The awareness that human behavior is affecting the environment is not a unique attribute of these days. Probably, the human-environment interrelationship is present as never before due to the tremendous adverse effects of climate change in the near future. Humans have become the single most influential species on earth so scientists are even discussing to assign the term "anthropocene" (stating that human activities are the dominant influence on climate and environment causing land surface transformation and changes in the composition of the atmosphere) to the current geological epoch (Crutzen and Stoermer, 2000; Lewis and Maslin, 2015). Basically, the general human-environment interrelationship has been recognized and described for a long time in history. For example, Seneca the Younger (4 B.C. - 65 A.D.) already notes that there exists a relationship between human activities and environmental phenomena. Specifically, he argues that household cooking fires, traffic or burning of dead bodies is correlated with the pollution in Rome (Seneca, 1971[62 A.D.]). Thomas Malthus (1766 - 1834) is another popular example being part of this tradition. He asks how population growth affects the availability of resources needed for human welfare (Malthus, 1960 [1798]). Though Malthus’ concerns, which basically address the incapacity of fixed land feeding an exponentially growing population were considered to be wrong due to even higher growth rates of total production (Galor and Weil, 2000), the debate regarding the human-environment nexus does not disappear. Rather, it shifts to topics like the depletion of natural resources (e.g. fossil fuels) or the degradation of renewable resources (Panayotou, 2000). With the writings of classical economists like Malthus, "the population-resource link receives systematic attention" for the first time (Dietz and Rosa, 1994, p. 278). In addition, also scientists from other disciplines were inspired by Malthus. For example, Charles Darwin (1809 - 1882) is driven by the same basic thought when he argues that population pressure on critical environmental resources drives evolutionary changes (Darwin, 1958 [1859]). So, various sciences (like social or biological ones) try to understand a similar phenomenon from their point of view throughout history. But, a systematic investigation of interrelations between human behavior and the environment is ignored for a long time. The raising discipline of ecology in the 20th century investigating the relations between organisms and environment slowly changes this status (Dietz and Rosa, 1994). Since then, many efforts are made in order to understand the mechanisms of human activities on the environment by combining the insights of biologists, ecologists and environmental scientists. Theoretical ideas are gradually transformed into models in order to determine and analyze the response of environmental change to a set of potential anthropogenic factors. One important objective of these models is to deliver policy recommendations based on robust (empirical) estimations (Schneider, 2022). About fifty years ago, Ehrlich and Holdren (1971) propose the idea of IPAT (environmental Impacts of Population, Affluence and Technology) in order to formalize the relationship between human activities and the environment. The IPAT model is based on the simple but plausible assumption that population, affluence and technology must be part of any serious effort to understand human impacts on the environment. Thus, the IPAT model provides a useful starting point for structuring this debate (Dietz and Rosa 1994). Shortly after, Commoner at al. (1971) firstly formulated the IPAT model as algebraic equation. This mathematical accounting identity specifies that environmental impacts are the multiplicative product of population, affluence and technology that allows to solve for any variable of interest. For example, the IPAT identity has often been used to calculate the term of technology (given the remaining components; e.g. Raskin, 1996). Specifically, population is conceptualized as population size, affluence as per capita consumption or production and technology as environmental impact per unit of production. The main strengths of the IPAT model are the parsimonious and clear specification of anthropogenic driving forces affecting the environment as well as the implication that these driving forces do not influence impacts independently due to their multiplicative interconnectedness. However, the pure formalization of a functional relationship between variables does not allow any hypothesis testing or causal interpretations (York et al., 2003). Further, there may be different underlying functional assumptions (e.g., non-linearities) or other potential driving forces affecting the environment. Obviously, the relative tight framework of the IPAT model cannot address these issues. Consequently, Dietz and Rosa (1997) develop a stochastic version of the IPAT model by transforming it into the STIRPAT (STochastic Impacts on the environment by Regression on Population, Affluence and Technology) model. The STIRPAT model allows empirical analysis and thus builds a powerful and flexible framework for hypothesis testing (Liddle and Lung, 2010). Indeed, many studies use the STIRPAT framework for broad empirical applications, such as global and regional analyses or the assessment of various anthropogenic driving forces (e.g., see Vélez-Henao et al., 2019 or Schneider, 2022 for literature reviews regarding STIRPAT studies). The major part of studies using the STIRPAT approach estimates environmental impacts with respect to the principal of greenhouse gas emissions, i.e. CO2 emissions. CO2 emissions are a globally accepted measure of environmental outcome in order to quantify climate policy goals. Moreover, there exist accurate and sound data for CO2 emissions for almost all parts of the world. However, also alternative measures for the environmental outcome like (variants of) the ecological footprint or different air pollutants (e.g., NOx or SO2 emissions) are analyzed within STIRPAT applications (Vélez-Henao et al., 2019). Traditionally, the STIRPAT approach is used to estimate the so-called ecological elasticities. The ecological elasticities indicate the percentage change in the environmental variable associated with a one percentage point increase in the respective explaining variable, holding the effects of the other explaining variables constant (Knight et al., 2013; for example, many studies found that a 1 percent increase in population increases CO2 emissions by about 1 percent). Generally, studies find that both population and affluence (typically operationalized as number of residentials and GDP per capita, respectively) are significant drivers of emissions. In contrast, most applications do not explicitly estimate the impacts of technology mainly due to the missing consensus on valid indicators for technology (Knight et al., 2013). So, technology is usually seen as included in the error term of the regression equation or (partly) captured by additional explanatory variables. Therefore, the empirical application of the STIRPAT model allows for the inclusion of additional potential driving factors into the analysis beside the three core components (population, affluence and technology) and thus offers a high potential for extensions compared to the benchmark framework (Wu et al., 2021; Schneider, 2022). Additionally, the approach encourages the investigation of environmental impacts regarding the three core components in more detail. For example, the components can be disaggregated into forms that have more social meaning (Rosa and Dietz, 1998). All in all, the STIRPAT model represents a strong tool for various applications. However, "[…] despite the multiple applications and the high potential of the STIRPAT model, inconclusive results and/or knowledge gaps remain […]" (Vélez-Henao et al., 2019, p. 1). The inconclusive results are mainly due to different model specifications (e.g. the treatment of technology), different underlying samples (regional or global data), different estimation techniques or different periods of time. This dissertation (consisting of five related but individual contributions) contributes to the existing STIRPAT literature methodologically as well as conceptually in several ways. The first two contributions (sections 2. and 3.) principally address methodological challenges whereas the last three contributions (sections 4., 5. and 6.) mainly address conceptual issues of STIRPAT modelling and/or provide novel variations of application. The first contribution (section 2.) gives a complementary perspective when dealing with the relative importance between environmental impacts. The second contribution (section 3.) presents an alternative way of using the STIRPAT model with the focus on reversed causality. The third contribution (section 4.) deals with the varying roles of the dimensions of affluence on the environment. The fourth contribution (section 5.) differentiates between settlement structures when analyzing human impacts on air pollution. The last and fifth contribution (section 6.) covers the effects of technological progress on the environment. To begin with, the first contribution (The role of demographic and economic drivers on the environment in traditional and standardized STIRPAT analysis; see section 2.) shows that the STIRPAT analysis should at least be complemented with standardized coefficients if the research focus lies in the assessment of the relative importance between the driving forces. Most studies find higher ecological elasticities related to population compared to GDP per capita growth. Hence, some authors suggest to mitigate the trade-off between economic growth and environmental pressure by giving priority to population policies and reducing population growth in first place. However, the question of the predictor variables’ relative importance cannot be finally answered by this approach. In response, the contribution complements the traditional ecological elasticities by the calculation of standardized β-coefficients (for a sample of 84 countries and the period between 1980 and 2014). Results indicate that GDP per capita rather than population growth matters more for explaining environmental impacts. Admittedly, interpretation of standardized coefficients is not without limitations. First, they are sample-specific and cannot be compared across different studies. Second, a predictor variable might not affect the environment only on its own, but joint impacts could be present. The contribution addresses these problems and provides a careful interpretation of and comparison between non-standardized and standardized β-coefficients. Overall, there is good reason to assume that environmental impacts can be reduced more readily by a policy giving priority to economic rather than population growth. The second contribution (Reversed STIRPAT modelling: the role of CO2 emissions, population and technology for a growing affluence; see section 3.) challenges the prevailing assumption of STIRPAT modelling, which in most cases proposes a one-way causality running from the anthropogenic factors to the environment. However, the rich portfolio of theoretical and empirical studies reveals no universal direction of causality between economic growth and the environment, findings rather depend on the considered time periods and countries’ stage of development and sectoral structure (Costantini and Martini, 2010; Ozturk, 2010). Consequently, the contribution proposes to add a new perspective to the IPAT/STIRPAT approach by setting up a stochastic model that explains impacts on economic growth (affluence) by regression on population, CO2 emissions (as a proxy for energy use or ecosystem services) and technology. Indeed, the applied Granger-causality tests indicate a reversed causal relationship. Therefore, the relationship between economic growth, demographic development and CO2 emissions for 30 industrialized countries using time-series data from 1982-2014 in the IPAT/STIRPAT setting is analyzed. The results confirm that GDP per capita growth rates of highly industrialized economies are significantly driven by the development of CO2 emissions, population and energy intensity. Coefficients remain robust with or without integrating structural and energy variables and for the short- and long-run perspective. Thus, the significant and robust regression results in all model variants demonstrate the reasonableness of applying this setup in addition and complementary to the traditional STIRPAT model. In addition, the findings confirm the ongoing high dependence of advanced economies on the availableness and consumption of cheap energy. The empirical findings of most STIRPAT studies show positive impacts of both population (commonly measured as number of residentials) and affluence (commonly measured as GDP per capita) on the environment independent of the model setup or underlying dataset. Furthermore, some studies are examining the effects of population on the environment in more detail. Thereby, authors differentiate population by region, economic status, settlement structure, age group or educational achievement (e.g., Cole and Neumayer, 2004; Liddle and Lung, 2010). In contrast to the more differentiated investigations of the environmental effects of population (and technology), affluence is almost only analyzed by GDP per capita. Obviously, GDP per capita is a very convenient measure of affluence. But, this measure alone potentially neglects the possibility that increasing affluence affects the environment in varying - even opposing - ways. Interestingly, already the initial concept of the IPAT model suggest to think about affluence as some measure of (national) production and consumption patterns (Dietz and Rosa, 1994). Hence, the third contribution (The varying roles of the dimensions of affluence in air pollution: a regional STIRPAT analysis for Germany; see section 4.) analyzes the role of affluence for the production of local NOx emissions in a more differentiated way. The study addresses this gap by decomposing affluence into three dimensions-income per taxpayer, private car ownership, and the share of single-family houses-and analyzing their roles in the production of local NOx emissions. Results for 367 German districts and autonomous cities between 1990 and 2020 indicate that private car ownership per capita and single-family houses per capita can indeed be considered drivers of local pollutants. In contrast, income per taxpayer has a negative impact on NOx emissions. While private car ownership and single-family houses could reflect the material- and energy-intensive part of affluence, taxable income per taxpayer might cover (if we control for car ownership and the housing situation) expenditures for material (e.g., food, consumables) as well as types of consumption more common among the financially affluent (e.g., services, cultural activities). The fourth contribution (Drivers of local air pollution: a regional STIRPAT analysis for Germany; see section 5.) offers an assessment of the role played by population, economic growth and technology change in the evolution of local air pollution, using the STIRPAT approach at the district level (NUTS 3). The analysis covers the development of 367 German districts and autonomous cities between 1990 and 2020. This procedure does not only allow for an analysis of the cities but also the rural districts. Further, the contribution analyzes the estimated environmental elasticities in detail by controlling for non-linear impacts. In this context, predicted margins of environmental elasticities are calculated. Results indicate that the development of local pollutants (NOx emissions) is clearly related to car ownership, regional population and industrial manufacturing. While the findings largely hold for urban and rural districts, they also indicate that environmental impacts depend on the types of regions for GDP per capita and urban density. For example, a negative environmental impact of urban density can be shown for rural but not for urban districts. Finally, the predicted margins analysis indicates that the effect of population on the environment strongly depends on its respective level. So, high percentiles of population reveal a (much) higher marginal impact compared to low percentiles. Finally, the STIRPAT model "[…] offers a valuable yet underused platform to address the (environmental) rebound effect […]" (Vélez-Henao et al., 2019, p. 1378). Traditionally, the rebound effect describes the change in overall consumption and production as a consequence of a change in economic variables induced by a change in the energy efficiency (Font Vivanco and Voet, 2014). The environmental rebound effect provides a more holistic perspective and thus expresses the rebound effect through different environmental dimensions like emissions (Vélez-Henao et al., 2019). Typically, improvements in productivity induced by technological progress are seen as promising measure in order to mitigate the adverse effects of climate change (IPCC, 2018). Against this background, the fifth contribution (The effects of technological progress on CO2 emissions: a macroeconomic analysis; see section 6.) analyses the effects of (technological) productivity increases on the environment and thus tries to clarify whether an environmental rebound effect exists. Specifically, the effect of improvements in carbon intensity (defined as "resource productivity" and estimated directly by CO2 emissions per GDP) and in overall productivity (defined as "factor productivity" and estimated indirectly by decomposition of a production function) on CO2 emissions is investigated. Therefore, data from 118 countries between 1962 and 2014 are analyzed. Findings indicate that there exists an environmental rebound effect regarding an increasing carbon intensity. So, improvements in carbon intensity lead to comparable underproportional decreases of CO2 emissions. Further, improvements of overall productivity could even lead to higher CO2 emissions (backfire-effect, i.e. rebound effect > 100 percent). In summary, technological progress in terms of productivity improvements cannot solve the environment-growth trade-off per se. All in all, the five contributions of this dissertation address several research questions regarding the human-environment nexus in the context of STIRPAT modelling. Moreover, all contributions provide environmental policy implications by taking the interaction between the effects of population, affluence and technology into account. Obviously, theoretical and empirical questions still remain to be solved in order to fully understand the complex human-environment relationship. There is no doubt that there are “many avenues for future expansion of the STIRPAT model” due to its wide flexibility in application (Kilbourne and Thyroff, 2020, p. 360). So, future STIRPAT studies can play a crucial role in gaining a comprehensive understanding of anthropogenic impacts on the environment and thus can help to deal with prospective challenges related to the multifaceted human-environment nexus.    «

The awareness that human behavior is affecting the environment is not a unique attribute of these days. Probably, the human-environment interrelationship is present as never before due to the tremendous adverse effects of climate change in the near future. Humans have become the single most influential species on earth so scientists are even discussing to assign the term "anthropocene" (stating that human activities are the dominant influence on climate and environment causing land surface trans...    »