Impact of urbanization on construction material consumption: A global analysis

Urbanization is considered a main driver of building material consumption. Nevertheless, statements on links between urbanization and resource consumption remain qualitative. This study aims to globally quantify the links between urbanization and non‐metallic mineral resource consumption at the level of nations. Based on hypotheses, we have investigated the relationship between construction material consumption and urbanization and further impact variables. Data were examined using descriptive and analytical statistical methods, by developing step‐by‐step regression models and representing them in a path diagram. The results show that urbanization alone does not adequately explain consumption of construction materials. Prosperity has a strong impact, too, but also does not have sufficient explanatory power in itself. Only the combination of both variables reveals their complex interrelationships. In principle, more prosperous societies consume more construction materials than less prosperous ones, regardless of the degree of urbanization. With low prosperity, however, material consumption per capita rises with increasing urbanization; with high prosperity, the effect is reversed. Developed societies are the problem today. However, with increasing urbanization and prosperity, dynamically growing societies will dramatically exacerbate the carrying capacity problem in the future. Then again, this is by no means inevitable. Rather, the key is to move from linear to consistently circular models of urbanization. These models must be comprehensive and spatially specified in order to develop sufficient clout in terms of effective resource protection through “circular urbanization.”

especially in North America and Europe, the current process is particularly taking place in countries of the global South (Swilling et al., 2018). While urbanization usually developed alongside industrialization in the first wave, urbanization is currently evident even without economic growth and industrialization (Gollin et al., 2016). Current urbanization patterns are therefore quite different from previous ones. Urbanization poses immense environmental challenges. It is considered one of the main drivers of natural resource use (Fritsche et al., 2015;Satterthwaite, 2011). With regard to material resources, non-metallic mineral construction materials are particularly relevant. More than 90% of the resources used in construction are of non-metallic mineral origin (Schiller et al., 2017a). Compared to other societal materials, the consumption of this material group is increasing most strongly worldwide (Krausmann et al., 2009).
Environmental and socioeconomic implications of urbanization have been widely discussed in relation to the consumption of non-metallic mineral resources. Local resource scarcity is highlighted (Tuoi Tre News, 2017), as are the negative consequences of resource extraction such as erosion (Sing et al., 2012), degradation of landscapes, ecosystems and natural habitats (Kittipongvises, 2017;Schneider et al., 2018), and land use conflicts (Ortlepp et al., 2015b). At the global scale, the focus is on climate change due to greenhouse gas emissions as well as on non-compliance with ecological limits (Satterthwaite, 2011;Soonsawad et al., 2022;Yeh & Huang, 2012;Zhang et al., 2021;Zhao & Zhang, 2018). However, statements on the link between urbanization and resource consumption remain qualitative. Analyses of material consumption in the context of urbanization usually refer to case studies from regions with high urbanization dynamics. Examples for China can be found in Fernández (2007) and Yeh and Huang (2012), while Choplin (2019) and Byiers et al. (2017) provide insights into Africa's cement demand. Urbanization itself is not usually the object of study but provides background for discussion about the urban metabolism. Schiller et al. (2020), for example, quantify material requirements of the Hanoi region in order to discuss their impact on the supply of raw materials from the hinterland.
This study is the first attempt to explicitly and globally investigate and quantify linkages between urbanization and non-metallic mineral material consumption using statistical methods. We reflect on whether the inclusion of other variables, in particular prosperity, would allow a more robust assessment of the impact on construction material consumption in the context of urbanization. This is done at the level of nations, including all states in the world where corresponding data are available. In Section 2 the research concept and the analytical methods are described. In Section 3, we outline the operationalization of the variables and present the data sources. The focus of Section 4 is on the results of the step-by-step regression models and their representation in a path diagram. In Section 5, we reflect on the selection of indicators and discuss the results in a broader context. Finally, we address the need to understand complex interrelationships between prosperity, operationalized by the human development index (HDI), and urbanization and draw conclusions with regard to the need to shift from linear to spatially differentiated circular urbanization.

Overall concept
Our research concept is applying descriptive and analytical methods. We use secondary data from publicly available sources (Section 3). The aim is to draw a comprehensive picture by including all countries of the world for which data are available. We describe the consumption of raw materials induced by construction materials using the indicator domestic material consumption of non-metallic mineral construction materials per capita (DMCm) (Section 3.1). The starting points are literature-based hypotheses that describe a relationship between selected indicators and DMCm (Section 2.2). Based on this, two analytical steps are carried out: (1) descriptive analyses using graphs and tables to illustrate the distribution of the data, and (2) analytical methods for hypothesis testing using bivariate correlations and regression models. The complete regression model is presented using a path diagram. Some passages refer to tables and figures that can be found in Supporting Information S1. These are marked with an "S," for example, "Table S4; Figure S5."

Hypotheses
Based on indications from the literature on variables influencing DMCm, the following hypotheses were derived and are specified below: -Urbanization is a main driver of construction material consumption.
-Prosperity is a main driver of construction material consumption.
-Geographical contextual factors influence the consumption of construction materials in a country.

2.2.1
Urbanization is a main driver of construction material consumption H1.The level of urbanization correlates positively with the DMCm Specific references to the relationship between urbanization and construction material consumption are provided by Swilling et al. (2018), who show a positive relationship between the level of urbanization and construction material consumption. Fernández (2007, p. 107) discusses the high consumption of raw materials in the first phase of advancing urbanization, described as "rapid urbanization." In Miatto et al. (2016, p. 924 Transport infrastructure is a mass-intensive part of built environment. Schiller et al. (2017b) show that, for Germany, almost half of the anthropogenic non-metallic mineral material stock is embodied in roads. Miatto et al. (2017) estimate that road infrastructure accounts for about 15% to 20% of total US domestic consumption of non-metallic construction materials. Accordingly, a positive correlation between road networks and DMCm is expected.

2.2.2
Prosperity is a main driver of construction material consumption H2.A country's prosperity correlates positively with its DMCm Choplin (2019) takes Africa's cement industry as an example to show the link between rising wealth, the building boom, and construction materials. According to this, a person's social and economic position can be assessed from the cubic meters of cement used in their house. Schandl and Eisenmenger (2006, p. 134) confirm the close connection between DMC and economic prosperity: "DMC is discussed as a potential biophysical equivalent to gross domestic product (GDP)." Industrialized nations, accounting for about 15% of the world's population, consume about 50% of key resources such as fossil fuels, industrial minerals, and metallic ores (Krausmann et al., 2009). A positive correlation between DMC and prosperity is also reported by Giljum et al. (2014).

H2.1.Depending on the level of prosperity, urbanization affects the DMCm
Cities around the world differ significantly in their prosperity levels. In poverty contexts, cities grow differently, are structured differently, are organized differently, and look different from rich countries (Swilling et al., 2018). It can be expected that these differences influence the relationship between urbanization and construction material consumption. It is therefore assumed that the level of prosperity has a moderating effect on the relationship between the degree of urbanization and construction material consumption.

2.2.3
Geographical context factors influence a country's consumption of construction materials H3.Frequent earthquake events drive the DMCm of a country Swilling et al. (2018) point to a potential increase in material consumption due to adaptations to climate-related impacts such as rising sea levels or heat stress. In addition, cities prevent or mitigate other natural disasters, such as heavy rain, hurricanes, earthquakes, or floods, with structural measures. The impact of earthquakes on the consumption of construction materials will be considered here as an example taken from the group of geographical context factors. We assume that in countries with frequent earthquake events, constructional adaptations to this threat will be developed. This is done by building more massive, particularly deep foundations to anchor tall buildings, which are predominantly located in urban areas Tokimatsu et al., 2013). Such adaptations can increase the DMCm.

Descriptive statistics
In descriptive statistics, the data set is characterized in terms of the distribution of its variables. The aim is to achieve the clearest possible presentation of the available data using tables and graphs.

Analytical statistics
The analytical examination of the data aims to draw generalizable conclusions from the data. Correlations and regression analyses are used to test the hypotheses.
Step by step, a regression model is built to track changes in the coefficients and their significance values. The final model is presented in a path diagram.
In the following calculations, the standardized regression coefficients (beta) are presented, as the comparison of the coefficients with each other is of particular interest.
Linear regressions provide meaningful and unbiased results if best linear unbiased estimator (BLUE) assumptions are met. This check was carried out (for more details see Supporting Information). High correlations among the independent variables may indicate multicollinearity. For micro-multicollinearity reduction, the independent variables involved in the interaction term were centered (Iacobucci et al., 2015) and denoted by c_variable. This concerns the variable level of urbanization and prosperity.
In the present data, there is a higher correlation coefficient (r = 0.714; p = 0.000) for the variables c_prosperity and c_urbanization, which could lead an indication of multicollinearity and related imprecise p-values. Therefore, in addition to the correlation coefficients, possible multicollinearity was also tested with the variance inflation factor values (VIF) (VIF max: 3.19) and condition indexes (CI max: 4.023) (Supporting Information S2, Tables S1, S2) with the result that multicollinearity can be almost ruled out.
Deviation from the required prerequisites was found with regard to homoscedasticity. Following Field (2018), this could be significantly minimized by a logarithmic transformation of the dependent variable DMCm ( Figure S2). In the regression equations, the consumption DMCm is therefore expressed as Lg DMCm. The general linear regression model is represented by the following equation (Chatterjee & Hadi, 2012) (Equation 1).
The equation describes the regression line that best anticipates the data. Y is the anticipated value of the dependent variable Lg DMCm. X (1...n) represents the independent variables. The standardized coefficients β (1...n) of the independent variables 1 to n, describe the slope of the regression line. The sum of the deviations of individual values from this calculated ideal regression line, which are not explained by the equation, is the residual e.
We hypothesized in Section 2.2.2 that prosperity moderates the relationship between urbanization and DMCm. To test this statistically, an interaction term is included. This is a product of the independent variables urbanization (x 1 ) and prosperity (x 2 ) (Preacher et al., 2006). The resulting extended equation (Equation 2) is: Following the hypotheses, the final regression model integrates the influences of urbanization, prosperity, and the interaction of urbanization and prosperity, as well as road network and the number of earthquake events. As another independent variable, the size of the country was included as a control.

Dependent variable-Construction material consumption
As mentioned above, we focus on non-metallic mineral construction materials. Lieber (2019) and Lutter et al. (2016) include the following materials in this group: asphalt, chert and flint, common clay, clay for bricks, crushed stone, igneous rock, lava sand, limestone, marl, shell, loam, marble, travertines, sand and gravel, sandstone, slate, and turfaceous rock. In addition, the trade balance of the raw materials is taken into account for the calculation of the variable domestic material consumption. The calculation follows the formula used in economy-wide material flow analysis (ewMFA) according to the OECD (2008) (Equation 3).
DMCm: Domestic material consumption of non-metallic mineral construction materials.

Extraction (m) (used):
Extraction of non-metallic mineral raw materials mainly used in the construction sector.

Exports (m): Exports of non-metallic mineral construction materials
Imports (m): Imports of non-metallic mineral construction materials.
Data on DMCm is available from the Global Material Flows Database for 216 countries (UN Environment, 2018;International Resource Panel, 2023). This corresponds to 99.3% of the world's population. A description of the data is given in Lutter et al. (2016). We relate the DMCm to the total population of the respective country as a specific quantity with the unit tons of material per capita. The population figures for the countries are based on the World Urbanization Prospects (UNDESA, 2020) for the reference year 2010.
In this study, we describe the level of urbanization following the UN definition (UNDESA, 2019c). This is due to the wide availability of data for this indicator (number of countries with available data or share of world population covered), as well as its potential suitability for time series analysis in future studies. Level of urbanization provides a basis for the United Nations Urbanization Reports (UNDESA, 2018a). The indicator expresses urbanization as the percentage of urban population in the total population of a country. Current level of urbanization values are available for 232 countries (99.9% of the world's population). The main criticism of this indicator is that "urban" is defined according to the respective country's own definition (Buettner, 2015;Cohen, 2006). The values of the indicator are available from 1950 onward (UDESA, 2018b). A discussion of the other indicators mentioned is given in the Supporting Information of this paper.

Road network
Since the length of the road network strongly depends on the country size, it is normalized to km road network km 2 land area . The data for this is taken from the CIA World Factbook (CIA, 2019) and is available for 216 countries, which corresponds to 98.5% of the world's population. GDP focuses on the economic performance of societies. However, this is driven by individual sectors without taking non-economic welfare aspects into account. Indicators such as the HDI, Prl, and MPI follow a broader understanding of prosperity and take into account non-monetary factors such as educational opportunities, health care, life expectancy, or access to drinking water (UNDP, 2018) and thus better represent the living standard in a country (Legatum Institute, 2021;OPHDI, 2018;Sever, 2013;UNDP, 2019). We assume that this broader understanding of prosperity is reflected more comprehensively in the materiality of buildings and infrastructure. Due to that and the better data situation (Sever, 2013)

Control variables
Country size is included as control variable in the regression models.

Variables and data used-General overview
All data on the variables described are taken from open data sources (Table 1). In each case, the reporting year closest to 2010 was chosen.

Results of the descriptive statistics
On average, the DMCm is 3.76 tons per person per year ( Table 2). The maximum is 18.860 t/cap in Singapore. 1 In extreme contexts, the DMCm tends toward zero (Palau, a Pacific Island country). A graphical representation, as well as a description of the frequency distribution of the variables DMCm, level of urbanization, Lg DMCm, and c_level of urbanization can be found in detail in Supporting Information S1, Figure S5. Table 2 shows the results of the descriptive statistics of all variables.

Results of the correlation analysis
Scatter plots give an impression of the correlations between the variables. We illustrate this using the focus variables level of urbanization and prosperity (raw data), each plotted in relation to the DMCm (Figure 1a,b). Since some variables had to be transformed and cleansed (consider- Pearson's correlation coefficient (p ≤ 0.05)*, (p ≤ 0.01)**, (p ≤ 0.001)***, two-sided; N = 123.
ing data gaps and outliers) to do the regression analysis (Section 2.4), the scatter plots of the transformed variables are also shown (Figure 1c,d).
DMCm tends to increase with increasing urbanization rates, but the spread of DMCm over the range of level of urbanization values is very broad (Figure 1a). This indicates the significant influence of further variables. In high prosperous countries, high consumption of construction materials is possible, but does not necessarily occur. In low prosperous countries, on the other hand, construction material consumption is limited to low values ( Figure 1b).
The correlation analysis shows a moderately strong positive correlation between c_level of urbanization and Lg DMCm (r = 0.590; p = 0.000).
Centered prosperity and construction material consumption have a very strong positive correlation, as already indicated in the scatterplot ( Figure 1b) (r = 0.808; p = 0.000). However, the correlation between the length of the road network and the consumption of construction materials (hypothesis 1.1) is only weakly positive (r = 0.335; p = 0.000). Furthermore, the strong positive correlation between the centered urbanization rate and centered prosperity (r = 0.714; p = 0.000) is noteworthy. All other independent variables show correlations below r = 0.5. Table 3 summarizes the results of the Pearson correlation analysis for all variables examined on the final regression model 3. Figure 1 are available in Table S1 of Supporting information S2.

Results of the regression analysis
Hypotheses H1, H2, H2.1, and H3 are tested using the regression model presented below. In order to enable comparability between the coefficients, standardized regression coefficients (beta) are given (Table 4). The significance level is indicated by asterisks. A detailed description of the individual models, with non-standardized coefficients and exact p-values, is provided in the Supporting Information (Table S6). In step 2 (M2) we include prosperity, which is addressed in H2. Here the influence of the urbanization rate is considerably lower than in M1 (β = −0.004) and has become statistically insignificant (p = 0.927). In contrast, the strongest and highly significant influence on the dependent variable is shown by the added variable prosperity (β = 0.894; p = 0.000). The effect has thus shifted from urbanization rate to prosperity. The introduction of the prosperity factor also results in an increase in the corrected R 2 to 0.65. The control variable shows no significance in this model. A significant influence of earthquake frequency on construction material consumption (H3) could not be proven in the models. Road network and total land area also show no significant influence in the models.

Results of the moderator analysis
The regression analysis indicated a significant moderator effect of prosperity and urbanization level in M3 with a negative sign (β = −0.119; p = 0.017) (more information about significant regions of the effect is provided in Table S8 of Supporting Information S2). This implies that as prosperity increases, the coefficient between urbanization level and DMCm decreases. The effect of urbanization on construction material consumption thus changes depending on the level of prosperity, as illustrated in Figure 3, taking three prosperity levels as an example. The solid line shows the effect for high prosperity, the dashed line for medium prosperity, and the dotted line shows the relation between urbanization and DMCm under low prosperity.
Material consumption per capita is at a high level in countries with high prosperity and correspondingly lower in medium-and low-prosperity countries-regardless of the level of urbanization. However, the effect of a change in the level of urbanization on material consumption varies greatly. In countries with high prosperity, material consumption per capita decreases as the level of urbanization increases. In contrast, in lowprosperity countries material consumption rises with increasing urbanization. In countries with medium prosperity, there is no evidence of the influence of urbanization on material consumption.

F I G U R E 3
Visualization of the moderator effect (c_level of urbanization * c_prosperity on Lg DMCm). Underlying data for Figure 3 are available in Table S3 of Supporting Information S2.

Selection of indicators
The three indicators considered to describe urbanization differ in their definition, their coverage of countries, and the timeliness of available data or the periodicity of their reporting. Each of the indicators shows weaknesses and imprecision. The heterogeneous definition of "urban" may lead to uncertainties in the data, but comparison with alternative urbanization indicators based on objective data (agglomeration index, gridded population) showed very high comparability (Supporting Information S2, Table S4; Supporting Information S1, Figure S4). We therefore consider these uncertainties to be acceptable. Thus the decision on which indicator to use can be made primarily with regard to the availability of data.
With respect to prosperity we argue in favor of the HDI because it reflects a broad understanding in this regard. Also, in this case, the three indices strongly correlate with each other (Supporting Information S2, Table S5). Thus, similar to the statements made above, we assume a robust baseline for the investigations carried out and again emphasize the importance of data availability for the selection of indicators.
Data for DMCm were taken from a globally recognized information source that follow the logic of an ewMFA, which is guided by top-down principles (Augiseau & Barles, 2016;OECD, 2012). The strength of this approach is the comprehensiveness in its coverage of societal material flows. In contrast, the degree of differentiation of material flows is limited (Schiller et al., 2017a). However, for non-metallic mineral materials, we estimate the resulting blurring to be very low. For example, data from Haas et al. (2015) referring to the EU-27 show that approximately 92% of all non-metallic mineral building materials can be allocated to the construction sector. Schiller et al. (2015) assume significantly higher shares.
Uncertainties also arise from incomplete data. This occurs, for example, when there are activities outside the legal framework, such as illegal mining, which is particularly widespread in developing countries (Filho et al., 2021). Data gaps can also be observed in trade balances, but since the bulk construction materials considered here generally have fairly short transport distances (Giljum et al., 2014;Schiller et al., 2017b), it can be assumed that these uncertainties will only be significant in very small countries.
The uncertainties mentioned can be overcome if the building material consumption is calculated according to the bottom-up principle, using material-intensity coefficients for typical structures to make projections for specific reference areas (Ortlepp et al., 2015a). Applications can be found at different levels, from neighborhoods, cities, and regions to global perspectives (Pauliuk et al., 2021;Peled & Fishman, 2021;Schiller, 2007;Stephan et al., 2022). An advantage of these approaches is that they provide a high degree of differentiation of materials; the disadvantage is mainly the incomplete representation of material stocks and flows, especially due to gaps in knowledge about the full extent of built structures and their dynamics, the crucial basis for extrapolation. Schiller et al. (2017a) assume an under-reporting of material flows by more than 40% compared to the complete picture provided by an ewMFA. Further challenges lie in locally different construction methods, which can only be approximated with the material indicators used and thus represent a further source of uncertainty (Heeren & Fishman, 2019;Schiller et al., 2019). In this light, the choice made in this paper to operationalize building material consumption through the DMCm is a more appropriate approach.
The other indicators used are less critical. One exception is the indicator "earthquake frequency." We use this indicator because correlations between the earthquake risk and the materiality of buildings are reported in the literature. In our study, we could not prove this correlation. We assume this is because earthquake risk areas are usually much smaller than the nations boundaries and found confirmation for that in CEN (2004) where it is mentioned that seismic risk greatly varies within countries. Thus, it is the study design or the spatial reference unit used that hampers the detection of a correlation. Nevertheless, we assume that due to the clear correlations between urbanization, DMCm, and prosperity no fundamental biases in the results would be expected, even with smaller-scale studies.
For some indicators, different reference years for the data (indicated in Table 1) are further potential sources of uncertainty. This may have different effects depending on the indicator and the development phase of the countries concerned. In our data and model applications we could not recognize systematic distortions and found that deviations from the general base year (2010) mostly affect only a few individual countries; thus, we conclude that this phenomenon is unimportant with regard to the general statements of the study.

Urbanization alone does not adequately explain the increasing consumption of construction materials
The results of both the correlation analysis and the regression calculations (M1, Sections 4.2.1 and 4.2.2) show significant influence of urbanization on material consumption. This statement can be found implicitly or explicitly in numerous sources in the literature (Ansari et al., 2020;Huang et al., 2010Huang et al., , 2018Sahoo et al., 2021;Yutong et al., 2021;Zhao et al., 2020). Our results clearly demonstrate that urbanization alone does not adequately explain the increasing consumption of construction materials and that the relationship is overestimated with simple models. More complex models (M3) have significantly stronger explanatory power. Considered separately, the influence of urbanization is very small, which could be due to collinearity. However, this does not affect the explanatory power of the overall model. Only the interpretation of the influence of individual variables would be limited. The interaction effect between urbanization and prosperity, on the other hand, can be demonstrated and interpreted contextually.

Prosperity level and consumption patterns induced by it are crucial
M2 indicates that prosperity has a stronger influence on material consumption than urbanization. We were able to specify this further by including an interaction term in M3 and applying a moderator analysis (section 4.2.3).
In nations with high prosperity, each individual consumes significantly more materials than in nations with low prosperity, regardless of the level of urbanization. Swilling et al. (2018) share similar results on the relationship between urban-specific DMC and prosperity by world regions. In more prosperous nations, construction is more solid, more generous, and thus more material intensive, in both rural and urban regions, than in less prosperous nations.

Complex interaction between prosperity and urbanization
In a context of high prosperity, specific material consumption decreases as the level of urbanization increases. This reflects efficiency effects resulting from compact construction methods with high settlement densities and decreasing floorspace consumption per capita, compared to dispersed low-density structures; associations that were intensively discussed in particular in the cost of sprawl debate at the turn of the millennium (Burchell et al., 1998) and further debates more recently (Duranton & Puga, 2020;Glaeser, 2011).
In contrast, among less prosperous countries, highly urbanized countries show higher material consumption values per capita than countries with a rural structure. Overall, however, material consumption in this group of countries remains at a low level. Reasons for this could be the predominance of simple construction methods in rural areas or informal settlements, with a high proportion of non-mineral materials and undocumented construction. Massive "urban" construction methods, on the other hand, result in significantly higher demand for non-metallic construction mineral materials (UN Habitat, 2014).
Urbanization policies in less prosperous countries are often aimed at improving the prosperity of the population (Huong et al., 2010;UNDESA, 2019b). If an increase in prosperity occurs in the course of urbanization, urbanization and prosperity effects reinforce one another and material consumption increases dramatically. Schiller et al. (2018) show that countries with high urbanization dynamics have significantly lower material consumption than countries with low urbanization dynamics. The latter tends to have high urbanization rates. This indicates the dramatic growth in  Table S4 of Supporting Information S2.
demand for construction materials and increase in regional scarcities that will occur if we do not succeed in decoupling the resource requirements of building structures in countries with high urbanization dynamics from desired prosperity effects.

Regional disparities
The studies presented in this paper cover almost all nations of the world. Regional analysis was not the subject, but it is possible to get a first impression by a simple spatial representation of the indicators used, which have emerged as central to the results. Figure 4 shows a chart of the indicators level of urbanization, urbanization dynamic, HDI, and DMCm.
A visual view provides an impression of the existing spatial disparities in the manifestations of these characteristics: high levels and low dynamism of urbanization, high prosperity and quite high material consumption in central and northern Europe and North America; high urbanization dynamics, comparatively low prosperity and low to medium material consumption in many regions of Asia and Africa; moderate urbanization and prosperity levels but very high urbanization dynamics and high material consumption in South-East Asia; high urbanization levels with low prosperity and medium material consumption in South America. These spatial differences, as well as the different effects of urbanization depending on the context of prosperity, indicate that approaches to global resource conservation must be differentiated according to the specific spatial context.

5.6
Declining flows due to saturation effects?
At least for construction material stock, there are indications that stock-related saturation processes are occurring in highly urbanized regions and countries such as "the Industrial Old World" (Wiedenhofer et al., 2021), or European countries and Japan (Miatto et al., 2016). However, our results suggest that this saturation of stocks does not necessarily lead to a reduction in flows, but rather that material flows remain at a high level. This could be due to high throughput induced by maintenance and renewal of the existing building fabric and the underlying linear material management concepts. Circular management concepts that aim for consistent and circular-oriented management of the existing building fabric could provide a remedy here.

CONCLUSIONS
The construction industry is one of the main consumers of non-metallic mineral materials and urban construction is usually quite different from construction in rural structures. In this respect, it is obvious and confirmed in this paper that urbanization influences the consumption of these materials.
Urbanization is also closely related to prosperity. Prosperity, in turn, has a great influence on the consumption of mineral construction materials and also has a moderating effect on urbanization and material consumption. The aim of this paper was to analytically explore these interrelationships in their complexity, to empirically test them and find explanations.
A focus on urbanization as the central driver of the worldwide increase in material consumption falls short and leads to overestimation. The same applies to the prosperity factor. Only the combination of both characteristics provides convincing explanatory patterns and offers a starting point for meaningfully differentiating strategies for sustainable construction in light of the megatrend of urbanization.
The figures indicate that the "hunger for resources" of the urban environment will not diminish with a "business as usual" approach. This is by no means natural or inevitable, however. Two points must be given particular attention: First, circularity should not be limited to the recycling of materials. Rather, the entire range of strategies currently being discussed under the concept of circular economy must be taken into account, slowing and narrowing are required in addition to recycling or closing (Bocken et al., 2016;Gallego-Schmid et al., 2020). Second, it is imperative to differentiate corresponding strategies and circular business models contextually, that is, spatially. For science, the study provides new empirical and methodological findings, and accordingly spatially specific circular economy strategies can be differentiated and specified. Without the involvement of political and societal actors, such a far-reaching transformation process cannot succeed, which is the immediate conclusion on the paper's political relevance: approaches such as the decoupling of material consumption and prosperity or the implementation of circular concepts in built environment must adequately acknowledge these different framework conditions in order to develop sufficient clout in terms of effective resource protection through "circular urbanization."

ACKNOWLEDGMENTS
The research was supported by basic funding of Leibniz Institute of Ecological Urban and Regional Development, Dresden. We would like to thank Tamara Bimesmeier, Marius Röhnisch and Anna-Maria Weber for their manifold support in the acquisition of data and in the selection of indicators.
Thanks are also due to Jörg Hennersdorf for the preparation of the maps and to Martin Behnisch for fruitful discussions on the regression models.
Open Access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data supporting the findings of this study are freely available in UNDESA (2019a) (https://comtrade.un.org/data), UNDESA In advance of the calculations, four extreme outliers of the DMCm were excluded (for details see Supporting Information S1, Figure S6).