026-06903-2.