A systematic review of forest area development drivers estimated under the concepts of environmental Kuznets curve and forest transition hypothesis

This systematic review follows the principle guidelines of the environmental evidence reporting standard 'RepOrting standards for Systematic Evidence Synthesis (ROSES)' (Haddaway et al 2017). In the sense of the ROSES guidelines, this study applies as a basis the article selection from the EGM published by Tandetzki et al (2022). Thereby, the EGM functions as research protocol and outlines the scope of research, the systematic search and selection strategy for relevant articles as well as the strategies for coding, and data extraction. In the present systematic review, the 46 articles identified as relevant by Tandetzki et al (2022) were used for further analysis. Additionally, before analyzing the content of these articles, an update of the article search was conducted to keep the article selection up-to-date. A brief overview of the basic methods used for carrying out the EGM and the update is given in section 2.2.

The EGM identifies which and how many articles have been published related to a particular topic, and, conversely, where there are publication gaps on the topic at article level. The EGM serves as an upstream methodological stage of the subsequent systematic review. The systematic review method is an comprehensive in-depth approach and synthesizes the evidence from the articles for a specific research question (Haddaway et al 2018). The differences between the EGM and a systematic review lie in the approach and purpose. The EGM represents the breadth and mass of literature on a topic, while a systematic literature review is focused on the depth and quality of the evidence and analyses the methodologies, assumptions, data and results extracted from the content of the articles.

While the EGM and its update identified the relevant research literature for the topic under consideration at the article level, the systematic analysis of the content of these articles is done in the present systematic review. Some articles contain several studies, e.g. different estimates of forest area development with different independent variables. To avoid multiplicity, we performed a study selection procedure as described in section 2.3 to extract relevant estimation studies from the preselected articles. After selecting all studies, we extracted the data, in particular the variations of the independent variables for each study, and then analyzed all studies comparatively. To accomplish this, we utilized a distinct data coding and extraction process (section 2.4) based on Tandetzki et al (2022) and a data grouping strategy (described in section 2.5) developed collaboratively within the author team of this systematic review.

2.2.1. Summary of methodologies applied in the EGM

2.2.2. Updating the database till 2022

2.2.3. Critical appraisal and final article selection

In contrast to the previously conducted EGM (Tandetzki et al 2022), the analysis of results was conducted at the study level. Therefore, a selection of relevant studies under avoidance of multiplicity was conducted in this systematic review. Thus, the unit of analysis is the individual study, defined as the single estimate of forest area development contained in the selected articles. Forest area development can be expressed as forest area, forest cover, change in forest area/cover, deforestation rate or any other specification that expresses the development of forest area (decline or expansion). Different studies presented within one article can be independent from each other and consider for instance different geographical regions, but may also multiply the same study concept several times by just exchanging one control variable. Thus, careful study selection was necessary to neither omit a relevant study, nor include studies that face multiplicity (López-López et al 2018).

As shown in figure 1, we considered six types of differences in study design as selection criteria across the selected articles. The selection criteria were applied to extract only those studies, that are unique in their characteristics and do not cause multiplicity, which would bias the analysis results. Thus, if individual studies within one article do not overlap, as they, e.g. consider different country groups or different time periods, we included all of them. For example, if one article examines low, middle, and high-income countries in individual studies, we selected all three of them. But, if the article also provides a study analyzing all countries together, we preferred this study to the studies of the individual country groups (see figure 1). For studies that utilize the EKCd concept, we always selected studies that examined both income and income² to fulfill concept's condition. However, if estimation methods are chosen that calculate a threshold, e.g. to satisfy the condition of the EKCd curve pattern, we considered those as well.

Zoom In Zoom Out Reset image size

2.3.1. Selection of basic and advanced studies

The analysis objective is to provide an overview of the independent variables tested in estimation studies to explain forest area development. The analysis aims to show which variables and variable combinations have been tested significant and how this is influenced by variations of model and variable specifications, and input data applied. Variations were analyzed with regards to different regions considered, dependent variable definitions, statistical estimation methods chosen and forest area development concepts analyzed. Advanced studies were examined separately from basic studies (simple models).

Judging the influence of socio-economic drivers on forest area development by comparing the studies is challenging due to the diversity of methods and assumptions. On the basis of their heterogeneity, studies cannot be merged in a unified data frame. Thus, we compared them using descriptive analysis and refrained from further methods such as meta-analysis. As the studies are heterogenous regarding the study design ((a) the chosen region, (b) the dependent and their tested independent variables, (c) the chosen concept, and (d) the estimation method), we examined these criteria separately.

The data extraction and coding strategy applied for this systematic review is based on prepared templates and was conducted in three screening steps (two on article level and one on study level). The screening procedures built on the methods applied and described in Tandetzki et al (2022). The data extraction within the screening steps is displayed in table 1.

While the screen on article level revealed the evidence base of the targeted forest area development concepts, the screen on study level focuses explicitly on the targeted estimation models and the examination of independent variables. During the latter, we systematically extracted all independent variables applied in the studies selected and merge them into a prepared coding sheet (see table 1). We distinguished between tested and statistically significant variables. All tested variables were extracted, even if not significant, to determine the causal relationship between variable elimination and variable adding. In addition, the significance level stated for each independent variable in each study was noted. Variables were ordinally coded according to the following scheme: (i) '1' for tested and not significant, (ii) '2' for tested and significant at the 10% interval, (iii) '3' for tested and significant at the 5% interval, and (iv) '4' for tested and significant at the 1% interval (more information in supplementary material (Tandetzki et al 2023)). If a variable of the coding list was not considered in the study, it was coded with zero.

We also extracted the coefficients of the individual drivers, but due to their heterogenous nature, we did not analyze them further. The coefficients undergo significant changes influenced by the study design. A cross-sectional analysis across all studies was not feasible. We only examined the signs of the individual coefficients in relation to basic studies, as the limited number of independent variables in these studies allowed for comparison.

2.5.1. Categorizing of variables

In the screening process, we identified 55 articles as eligible for the review analysis. From the 55 articles, 85 studies were extracted. The 85 studies were categorized into 14 basic studies and 71 advanced studies (figure 2). From the majority of articles, we chose one advanced study per article for subsequent study analysis. From eight articles (table 3), more than one advanced study was selected, as e.g. these articles presented numerous studies with different regional scope or for different time periods.

Zoom In Zoom Out Reset image size

An overview of all selected studies and the extracted data is available in the supplementary material (Tandetzki et al 2023).

In total, the 85 studies assess 157 ⁴ different independent variables. These 157 variables were categorized into 12 thematic categories (as displayed in figure 2) ⁵ . Across all selected studies of the systematic review, on average, basic studies tested two independent variables and advanced studies 7.2 independent variables. Not all of the tested variables were tested as statistically significant in the estimation studies considered. In total, 65% of variables proved to be statistically significant, of which 70% are highly significant at the 5% or 1% significance level.

During the screening process, we identified 14 basic studies (figure 2). Within one article, these basic studies have the same study design as the advanced studies in regard to, e.g. region and period selection and the choice of the dependent variable. The estimation method is also the same, except for Caravaggio (2020a), who uses a fixed effect estimator (FE) estimation for the basic studies and a PMG estimation for the advanced studies.

Based on the summary of the study designs of each basic study (see table 4), we appraised the explanatory content of the basic studies. Except for Köthke et al (2013), who analyze the FTH concept, all basic studies estimate the EKCd concept and thus the influence of income (expressed as GDP, GDP pc, GDP pc² or a combination them) on forest area development. Thereby, the influence of at least one of these income variables was tested as statistically significant in 11 cases. This finding underlines the important role of income in explaining forest area change. Judging the explanatory power of the basic models we found that five out of the 14 basic studies specify a R² as statistical measure for the goodness of model fit (Meyer et al 2003, Köthke et al 2013, Andrée et al 2019, Salahodjaev and Jarilkapova 2020, Assa 2021, Aquilas et al 2022) ranging from 0.234 to 0.726. The highest R² (0.726) was found for the global basic study by Köthke et al (2013), who estimate the influence of demography on forest cover (applying FAO data) for 126 countries for a 20 year period using the FTH concept. On the contrary, the three other global basic studies carried out by Meyer et al (2003), Salahodjaev and Jarilkapova (2020) and Assa (2021), who applied income instead of demographic variables and test the EKCd concept, yielded lower R² of 0.36, 0.23 and 0.24, respectively. Even though the limited availability of R² measures made it difficult to draw firm conclusions on the explanatory power, the low goodness of fit of the global EKCd basic studies is an indication that income is insufficient as a sole explanatory factor for EKCd development, and that the incorporation of other drivers can enhance the explanatory power. In the studies that provide explanatory values in the form of R² this value is generally relatively low overall, (except for Köthke et al (2013) and Aquilas et al (2022)). Only few studies display an adjusted R² value, which is slightly higher in advanced models. The sign of the coefficients provides insights into the direction in which drivers influence forest area development and, additionally, whether the EKCd hypothesis is confirmed or not. It is important to note the specific dependent variables under consideration. For the rate of deforestation (n = 6), approximately half of the signs are either positive or negative. The results are so heterogeneous that no definitive statement can be made.

In a first step, we analyzed the 12 thematic categories to which all independent variables were assigned in regard to the number of tested and statistically significantly variables. In the advanced studies on average 7.2 independent variables were tested, i.e. not every study tested every variable. The most frequently tested are income variables, with an average of two income variables per study (n = 71). In some cases, study authors use more than one variable of the same thematic variable category within a study. This is evident, e.g. in figure 3(a), where 63 of 71 studies included in this systematic review examined the influence of income and tested a total of 145 income variables. Of these variables, 66% showed statistically significant influence. Demographic variables were similarly tested at least once in 63 out of 71 studies, with a total of 97 demographic variables, of which 60% were statistically significant tested. It is noteworthy that some variable categories were rarely tested, but are strikingly often statistically significant, e.g. 14 out of 16 institutional economic variables, 10 out of 12 education variables, 6 out of 7 consumption-related variables, and 5 out of 7 production forest land variables were tested statistically significant.

Zoom In Zoom Out Reset image size

In a second step, we considered the individual independent variables. Figure 3(b) displays the 10 most often estimated individual independent variables (out of 157 different variables tested in total). GDP pc and GDP pc² were most frequently applied. In approximately 70% of cases these variables were tested as statistically significant. Even though less frequently tested, population density, trade openness, and institutions appear to be statistically significant in more than 60% of cases. Agricultural land area and rural population as percentage of total population were statistically significant in less than 50% of cases tested (figure 3(b)).

To better grasp and analyze the combinations of independent variables tested within one study, we used a network graph displaying the statistically significant variable combinations. The graph accounts for statistically significant variables, grouped at category level. The combination of significant variables of different categories within one study is represented by a connection (edge) within the graph. The more often a combination of variables has been shown to be statistically significant in studies, the thicker the edge. For example, income variables proved to be statistically significant most often in combination with demographic or trade variables (see figure 4).

Zoom In Zoom Out Reset image size

In the center of the network diagram are the most common and frequently estimated variable categories, e.g. income-related variables and demographic-related variables. The variable categories that have been statistically significant in only one or two studies are displayed in the outermost ring (e.g. education-related variables, consumption-related variables). If multiple variables from one category are tested within one study, a ring is represented above the nodes. This is particularly common for the income variables, where GDP pc and GDP pc² were jointly investigated in 86% of studies (figure 4).

The studies used in this review cover specific regions such as Middle East and North African (MENA-), Sub-Saharan African (SSA-), the organization for Economic Co-operation and Development (OECD-) countries, or continents and larger areas, such as developing countries or global data sets. We compared the study results from different geographical regions (see also figure 5(a)).

Zoom In Zoom Out Reset image size

The figure 5 displays the cumulative relative distribution of statistically significant drivers according to the 12 thematic categories. The drivers are represented in different colors in the bar chart, and the bars are thematically subdivided.

Figure 5(a) presents the distribution of influential drivers for larger spatial boundaries (global studies, developing country studies, and continental studies). Global datasets are most frequently analyzed among all studies (n = 23), followed by developing country studies (n = 9). The most frequently tested drivers for these large-scale studies are income-related, demographic, direct area-related, institutional, educational, and trade-related variables. Significant drivers from each thematic category were found for global studies (except for the category of other variables), while significant drivers were found for 9 out of 12 categories for developing countries. A comparison of the two large-scale groups (global and developing) shows that developing country studies more often identified trade as an influential driver than global studies, while the latter more often identifies direct area-related drivers to have an impact.

At the continental level, datasets are frequently examined for Asia (n = 4), Africa (here we also counted SSA-, MENA and the Congo Basin countries) (n = 9) and Latin America (n = 6), with most studies focusing on the African continent. Other continents (East Europe) were only considered in one study.

Africa was found to have the highest diversity of influencing factors in the studies reviewed, with drivers from 10 out of the 12 thematic categories showing a significant impact on forest area development. Studies on Latin America and Asia, in contrast, have shown no significant influence of consumption and other economic factors besides income.

As figure 5(a) shows, income and trade are the most dominant drivers in all three continents. Demographic factors, however, mainly influence forest area development for the Asian and African continents, and less frequently in Latin America. In contrast, direct area-related drivers were frequently identified in studies on Latin America, but seldom in studies on Asia and Africa. In addition to the described factors, institutional factors, institutional economic factors, and other factors have an influence on all three continents.

The heterogeneity in estimation models is also reflected in the choice of the dependent variable as representative of forest area development. A distinction is not only made between the definitions of the different dependent variables, but also between the data sources. In regard to the data source applied to determine forest area development we found that 40 studies use the FAO database, 16 studies use the World Bank data, five studies draw their data from the publication of Hansen et al (2013), three from the World Resources Institute, three do not clearly indicate the source and one study each uses the data from Zon and Sparhawk (1923), Sten (1956), the Environmental Performance Index (EPI), and from the land cover climate change initiative of European Space Agency.

The definition of the dependent variable is subject to the author's specification in each individual study, which we categorized into eight groups. The studies considered apply five forest variables already distinguished in Tandetzki et al (2022) (forest area, annual rate of forest change, forest cover, forest cover change, and rate of deforestation), net forest depletion (Amirnejad et al 2021, Ibrahim and Ajide 2022), tree cover loss from EPI (Le and Le 2022), and carbon emissions generated by deforestation per hectare (Zafeiriou et al 2022). Three studies apply agricultural variables (arable land and agricultural area change) as proxies for tropical deforestation (compare figure 5(b)). To detect the influence of socio-economic drivers depending on the definition of the dependent variable, we analyzed the role of the significant tested drivers with regard to the specification of forest area development.

In figure 5(b), the relative distribution of significant drivers with respect to the specification of the dependent variable are displayed. The rate of deforestation as dependent variable is included in most (n = 34) studies. The second most frequent applied independent variable is forest cover (n = 15), followed at a distance by 'all others', which assemble all studies that cannot be clearly assigned to any group (n = 6), and then by forest area (n = 5). The remaining specifications are rarely applied (less than five times).

For the rate of deforestation, drivers from each thematic category were tested significant at least once. Forest cover studies show a similar pattern expect from institutional and institutional economic-related factors, which were not tested significant.

The most frequently tested dependent variables, namely the rate of deforestation and forest cover, are dominated by income, demographic, and trade-related variables. A significant influence of direct area-related variables was found strikingly often on forest cover compared to rate of deforestation. Whereas institutional and institutional-economic drivers were more often found to influence the rate of deforestation.

Dependent variables which were applied by a few studies only exhibit a more diverse pattern. Due to the low number of observations, it is however difficult to draw valid conclusions on these. The pattern displayed for studies that examine forest cover change is strikingly different from that emerging from the other studies. These four studies are from the article of Rudel (1998), who considers in all studies an identical study design in different time periods, including the same tested variables. The low variety in identified drivers might therefore be attributed to the low variety of underlying articles and authors. When it comes to the agricultural variables serving as proxies, the influence of trade is dominate. However, these specifications of the dependent variable have only been applied three times in total.

The theoretical concept chosen by the authors of the studies selected was not always clearly assignable to the framework of the EKCd or FTH. Thus, although the concepts are differentiated in theory, a clear distinction was not always visible in the estimation. We assigned the studies to the concepts according to the definition the authors gave in the respective article. However, due to missing information (n = 3) or both concepts being investigated in a single study (n = 4), we assign seven studies to a 'mixed' category. The rest were divided into 76% EKCd studies and 16% FTH studies. Analyzing the sum of ECKd studies, we found that at least one driver out of each thematic category was significant. Income, demography and trade were most frequently found to influence forest area development in EKCd studies. In contrast, only variables from six thematic categories were statistically significant in FTH studies. While the drivers in EKCd studies are very diverse, with a slight dominance of income, demography and trade, FTH studies are clearly dominated by demographic and direct-area-related drivers, which account for 33% and 37% of significant variables respectively (figure 5(c)).

The studies applied different statistical estimation methods. We extracted all estimation methods, which are in order of frequency of appearance: OLS, generalized mixed methods (GMM), generalized least squared (GLS), PMG estimation; spatial models (including spatial durbin and spatial panel models). Additionally, and building on that, we analyzed which studies prefer FEs and which prefer random effects (RE) based on respective specification tests (e.g. Hausmann test).

The estimation methods chosen by the study authors also vary with the timing of publication. Historically, OLS has been frequently used in the early studies while in more recent publications, the range of estimation methods has become more heterogeneous (see also Tandetzki et al (2022)). To minimize a potential bias and ensure better comparability, we separately investigated the influence of the chosen estimation method on the drivers. Across all studies OLS was used most frequently (n = 21), followed by GMM (n = 13) and GLS (n = 7). Under these three estimation methods a high diversity of drivers (from 10 to 12 thematic categories) was identified. The relative frequency distribution of influential drivers indicates that estimations with OLS, GMM, and GLS similarly identified income and demography as main drivers. Direct area-related drivers were, however, identified as influential in 16% and 13% of studies using OLS and GMM respectively, and only in 3% of GLS studies. Conversely, GLS has a higher proportion of significantly tested other factors and institutional economic factors (nine and 10 times, respectively). Estimation methods used less often (PMG and spatial models) show a lower variety of drivers (figure 5(d)). Drawing definitive conclusions proves challenging in this context due to the limited strength of the mean value.

In almost all articles that conduct comparative studies between FEs and REs, the FE method is preferred. In 32% of the selected studies, both FE and RE were considered, with a decision in favor of FE in 91% of the studies.

In order to investigate potential variations resulting from variable specification within a thematic group, we delved into the analysis of the income and demographic variables most commonly assessed (as depicted in figure 3(b)). To assess the impact of income, study authors most frequently applied GDP pc, GDP pc² and GDP growth rate. Figure 6 shows the application of the different specifications of income in studies with different scope. Income variables are predominantly applied in EKCd studies (see figure 6(a)), but turned out to be significant only in 44%–66% of cases. In the FTH studies GDP pc and GDP pc² were statistical significant in four and two of 12 studies but income was rarely tested over all. In the baseline studies, where only income was tested, it was significant for all income variables over at least 78%. In these groups (figures 6(a) and (b)), there are no great differences between the individual variable specifications. A different picture emerges when we differentiate regionally (figure 6(c)), where in large-scale studies (global and developing), GDP pc growth never showed a significant impact, and GDP pc, and GDP pc² were less often statistically significant (0%–60%) than in studies considering smaller regions (71%–100% in continental studies). Significant effects of income were most frequently found in studies of Latin America.

Zoom In Zoom Out Reset image size

The influence of demographics was mainly tested using the variables of population density, population annual growth rate und rural population. These variables were tested in at least every FTH study, but only in 79% of the EKCd studies. In studies of both concepts, population density and annual growth were statistically significant in at least 53%–100% of cases, while rural population was significant in a maximum of 27% of cases (figure 7(a)).

Zoom In Zoom Out Reset image size

In global studies, a statistically significant impact of population density and population annual growth rate was found in 71% of the cases, whereas the impact of rural population density was only confirmed in 27% of the cases. In developing countries only, the demographic variable population density was tested as statistically significant (75% of cases). Again, consideration of continental studies provides a different picture. Rural population consistently exhibits statistical significance when tested in the regions of Asia, Africa, and Latin America, whereas its statistical impact in global studies is observed in only 27% of cases. Population density did not show any significant impact in any study conducted on Asia and Latin America. In Africa, all three demographic variables were consistently tested with statistical significance impact (figure 7(b)).

To hold deforestation is one of the main goals to combat global climate change and foster forest-based mitigation measures as well as e.g. loss of biodiversity. Both the EKCd hypothesis and the FTH try to explain global forest area dynamics as a function of socio-economic variables. Knowing which of these variables significantly drives forest area development could enable policy and stakeholders to effectively intervene and reverse deforestation. However, overarching key drivers confirmed by our research such as population growth and GDP pc are difficult to shape for individual policy-makers or other decision holders in the short run. The statistical relevance of income, especially on regional scale in the southern hemisphere, confirms the important role of poverty alleviation, reduction of inequalities on the global agenda (United Nations 2015). Our analysis further revealed that the international exchange of goods and financial assets significantly impact forest area development. The governance of related economic variables, e.g. international trade or foreign direct investments appears to be possible but requires, due to its cross-border nature, the mutual effort of politics and economic actors in both the originating and receiving regions. According to our results, institutional indicators are remarkably often tested as statistically significant and thus, seemed to be the promising tools for controlling forest area development at the country or regional level. Thus, policymakers looking for effective ways to influence forest area development in a given region should focus on strengthening institutional drivers including, e.g. fostering political stability, civil freedom, political rights and democracy, while combating corruption. However, in a globalized world, worldwide deforestation cannot be controlled through the efforts of specific countries alone. Thus, only the simultaneous addressing of local and transregional drivers, such as trade and economic drivers, can truly alter forest area development. Again, noteworthy is the variable category education which showed to be significant in almost all cases when included into estimations in particular in global studies. Thus, to invest in and enable access to education is concrete and tangible option for action to be pursued from regional to global scale to fight deforestation. This systematic review is not a meta-analysis or meta-regression. While we identified which drivers are statistically significant in influencing forest area and provide a synthesis of patterns within these drivers, we do not conduct a meta-analysis of the strength and direction of effect of each influencing variable or a re-estimation of forest development based on the identified variables.

This synthesis built on the established methodology for systematic reviews of RoSES guidelines and thus, followed a clear protocol (Haddaway et al 2018). A two-stage approach that involves an initial EGM followed by a systematic review is not a novel but rarely employed method outside medical research. In this context, the EGM serves as a protocol, given its significant resemblance to the protocol of a systematic literature review. Initially, the breadth of evidence on a research topic is outlined using the EGM, which maps the existence and non-existence of literature on a certain topic. Subsequently, the systematic review approach delves deeper into the approaches and results applied in the literature and synthesizes them. This two-stage method allows for a more comprehensive and precise examination of the research (James et al 2016). Since the selection strategy is clearly defined, comprehensible, and repeatable, the reader can retrace how, at what time, and with what criteria the literature search was conducted. However, an unknown publication bias is inherent to this kind of methodology: systematic reviews can only make use of published research that fulfills the eligibility criteria (e.g. English and peer-reviewed). Anyway, this is only a subset of all work conducted in the field since, e.g. researchers do not publish their insignificant findings or publish their results in another language or format (e.g. non-reviewed working paper, book). In addition, also English and peer-reviewed articles face this reporting bias through individual editors' and reviewers' choices (see e.g. Choumert et al (2013) on a discussion of selection bias in meta-analysis).

Due to the heterogeneity of concepts and study designs (e.g. geographical scope, estimation method, measurement of model fit) used for the different estimations, comparison of study contents was challenging. In order to be able to compare the studies at all, we focused on the independent variables used in each study and related their occurrence and statistical significance to the study region, the dependent variables, the forest area development concept, and the estimation methods.

For a straight comparability of the studies, it would be necessary to re-estimate some of the studies and align data and variable sets with similar statistical indicators and estimation methods. However, this is beyond the scope of this theoretical systematic review.

Over time, estimation methods have seen ongoing enhancements, enabling more recent publications to construct more dependable models. Future work may explore whether the earlier estimation methods merit a more critical scrutiny. For instance, spatial analysis techniques were used only in two studies, revealing a distinct pattern of forest development drivers compared to the other studies.

Bi- and multilateral trade linkages can drive deforestation (see e.g. Pendrill et al 2019, 2022), which is often neglected in cross-country studies that make up the majority of studies identified in this review. Future studies could also focus on these aspects.

Additionally, research gaps arise from the separate examination of different forest area types (e.g. natural forest, planted forest and plantations (FAO 2020)). Also, a more comprehensive investigation of institutional and economic (e.g. debt level), education, consumption (e.g. fossil fuel energy consumption) and production forest land-related (e.g. labor per unit forest area land) drivers is needed. These drivers, although relatively rarely tested, have a statistically significant influence on forest area development.

Even though other land use theories explaining land use change especially for smaller regions exist (Meyfroidt et al 2018), the hypotheses of EKCd and FTH provide an overarching framework that is applicable from regional to global studies and whose statements are universally valid independent of the study region or design. Based on our synthesis we concluded that the investigation of EKCd and FTH is not yet completed. In line with previous research on the validity of the concepts we confirm that—based on the present findings—neither a rejection nor an acceptation of the hypotheses is possible. In summary, our findings showed that many drivers influence the development of forest areas. Thus, future research dedicated to the hypothesis of the EKCd is needed to amend and refine the mere concept of an U-shaped curve. The FTH features a more comprehensive theoretical explanation for forest area development but has been insufficiently applied on a global scale till now.