Although sustainability-related efforts remain central to development, their accomplishment varies across places for a variety of reasons including climatic and geographic differences. This variability makes a regional focus important. In this paper, we investigate ecological footprints in both total and sub-footprint forms as measures of environmental sustainability over time in Africa. We examine economic, demographic, and ecological variables as key factors driving national-level environmental sustainability in Africa over nearly five decades. Our results reveal demographic attributes to be the primary but not the only forces affecting environmental sustainability. We situate our findings both in the context of prior studies and in relation to opportunities for further academic study.

From the Millennium Development Goals of 2000–15 to the Sustainable Development Goals since 2015, prominent international organizations including the Food and Agriculture Organization of the UN (FAO), United Nations, United Nations Development Programme (UNDP), and the World Bank continue to emphasize the central importance of the environment for development efforts (FAO 2016; UNDP 2014; World Bank 2014, 2015). Ensuring environmental sustainability has been on this global agenda since at least 2000 as one of the Millennium Development Goals. Although as of 2015 these goals have been reformulated with an eye to transforming development, many of the Sustainable Development Goals retain an environmental focus, including providing access to clean water and sanitation, maintaining healthy ecosystems, addressing climate change, and improving other aspects of well-being (UNDP 2014). Further, questions regarding how to examine spatial and temporal variability in the environment–development link, particularly as they are affected by climate change, remain core objectives of academic research. Our research contributes to this scholarship by examining a measure of environmental sustainability over time across five geographic regions in Africa.

Research on environmental sustainability seeks to answer the salient question of how much pressure nations exert on their surrounding environment through industrial processes linked with development and consumptive activities related to human populations. Research demonstrates that ecological footprints (EF)—a widely used measure of environmental sustainability—are driven by numerous economic, political, ecological, and demographic factors at varying—national, regional, and global—scales. Identifying these drivers, however, may require slight modification when applied to certain developing regions, given historical and ecological factors and pressing environmental challenges related to climate change. To date, limited attention has been paid to how these drivers operate in a single region over time in the context of regional heterogeneity (Dietz and Jorgenson 2014). Accordingly, this paper examines the driving forces of environmental sustainability over time in Africa, with emphasis on five geographic regions. A central objective is investigation of whether and how these dynamics play out differently across the six components of the EF and the extent to which it is possible to identify varying regional implications.

A data set of 46 African countries (grouped into five geographic regions) for 48 years (1,871 case-years) is used in empirical models integrating theoretical expectations from prior work. Accounting for regional variation in Africa is important, given its many ecosystem types as well as different cultural and historical experiences. Not only is there variability with regard to colonial history, in terms of colonizing power and year of independence, but also development has been uneven with regard to where economic growth has been concentrated and where different development projects were initiated as a result. Ecological conditions vary across the continent, as shown by the land-based resources available. The unique environmental challenges continent-wide related to sustainability efforts are important factors for this regional investigation, and thus for the coupled human and natural systems investigated in this research. Our results demonstrate the importance of demographic, economic, and environmental factors over time in these five regions, with some variation in the direction of factors both across regions and across the EF's subcomponents.

EXPLAINING ENVIRONMENTAL SUSTAINABILITY

Developing a better understanding of human actions and environmental changes is an essential task for academic research given an increasingly interconnected world and how these dynamics have become central features in the work of international organizations, including the FAO, UN, UNDP, and World Bank. Describing large-scale processes spanning human and natural systems requires an approach that is able to articulate how social structure and institutions, and the physical environment, are mutually reinforcing (Dietz and Jorgenson 2013; Liu, Dietz, Carpenter, Alberti, et al. 2007; Liu, Dietz, Carpenter, Folke, et al. 2007). Insights from environmental sociologists are especially well-placed as this work investigates how human activities, social institutions, and relations between social units at numerous scales—community, national, regional, and international, for instance—contribute to environmental impacts like those portrayed in a measure of environmental sustainability such as the EF.

EF is a widely used measure of sustainability because it is a standardized measure of consumptive impact with both national relevance and international comparability. The EF measure also encompasses multiple use functions of the environment—living space for human populations, source of natural resources, and place to absorb society's waste—and resulting domains of environmental impact, rather than a single one that might not convey the full imprint a society has on its physical surroundings. The total EF measure has six components: the cropland, grazing land, forest, fishing, built-up land, and carbon footprints (Ewing et al. 2009). Represented as a single metric in units of land area, which further attests to its utility and comparability, EF data are available for more than 160 countries and 53 time points (http://data.footprintnetwork.org/). Scholarly work is thus able to consider an EF as a consumption-based, national environmental impact that can be applied at multiple scales, including nations, regions, and the world; thus, it is a major indicator of societal impact on natural environments and, by extension, an appropriate measure of sustainability.1 

To determine how much pressure nations exert on their surrounding environment, social scientific scholarship examines economic, demographic, and ecological factors as primary drivers of change in environmental conditions. The conceptual models of Duncan (1959), Ehrlich and Holdren (1971), and York, Rosa, and Dietz (2003a, 2003b) provide necessary background, revealing emphasis from human ecology. To begin, the four shared pillars of these approaches—population, (social) organization, environment, and technology—form the core of Duncan's (1959) POET model. Each domain is posited to play a role in social and environmental change, including the ways that human societies and the physical environment are involved in reciprocal and/or interlinked relations. Ehrlich and Holdren (1971) developed the IPAT model from a similar background. According to this approach, as populations (P) grow they interact with changing levels of affluence (A) and technology (T) to yield an environmental impact (I). The third model, the STIRPAT2 model (York, Rosa, and Dietz 2003a, 2003b), has strong ties with both ecological models and uses the same dimensions—population, affluence, and technology as drivers of impact—recast in stochastic form.

Population plays a central role in these models due to the number of population measures available and the multiple dimensions they represent. In short, these measures reveal how different population attributes contribute to the shifting demands human populations place on the physical environment (Jorgenson, Rice, and Crowe 2005; Marquart-Pyatt 2010, 2015; York, Rosa, and Dietz 2003a, 2003b). Total population size is inherently related to a nation's total resource consumption, and population growth suggests continually increasing demand on natural capital or total environmental resources (Dietz, Rosa, and York 2007; Rosa, York, and Dietz 2004; York, Rosa, and Dietz 2003a, 2003b).

A nation's population composition has been used to understand resource utilization and, by extension, a nation's environmental footprint. For instance, a country with a higher proportion of the population that is of working age relative to that above and below the working age, has been found to increase resource consumption at the national level (Dietz, Rosa, and York 2007; Marquart-Pyatt 2015; Rosa and Dietz 2012; York, Rosa, and Dietz 2003a, 2003b). It is posited that the working-age population not only uses more resources, but also contributes to increased economic production (Crenshaw and Ameen 1997; Crenshaw and Robison 2010; Rosa and Dietz 2012). It is also possible that modernization, urbanization, and industrialization (all related to economic growth) are important drivers of the age composition of a population, in keeping with classic demographic transition theory (Crenshaw, Christenson, and Oakey 2000). The expectation is that as a country becomes more developed, both mortality rates (especially among children) and birth rates will decline, leading to a proportionally larger working-age population that is having fewer children.

Of the remaining two elements in the STIRPAT model, affluence is almost exclusively measured by GDP per capita, but the technology component is much more difficult to capture (York, Rosa, and Dietz 2003b). The portion of the population that lives in an urban area is often used as a measure of the country's modernization (Ehrhardt-Martinez, Crenshaw, and Jenkins 2002; York, Rosa, and Dietz 2003a, 2003b), which is expected to be correlated with technological sophistication. Measures of the relative importance of manufacturing and/or the service sector have also been used to control for differences in domestic economic structure that are associated with different stages of technological modernization and affluence (Dietz, Rosa, and York 2007; Jorgenson and Burns 2007b; Jorgenson and Clark 2011; Marquart-Pyatt 2010; York, Rosa, and Dietz 2003a, 2003b). Larger urban populations have larger environmental impacts (Jorgenson 2003, 2005; Jorgenson and Clark 2011; Jorgenson, Rice, and Crowe 2005; Rice 2007; York, Rosa, and Dietz 2003a, 2003b). However, the effects of urbanization may be curvilinear and start to decrease after a certain point (Ehrhardt-Martinez, Crenshaw, and Jenkins 2002).

Modernization, industrialization, and an increase in the relative size of the service sector may reduce certain kinds of environmental impacts, such as deforestation, within a country (Ehrhardt-Martinez, Crenshaw, and Jenkins 2002; Jorgenson and Burns 2007a), but increase others, such as CO2 emissions (York, Rosa, and Dietz 2003b). Shifting resource extraction and manufacturing to other countries will not reduce the calculated size of a country's EF, because the EF measures the consumption of a country rather than its production. Therefore, the combined effect of affluence and technology on the footprint measure is expected to be positive.

The EF, as a composite measure of environmental sustainability, has six components encompassing agricultural and other development-related measures, including cropland, grazing land, forest, fishing, built-up land, and carbon footprints. Few studies have investigated the factors shaping the sub-footprints on their own. Rosa, York, and Dietz (2004) found uniform effects of population and affluence on all five of the EF components they studied (forests, arable land, grazing land, fishing ground, and built-up area). Two of the four studies that examined these components separately examined single years of data (2000 and 2005), finding that population and economic measures affect some but not all sub-footprint measures globally and in specific regions (Jorgenson, Rice, and Crowe 2005; Marquart-Pyatt 2010). Jorgenson, Rice, and Crowe (2005) found that urban population and world economy position affect cropland, forest, built-up land, and carbon footprints, that within-nation income inequality shapes forest and carbon footprints, and that the factors shaping grazing land and fishing footprints were counter to predictions. Marquart-Pyatt (2010) found that the factors significantly affecting the carbon footprint are the most similar to those significantly affecting the total EF with regard to economic, demographic, ecological, and political attributes, noting that there are links across equations for the footprint components, specifically cropland, forest, and built-up land. Considering the largest time span with 46 years of data, Marquart-Pyatt (2015) demonstrated that demographic attributes are key factors shaping environmental sustainability in five West African countries. Analyses extending this work across additional spatial and temporal scales—across more countries and more time points—contribute to this growing literature.

Given prior scholarly work, our empirical model includes demographic, economic, and ecological indicators as central factors shaping environmental sustainability. In addition, two important principles from prior work are guideposts for this research. First, the importance of context cannot be overstated (Dietz 2013), and the importance of regional context, such as on the effect of urban slums on child mortality (Jorgenson, Rice, and Clark 2012), should be investigated. Second, careful attention is warranted regarding the specific types of demographic measures used in an empirical study (Marquart-Pyatt 2013, 2015; York and Rosa 2012). Following these two principles necessarily refocuses discussions about contexts to encompass societal and institutional attributes including demographic, economic, and ecological factors. The regional emphasis in Africa, to which the paper now turns, is another novel element of this work.

THE AFRICAN CONTEXT

The African continent contains 54 sovereign countries spread over 29.4 million square kilometers and is home to 1.15 billion people, or 14.4% of the world's population. As of 2008, Africa as a whole had a footprint that was smaller than what its biological capacity could support (WWF 2012). The UN divides the continent into five regions: Eastern, Middle, Northern, Southern, and Western (see the  appendix for a detailed list of countries by region). For context, Table 1 provides regional characteristics of a suite of variables, some of which are in addition to the variables included in the analyses (excluding island countries, consistent with the models). From this table we see regional differences in geographic size: in square kilometers the Southern region is the smallest and the Northern region the largest, while average country size is smallest in the Western region and largest in the Southern region. The Eastern region is the most populous (as of 2014), while the Southern region is the least populous. Fourteen of the countries are former British colonies. The majority of countries in each region have a coastline, except in Northern Africa, where all do. We discuss differences in the footprints later in the paper.

TABLE 1.
Descriptives as Five-Year Averages by African Region (Excluding Island Countries), 2010–14
EasternMiddleNorthernSouthernWestern
Region area (million sq. km) 6.11 6.50 7.47 2.65 6.06 
Region population in 2014 (millions) 357.00 147.00 219.00 62.00 343.00 
Number of countries 14.00 8.00 6.00 5.00 15.00 
Number of coastal countries 6.00 6.00 6.00 2.00 12.00 
Number of former British colonies 8.00 0.00 2.00 4.00 4.00 
Average country size (million sq. km) 0.44 0.81 1.25 12.40 0.40 
GDP per capita 493.20 3,207.62 2,975.13 4,071.69 558.36 
Percent urban 28.49 48.33 58.03 42.16 42.80 
Age dependency ratio (%) 88.36 87.70 56.07 64.37 87.89 
Rural population density 0.92 0.11 0.20 0.25 0.42 
Arable land per capita 17.74 4.20 8.47 6.34 19.30 
Land area per capita 2.34 7.74 7.22 13.54 3.85 
Precipitation (cm/y) 84.94 143.19 16.65 55.44 112.21 
Forest area (% of land area) 23.84 51.27 6.29 14.13 27.45 
Avg. quality of port infrastructure 3.49 2.86 3.84 4.21 3.91 
Avg. % of pop. with access to electricity 21.81 38.55 88.24 47.02 32.78 
 rural 6.88 14.24 85.41 29.34 14.51 
 urban 58.34 58.76 93.27 76.59 60.27 
Avg. % with access to improved sanitation 31.10 35.73 77.89 49.31 23.90 
 rural 24.12 24.79 70.44 39.56 14.27 
 urban 41.60 49.26 84.78 60.37 37.68 
Avg. % of pop. with access to improved water 66.36 63.34 84.07 86.22 72.04 
 rural 57.63 43.00 77.18 79.39 61.65 
 urban 86.00 84.25 90.06 96.90 88.17 
Total EF per capita (2007–08 avg.) 1.04 1.22 1.90 1.97 1.57 
Cropland EF per capita (2008) 0.31 0.38 0.61 0.37 0.57 
Grazing EF per capita (2008) 0.21 0.27 0.31 0.67 0.33 
Carbon EF per capita (2008) 0.09 0.06 0.73 0.66 0.14 
Built EF per capita (2008) 0.04 0.05 0.05 0.04 0.06 
Forest EF per capita (2008) 0.35 0.40 0.15 0.20 0.34 
Fishing EF per capita (2008) 0.04 0.07 0.04 0.03 0.13 
EasternMiddleNorthernSouthernWestern
Region area (million sq. km) 6.11 6.50 7.47 2.65 6.06 
Region population in 2014 (millions) 357.00 147.00 219.00 62.00 343.00 
Number of countries 14.00 8.00 6.00 5.00 15.00 
Number of coastal countries 6.00 6.00 6.00 2.00 12.00 
Number of former British colonies 8.00 0.00 2.00 4.00 4.00 
Average country size (million sq. km) 0.44 0.81 1.25 12.40 0.40 
GDP per capita 493.20 3,207.62 2,975.13 4,071.69 558.36 
Percent urban 28.49 48.33 58.03 42.16 42.80 
Age dependency ratio (%) 88.36 87.70 56.07 64.37 87.89 
Rural population density 0.92 0.11 0.20 0.25 0.42 
Arable land per capita 17.74 4.20 8.47 6.34 19.30 
Land area per capita 2.34 7.74 7.22 13.54 3.85 
Precipitation (cm/y) 84.94 143.19 16.65 55.44 112.21 
Forest area (% of land area) 23.84 51.27 6.29 14.13 27.45 
Avg. quality of port infrastructure 3.49 2.86 3.84 4.21 3.91 
Avg. % of pop. with access to electricity 21.81 38.55 88.24 47.02 32.78 
 rural 6.88 14.24 85.41 29.34 14.51 
 urban 58.34 58.76 93.27 76.59 60.27 
Avg. % with access to improved sanitation 31.10 35.73 77.89 49.31 23.90 
 rural 24.12 24.79 70.44 39.56 14.27 
 urban 41.60 49.26 84.78 60.37 37.68 
Avg. % of pop. with access to improved water 66.36 63.34 84.07 86.22 72.04 
 rural 57.63 43.00 77.18 79.39 61.65 
 urban 86.00 84.25 90.06 96.90 88.17 
Total EF per capita (2007–08 avg.) 1.04 1.22 1.90 1.97 1.57 
Cropland EF per capita (2008) 0.31 0.38 0.61 0.37 0.57 
Grazing EF per capita (2008) 0.21 0.27 0.31 0.67 0.33 
Carbon EF per capita (2008) 0.09 0.06 0.73 0.66 0.14 
Built EF per capita (2008) 0.04 0.05 0.05 0.04 0.06 
Forest EF per capita (2008) 0.35 0.40 0.15 0.20 0.34 
Fishing EF per capita (2008) 0.04 0.07 0.04 0.03 0.13 

While these regional divisions do not perfectly capture all the environmental or social differences across the African continent, they control for differences between regions that may influence environmental sustainability and that are not possible to explicitly include in our models due to data availability limitations. For example, Table 1 shows that while the Southern region has the fewest countries (five, of which South Africa is the largest), it is not as different from the rest of the continent as the Northern Region for most of the measures shown. Northern Africa as a region is the most developed, given the highest percentages of the total, urban, and rural populations with access to electricity and improved sanitation averaged across the six nations, the second-highest total EF per capita, and the third-highest GDP per capita. Northern Africa also has the lowest age dependency ratio of all the regions, pointing to a regional demographic distinctiveness. The Southern Region is unique for its small number of countries (five) and their variability in size (two very small countries, Lesotho and Swaziland, and a very large country, South Africa).

DATA AND METHODS

In this analysis we examine predictors of total EF and the subcomponent footprints on their own. The first set of models investigates the footprints for the continent of Africa; the second set of models breaks down the total EF by region. To accomplish this we employ a pooled cross-section time-series regression with data from 46 African countries over 48 years (1961 to 2008).3 The unit of analysis is the country-year, of which, after dropping cases with missing data, there are 1,871.4 To accommodate the presence of serial correlation we use a fixed effects Prais-Winsten model with panel-corrected standard errors (Beck and Katz 1995, 1996, 2004), which allows for heteroskedastic disturbances that are contemporaneously correlated across panels. With this model we specified first-order autocorrelation (AR1) within panels and used the pairwise option given unbalanced panels. All variables are log-transformed; we used Stata 14 to run the analysis.

Outcome Variables

The outcome variables included in this analysis are the total EF and its six components at the country level. As calculated by the Global Footprint Network (http://www.footprintnetwork.org/), EF is the land area required to support the consumption of the people living in a particular country (Ewing et al. 2009). The EF unit is global hectares—an averagely productive hectare based on world productivity. EFs are useful measures of sustainability because they are standardized measures of consumptive impact on the world at large. Due to international trade, the productive impact of a country would not necessarily be expected to reflect its consumptive impact. The consumptive impact must be measured against the productive capacity of the world to determine overall sustainability. The comparison between countries, however, provides useful insights into the portion of world resources used by each country. We examine the factors that drive this consumption, both in total and for each component.

The total EF measure is composed of six elements: the cropland, grazing land, forest, fishing, built-up land, and carbon footprints. The cropland footprint is the amount of productive land needed to produce food for human consumption either directly or indirectly through animal feed, as well as fibers and other materials. The grazing land footprint is the area needed to raise the livestock for the meat, milk, fiber or leather consumed, determined as the space required to feed animals above what is provided by available grain feed. The forest footprint is the area of forest harvested to make the wood and paper products used. The fishing footprint is the area need to provide the amount of seafood consumed. The built-up land footprint is the area (adjusted by productivity) that is occupied by human-built structures, including roads, buildings, and hydropower dams. The carbon footprint is the area of forest needed to absorb the human-released carbon emissions produced by the country and the products it consumes that are produced elsewhere.

We use the total country footprint rather than the per capita footprint in our analysis because a country's global sustainability (i.e. its use of world resources compared to world productivity) is a function of per person consumption and the size of the population, in keeping with using country as the unit of analysis. Using the country footprint also allows us to include the effect of total population in our model (in keeping with STIRPAT) and look for differences in effects of population size. Total population can only be included as a predictor in models using total EF as the outcome, rather than a per capital footprint measure. Total population is a logical control variable for total EF (Dietz, Rosa, and York 2007; Marquart-Pyatt 2015; Rosa, York, and Dietz 2004; York, Rosa, and Dietz 2003a, 2003b). However, for ease of comparison with other studies, we use per capita footprint when discussing EF in descriptive terms, as it is more intuitive to interpret, since it accounts for population size.

Independent Variables

The input variables in the models include measures of affluence, population, history, and environmental context. All values are from the World Bank's DataBank (http://databank.worldbank.org/data/home.aspx) unless otherwise noted. Gross domestic product (GDP) per capita in constant 2005 US dollars is included to measure economic development, an important influence on environmental impacts (Jorgenson 2005, 2009; Jorgenson and Burns 2007a, 2007b; Jorgenson and Rice 2005; York, Rosa, and Dietz 2003a, 2003b). Demographic factors have also been shown to influence EF and are included in our models. We include four demographic measures: total population, urban population, rural population density, and age dependency ratio. Urbanization is measured as the percentage of the total population of a country living in urban areas. More urbanized countries have larger environmental impacts (Jorgenson 2003, 2005; Rice 2007; York, Rosa, and Dietz 2003a, 2003b). The rural population density is measured as the population of the country that lives in a rural area divided by the total land area of the country. The age dependency ratio (ADR) is the ratio of individuals younger than 15 and older than 64 (dependents) to those between the ages of 15 and 64, expressed as a percentage. Whether a country has a British colonial history was coded as a dummy variable using data from the CIA World Factbook. We also include three land-based variables: the total land in the country per person and the arable land in the country per person, measured in hectares, as well as whether the country has a coastline (determined by examining Google Maps).

In the first set of models, we also include regional fixed effects (i.e. dummy variables for region), using West Africa as the reference region, given prior work (Marquart-Pyatt 2015) and the characteristics shown in Table 1. When we examine the measures that make up the EF individually, a different picture emerges than when we investigate them as an aggregate measure (Marquart-Pyatt 2010, 2015). Teasing out these relations at the continental or regional level over time is an important extension of prior work. The focus on African nations is unique and salient as well, providing a baseline for proposed future work, as articulated in the discussion and conclusion that follow.

RESULTS

Figure 1 presents the per capita EF measures for 11 time points (roughly every five years) over the 48-year time span from 1961 to 2008 for all five regions in Africa, and includes values for Africa as a whole and the world as reference points. Figure 2 shows the cropland, grazing land, forest land, and carbon footprints for the same 11 time points from 1961 to 2008 for all five regions. As shown in Figure 1, the per capita footprint for the African continent as a whole is much lower than for the whole world, nearly half a hectare per person smaller. Regional variation is also evident, especially for Southern Africa, which until the early 1980s had a per capita footprint higher than the world average. Footprint size for the Southern region has fallen to be much closer to that of the rest of Africa, though it remains the region with the highest footprint. Middle and Eastern Africa have smaller per capita footprints than Africa as a whole, while Western Africa has had an average or slightly higher footprint. Northern Africa is the only region that has had an increase in its per capita footprint over the time period shown, with all the other regions and Africa as a whole experiencing net decreases in total EF.

FIGURE 1.

Per Capita Total Ecological Footprint (in Global Hectares) for Africa, African Regions, and the World, 1961–2008

FIGURE 1.

Per Capita Total Ecological Footprint (in Global Hectares) for Africa, African Regions, and the World, 1961–2008

FIGURE 2.

Average Per Capita Footprint (in Global Hectares) for African Regions, 1961–2008 (Island Countries Not Included)

FIGURE 2.

Average Per Capita Footprint (in Global Hectares) for African Regions, 1961–2008 (Island Countries Not Included)

Figure 2 breaks down the total footprint and shows four of the sub-footprints for each region as per capita measures. The built-up land and fishing ground footprints are very similar across all the regions and have changed very little over the study period; thus, we have not included them in Figure 2. These graphs show the different footprint components and reveal which drive the total footprint trends. We can see that Northern Africa's total footprint increase is due to a large increase in its per capita carbon footprint. The carbon footprint has increased in all the regions over the study period. It increased earliest in Southern Africa but has not increased much recently in this region. The forest footprint has been steady or slightly decreasing in all the regions except for Southern Africa, where the peak in the 1970s followed by decline since is similar to the carbon footprint in the region. In all the regions the cropland and grazing land footprints have shrunk, and in Eastern and Middle Africa this decrease offsets the increased carbon footprint.

Table 2 presents the descriptive statistics for the variables used in the analysis. Table 3 gives the results of the panel regression models for the total footprint and all six sub-footprints for all the countries included in the analysis. Table 4 shows the results of the models for total EF by region. Because we have logged all variables (as appropriate), the model results are in the form of elasticities that indicate the percentage change between the independent and outcome variables, holding all other variables constant. As described more fully by York, Rosa, and Dietz (2003b), coefficients equal to 1 indicate a proportional relationship (a unit elastic relationship), coefficients between 0 and 1 indicate a less-than-proportional relationship (inelastic relationship), and coefficients larger than 1 indicate a greater-than-proportional relationship (elastic relationship). Negative coefficients indicate an inverse relationship, with the same types of elasticities. The independent dummy variables are the multiplier of the dependent variable when the dummy variable equals 1, once the antilog (ex) of the coefficient is taken (York, Rosa, and Dietz 2003b).

TABLE 2.
Descriptive Statistics for 46 African Countries, 1961–2008
MeanStandard deviationMinimumMaximum
Total ecological footprint 16.066 1.238 12.745 19.212 
Cropland footprint 15.077 1.329 11.551 18.698 
Grazing land footprint 14.007 1.559 0.000 17.157 
Carbon uptake footprint 12.616 1.420 8.928 16.470 
Built land footprint 14.176 2.556 0.000 17.542 
Forest footprint 11.549 3.205 0.000 16.531 
Fishing footprint 11.596 4.574 0.000 18.165 
GDP per capita 6.477 0.940 4.228 9.515 
Total population 15.619 1.281 12.304 18.834 
Percentage urban 3.220 0.664 0.749 4.441 
Age dependency ratio 4.497 0.134 3.823 4.728 
Rural population density 0.230 0.254 0.002 1.407 
Former British colony 0.367 0.482 0.000 1.000 
Land area per capita 1.797 1.042 0.088 4.767 
Arable land per capita 0.298 0.180 0.000 1.456 
Coastal 0.661 0.473 0.000 1.000 
MeanStandard deviationMinimumMaximum
Total ecological footprint 16.066 1.238 12.745 19.212 
Cropland footprint 15.077 1.329 11.551 18.698 
Grazing land footprint 14.007 1.559 0.000 17.157 
Carbon uptake footprint 12.616 1.420 8.928 16.470 
Built land footprint 14.176 2.556 0.000 17.542 
Forest footprint 11.549 3.205 0.000 16.531 
Fishing footprint 11.596 4.574 0.000 18.165 
GDP per capita 6.477 0.940 4.228 9.515 
Total population 15.619 1.281 12.304 18.834 
Percentage urban 3.220 0.664 0.749 4.441 
Age dependency ratio 4.497 0.134 3.823 4.728 
Rural population density 0.230 0.254 0.002 1.407 
Former British colony 0.367 0.482 0.000 1.000 
Land area per capita 1.797 1.042 0.088 4.767 
Arable land per capita 0.298 0.180 0.000 1.456 
Coastal 0.661 0.473 0.000 1.000 

Note: All variables (except dichotomous measures) logged; N = 1,871.

TABLE 3.
Prais-Winston Regression Coefficients for Total Ecological Footprint and All Sub-footprints in Africa (Excluding Island Countries, S. Sudan and Somalia), 1961−2008
Total EFCroplandGrazingBuilt landCarbonForestFishing
Constant 1.833** −1.153 −2.887 −1.881 0.147 4.710 −16.514* 
  (0.600) (0.686) (4.606) (1.222) (8.862) (10.614) (7.302) 
GDP per capita 0.104*** 0.191*** −0.018 0.069* −0.055 −0.302 0.822*** 
  (0.024) (0.020) (0.072) (0.033) (0.400) (0.386) (0.189) 
Total population 0.974*** 0.981*** 0.993*** 1.009*** 1.682*** 1.100** 1.096*** 
  (0.014) (0.016) (0.090) (0.020) (0.213) (0.416) (0.145) 
Percentage urban −0.177*** −0.321*** 0.013 −0.311*** 1.441** −0.397 0.457 
  (0.030) (0.037) (0.174) (0.049) (0.518) (0.566) (0.275) 
Age dependency ratio −0.252* 0.168 0.257 0.075 −4.113* −0.791 0.595 
  0.116 0.142 0.801 0.265 1.780 1.969 1.402 
Rural pop. density −0.188* 0.031 −0.684 −0.655*** 0.802 −1.005 1.855 
  (0.116) (0.142) (0.801) (0.265) (1.780) (1.969) (1.402) 
Former British colony 0.167*** 0.061 −0.014 0.159** 0.493 0.839 0.819 
  (0.034) (0.040) (0.609) (0.059) (0.658) (0.598) (0.457) 
Arable land per capita 0.318*** 0.824*** 0.094 −1.457*** 0.103 −0.088 −0.014 
  (0.074) (0.099) (0.397) (0.131) (1.392) (1.454) (1.260) 
Land area per capita 0.081** −0.022 0.464 −0.068 −0.406 −0.216 0.370 
  (0.026) (0.030) (0.429) (0.041) (0.540) (0.723) (0.256) 
Coastline −0.074 −0.092 −0.384 −0.143* −0.894* −0.274 1.616** 
  (0.041) (0.048) (0.398) (0.067) (0.435) (0.460) (0.470) 
East Africa −0.302*** −0.48*** −0.018 −0.453*** 0.519 −0.707 −1.052* 
  (0.043) (0.061) (0.375) (0.083) (0.671) (0.558) (0.506) 
Mid Africa −0.364*** −0.432*** −1.005* −0.355*** 0.550 0.199 −0.203 
  (0.035) (0.042) (0.458) (0.067) (0.949) (1.321) (0.391) 
North Africa −0.128** 0.069 −0.024 −0.380** −0.854 −0.625 −2.344*** 
  (0.049) (0.063) (0.290) (0.121) (1.390) (0.760) (0.668) 
South Africa −0.061 −0.518*** 0.696 −0.788*** 1.716 −2.001 −4.637*** 
  (0.092) (0.070) (1.735) (0.119) (1.152) (2.469) (1.027) 
                
R-squared 0.987 0.969 0.611 0.934 0.116 0.249 0.214 
rho 0.839 0.751 0.949 0.805 0.938 0.959 0.810 
Total EFCroplandGrazingBuilt landCarbonForestFishing
Constant 1.833** −1.153 −2.887 −1.881 0.147 4.710 −16.514* 
  (0.600) (0.686) (4.606) (1.222) (8.862) (10.614) (7.302) 
GDP per capita 0.104*** 0.191*** −0.018 0.069* −0.055 −0.302 0.822*** 
  (0.024) (0.020) (0.072) (0.033) (0.400) (0.386) (0.189) 
Total population 0.974*** 0.981*** 0.993*** 1.009*** 1.682*** 1.100** 1.096*** 
  (0.014) (0.016) (0.090) (0.020) (0.213) (0.416) (0.145) 
Percentage urban −0.177*** −0.321*** 0.013 −0.311*** 1.441** −0.397 0.457 
  (0.030) (0.037) (0.174) (0.049) (0.518) (0.566) (0.275) 
Age dependency ratio −0.252* 0.168 0.257 0.075 −4.113* −0.791 0.595 
  0.116 0.142 0.801 0.265 1.780 1.969 1.402 
Rural pop. density −0.188* 0.031 −0.684 −0.655*** 0.802 −1.005 1.855 
  (0.116) (0.142) (0.801) (0.265) (1.780) (1.969) (1.402) 
Former British colony 0.167*** 0.061 −0.014 0.159** 0.493 0.839 0.819 
  (0.034) (0.040) (0.609) (0.059) (0.658) (0.598) (0.457) 
Arable land per capita 0.318*** 0.824*** 0.094 −1.457*** 0.103 −0.088 −0.014 
  (0.074) (0.099) (0.397) (0.131) (1.392) (1.454) (1.260) 
Land area per capita 0.081** −0.022 0.464 −0.068 −0.406 −0.216 0.370 
  (0.026) (0.030) (0.429) (0.041) (0.540) (0.723) (0.256) 
Coastline −0.074 −0.092 −0.384 −0.143* −0.894* −0.274 1.616** 
  (0.041) (0.048) (0.398) (0.067) (0.435) (0.460) (0.470) 
East Africa −0.302*** −0.48*** −0.018 −0.453*** 0.519 −0.707 −1.052* 
  (0.043) (0.061) (0.375) (0.083) (0.671) (0.558) (0.506) 
Mid Africa −0.364*** −0.432*** −1.005* −0.355*** 0.550 0.199 −0.203 
  (0.035) (0.042) (0.458) (0.067) (0.949) (1.321) (0.391) 
North Africa −0.128** 0.069 −0.024 −0.380** −0.854 −0.625 −2.344*** 
  (0.049) (0.063) (0.290) (0.121) (1.390) (0.760) (0.668) 
South Africa −0.061 −0.518*** 0.696 −0.788*** 1.716 −2.001 −4.637*** 
  (0.092) (0.070) (1.735) (0.119) (1.152) (2.469) (1.027) 
                
R-squared 0.987 0.969 0.611 0.934 0.116 0.249 0.214 
rho 0.839 0.751 0.949 0.805 0.938 0.959 0.810 

*p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed)

Notes: Standard errors in parentheses; N = 1,871.

TABLE 4.
Prais-Winston Regression Models for Total Ecological Footprint in Africa (Excluding Island Countries, S. Sudan, and Somalia), by Region, 1961–2008
 EasternMiddleNorthernSouthernWestern
Constant 0.148 1.838 1.893 −2.139 −0.195 
  (1.243) (1.992) (1.603) (3.156) (1.533) 
GDP 0.261*** 0.159*** 0.193* 0.108 0.104** 
  (0.054) (0.045) (0.079) (0.105) (0.034) 
Total population 0.978*** 0.940*** 0.878*** 1.321*** 0.912*** 
  (0.040) (0.043) (0.048) (0.180) (0.037) 
Percentage urban −0.261*** −0.305** 0.058 −0.383** 0.017 
  (0.057) (0.092) (0.167) (0.137) (0.058) 
Age dependency ratio (%) −0.013 −0.047 −0.319* 0.677 0.166 
  (0.275) (0.351) (0.151) (0.368) (0.343) 
Rural pop. density −0.601*** −0.215 0.083 −7.197*** 0.570 
  (0.170) (1.823) (0.685) (0.902) (0.325) 
Former British colony −0.083 none 0.552*** −1.264 0.290*** 
  (0.067)   (0.094) (0.677) (0.078) 
Arable land per capita 1.132** 0.402 0.368 0.452 0.365** 
  (0.381) (0.242) (0.338) (0.654) (0.107) 
Land area per capita −0.213* −0.101 0.036 −0.704*** 0.286*** 
  (0.107) (0.154) (0.093) (0.126) (0.033) 
Coastline 0.011 −0.437*** none −1.696* −0.175 
  (0.044) (0.078)   (0.758) (0.099) 
            
rho 0.890 0.785 0.630 0.658 0.848 
Sample size 420 339 237 202 673 
 EasternMiddleNorthernSouthernWestern
Constant 0.148 1.838 1.893 −2.139 −0.195 
  (1.243) (1.992) (1.603) (3.156) (1.533) 
GDP 0.261*** 0.159*** 0.193* 0.108 0.104** 
  (0.054) (0.045) (0.079) (0.105) (0.034) 
Total population 0.978*** 0.940*** 0.878*** 1.321*** 0.912*** 
  (0.040) (0.043) (0.048) (0.180) (0.037) 
Percentage urban −0.261*** −0.305** 0.058 −0.383** 0.017 
  (0.057) (0.092) (0.167) (0.137) (0.058) 
Age dependency ratio (%) −0.013 −0.047 −0.319* 0.677 0.166 
  (0.275) (0.351) (0.151) (0.368) (0.343) 
Rural pop. density −0.601*** −0.215 0.083 −7.197*** 0.570 
  (0.170) (1.823) (0.685) (0.902) (0.325) 
Former British colony −0.083 none 0.552*** −1.264 0.290*** 
  (0.067)   (0.094) (0.677) (0.078) 
Arable land per capita 1.132** 0.402 0.368 0.452 0.365** 
  (0.381) (0.242) (0.338) (0.654) (0.107) 
Land area per capita −0.213* −0.101 0.036 −0.704*** 0.286*** 
  (0.107) (0.154) (0.093) (0.126) (0.033) 
Coastline 0.011 −0.437*** none −1.696* −0.175 
  (0.044) (0.078)   (0.758) (0.099) 
            
rho 0.890 0.785 0.630 0.658 0.848 
Sample size 420 339 237 202 673 

*p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed)

Note: Standard errors in parentheses.

Our model outputs represent “ecological elasticities” that indicate the relationship between the independent/impact variables and the environmental impacts measured by EFs (York, Rosa, and Dietz 2003b). The results of our analysis are directly comparable to a subset of past studies given decisions about model specification, years of data included in the study, and sample composition.

GDP per Capita

We use GDP per capita to measure affluence and economic development in our models. In our total footprint model for the continent (Table 3), we find that GDP per capita is inelastic, with a 1% increase in GDP per capita resulting in only a 0.1% increase in total footprint. For the total footprint by region (Table 4) we find a slightly larger inelastic, positive effect in the Eastern, Middle, and Northern regions, at 0.26%, 0.16%, and 0.19%, respectively. In the Western and Southern region models the effect is functionally the same as for the continent model, at about 0.1%, though it is not significant for the Southern region. For the sub-footprint models (Table 3), the significant effects of GDP per capita are positive and inelastic, with increases in the cropland, built land, and fishing footprints of 0.19%, 0.07%, and 0.82%, respectively, for a 1% increase in GDP per capita.

Our continent and region findings for the positive effect of GDP per capita on total EF are consistent with most prior work (Jorgenson 2009; Jorgenson and Clark 2011; Marquart-Pyatt 2010; Rice 2007; Rosa, York, and Dietz 2004; York, Rosa, and Dietz 2003a), but not all (Dietz, Rosa, and York 2007; Marquart-Pyatt 2015).5 Similarly, our significant positive effects of GDP per capita on three of the sub-footprints are generally consistent with prior studies (Marquart-Pyatt 2010, 2015; Rosa, York, and Dietz 2004).

Total Population

Total country population is in general functionally unit elastic, with a 1% increase in total population resulting in close to a 1% increase in the total footprint in the continent model (Table 3), the Eastern, Middle, and Western region total footprint models (Table 4), and the cropland, grazing land, built land, forest land, and fishing footprint models (Table 3). A 1% increase in total population is associated with only a 0.88% increase in the total footprint in the Northern region and with a 1.32% increase in the total footprint in the Southern region (Table 4). Total population has a notably elastic relationship with the carbon footprint (Table 3), with a 1% increase in total population being related to a 1.68% increase in the carbon footprint.

Findings with respect to measures of population are largely in line with previous research, where total population is a consistent driver of a country's total consumption/resource footprint, a finding that also holds for the region models. Our positive and significant findings on this variable are similar to prior work (York, Rosa, and Dietz 2003a, 2003b; Rosa, York, and Dietz 2004; Dietz, Rosa, and York 2007). Rosa, York, and Dietz (2004) and Marquart-Pyatt (2015) are the only other analyses to include total population in a sub-footprint model, and their results are consistent with ours.

Urban Population

The percentage of the population that lives in an urban area has a negative, inelastic relationship with the total footprint, with a 1% increase in urban percentage resulting in a 0.17% decrease in the total footprint (Table 3). In the region models for total footprint (Table 4), a 1% increase in urban percentage results in 0.26%, 0.31%, and 0.38% decreases in the total footprint for the Eastern, Middle, and Southern regions, respectively. For the sub-footprint models (Table 3) a 1% increase in urban percentage resulted in significant decreases in the cropland and built land footprints of 0.32% and 0.31%, respectively. Urban Percentage has a positive effect on the grazing, carbon, and fishing footprints, though only the effect on the carbon footprint is significant, with a 1% increase in urban percentage resulting in a 1.44% increase in the carbon footprint (Table 3).

The effect of urban percentage on total EF has mixed results in the literature. We find a significant, negative effect in the total footprint model and in three of the five regions. In contrast, York, Rosa, and Dietz (2003a) found positive effects of urban percentage on total EF. Other studies reveal mixed effects, from positive (Jorgenson and Clark 2011; Jorgenson, Rice, and Crowe 2005; Rice 2007) to negative (Marquart-Pyatt 2015) to non-significant (Dietz, Rosa, and York 2007; Marquart-Pyatt 2010).

More limited comparisons can be made between our results and those of previous analyses for the relationship between urban percentage and the sub-footprints. We found a negative and significant effect between urban percentage and the cropland and built land footprints, which is consistent with some work (Marquart-Pyatt 2015) but not all (Jorgenson, Rice, and Crowe 2005; Marquart-Pyatt 2010). For the carbon footprint we found urban percentage to have a positive effect that is consistent with York, Rosa, and Dietz (2003b) and Marquart-Pyatt (2015) and generally with Jorgenson, Rice, and Crowe (2005).

Age Dependency Ratio

ADR has a negative, inelastic effect on the total footprint at the continent level, with a 1% increase in ADR associated with a 0.25% reduction in the total footprint (Table 3). In the region models for total footprint (Table 4), ADR is only significant for the Northern region, with a decrease of 0.32% for a 1% increase in ADR. For the sub-footprints ADR has a positive, inelastic though not significant effect on all but the carbon and forest footprints. The effect of ADR on the forest and carbon footprints is negative, but only has a significant effect on the carbon footprint, where a 1% increase in ADR decreases the carbon footprint by 4.11%.

We found ADR to have a negative effect on total EF, a finding that holds in the Northern region. In contrast, York, Rosa, and Dietz (2003a) found a positive effect. Other work using similar indicators found no significant effects (Dietz, Rosa, and York 2007; Marquart-Pyatt 2015; York, Rosa, and Dietz 2003b); Marquart-Pyatt (2015) included ADR in models of sub-footprints and found a negative effect on the carbon footprint that is consistent with our findings.

Rural Population Density

Rural population density has a negative, inelastic relationship with total footprint for Africa as a whole, with a 1% increase in rural population density associated with a 0.19% decrease in total footprint. The effect of rural population density is significant for the Eastern and Southern regions, with a 1% increase in rural population density associated with a 0.6% decrease in the Eastern region and a 7.2% decrease in the Southern region. The effect is positive but not significant for Western Africa. For the sub-footprints, rural population density has a negative effect on the built land footprint, decreasing it by 0.66% for a 1% increase in rural population density.

We found rural population density to have a negative and significant effect on total EF, consistent with Marquart-Pyatt (2015). Results for the sub-footprints differ, as Marquart-Pyatt (2015) found rural population density to have a negative effect on the cropland and grazing land footprints (we find no significant effect) and a positive effect on the built land footprint, whereas we find a negative effect.

Former British Colony

Countries with previous British colonial influence overall have footprints that are 18% larger than countries that were not British colonies (e0.167 = 1.18). In the region models, the effect of being a former British colony is positive and significant for the Northern and Western regions, with former British colonies in those regions having total footprints that are 74% and 34% larger, respectively, than non-former British colonies. With some regional differences, being a former British colony, compared to not being one, increases the built land footprint by 17% on average, and has a positive though not significant effect on all the other sub-footprints, except the grazing footprint, where the effect is negative but very small. Our results approximate prior studies using integration into the world polity via world-system position (Jorgenson 2003, 2004; Marquart-Pyatt 2010; York, Rosa, and Dietz 2003a).

Arable Land per Capita

The amount of arable land per capita has a positive, inelastic relationship with the total footprint at the continent level, increasing the total footprint by 0.32% for a 1% increase in land area per capita (Table 3). In the region models (Table 4) the relationship is also positive but only significant for two regions—Eastern and Western—where a 1% increase in land area per capita increases the total footprint by 1.13% and 0.37%, respectively. Arable land per capita has a positive effect on the cropland footprint, increasing it by 0.82% for a 1% increase in per capita arable land. Arable land has a negative effect in the built land model, decreasing it by 1.46% for a 1% increase in arable land per capita.

We find arable land per capita to have a positive and significant effect on total EF, consistent with prior studies (Jorgenson and Clark 2011; Marquart-Pyatt 2015). For the sub-footprints, we find arable land per capita to have a positive effect on the cropland footprint and a negative effect on the built land footprint, which are both consistent with Marquart-Pyatt (2015).

Total Land per Capita

Total land area per capita, as a form of population density, has a positive and inelastic relationship with the total footprint for the African continent, as well as in the Western region, with a 1% increase in the land area per capita increasing the total footprint by 0.08% (Table 3) and 0.29% (Table 4), respectively. Land area per capita has a negative and inelastic relationship with total footprint in the Eastern and Southern regions (Table 4), with a 1% increase in the land area per capita decreasing the total footprint by 0.21% and 0.7%, respectively.

We also find total land per capita to have a positive and significant effect on total EF. York, Rosa, and Dietz (2003a) also found the relationship to be positive and significant overall, as did similar studies, such as Dietz, Rosa, and York (2007) and Marquart-Pyatt (2010). We do not find land area per capita to have a significant effect on any of the sub-footprints, which differs from prior work (Marquart-Pyatt 2010, 2015).

Coastline

A country's having a coastline had a negative but not significant effect on the total EF at the continent level (Table 3). The effect was negative and significant in the Middle and Southern regions, with coastal countries having total footprints that were 65% and 18%, respectively, the size of the footprints of countries without coasts in those regions (Table 4). Having a coastline is also associated with having built land footprints that are 87% the size and carbon footprints that are 41% the size of countries without coastlines (Table 3). Fishing footprints are 5 times as large in countries with coastlines than in those without (Table 3).

Coastlines have not been included in any of the similar types of analyses in the literature. Although climate-type dichotomous measures have been used (Dietz, Rosa, and York 2007; Jorgenson 2009; Jorgenson and Clark 2011), the absence of substantial climate variability within the African continent precludes its inclusion in our analysis.

Region Effects

We control for region in our models of the total footprint and sub-footprints at the continent level (Table 3). We find that the Eastern, Middle, and Northern African regions have total EF 74%, 69%, and 88% the size of Western Africa (the reference region), while the Southern region—despite its unique attributes, noted earlier—is not significantly smaller than the Western region. For the cropland footprint, countries in the Eastern, Middle, and Southern regions have footprints that are 62%, 65%, and 60% the size of the Western region, while countries in the Northern region have cropland footprints that are not significantly larger than the Western region. Only countries in the Middle region have grazing land footprints that are significantly different from the Western region, at 37% the size. Being in the Southern region compared to the Western region has a positive but not significant effect on the grazing footprint, while being in the Eastern, Middle, and Northern regions has a negative but not significant effect.

The Eastern, Middle, Northern, and Southern regions all have smaller built land footprints than the Western region, by 64%, 70%, 68%, and 45%, respectively. The Eastern, Middle, and Southern regions have carbon footprints that are not significantly smaller than the Western region, while the carbon footprint of the Northern region is not significantly larger. The Western region has a forest footprint that is not significantly larger than those of the Eastern, Northern, or Southern region and not significantly smaller than that of the Middle region. The Western region has the largest fishing footprint compared to the other regions, with the Eastern, Northern and Southern fishing footprints being 35%, 10%, and 1% the size of the Western region.

DISCUSSION

In concert with prior work (e.g. the STIRPAT model), we find that total population consistently increases environmental impact as measured by the EF and its subcomponents. The near-proportional (unit elastic) relationship between population and impact/consumption/footprint (except of carbon) is not surprising since total population is an element of the total EF of a country. Simply stated, one would not expect there to be any efficiency of consumption to be gained simply by having more people in a country. The other hallmark driver of environmental impact, affluence, measured as GDP per capita, has the expected positive effect in the models, although the inelastic effect of GDP per capita suggests in most cases a negligible effect compared to the other predictors in the models. Although GDP per capita is not shown to be as important a driver of EF as we might expect conceptually from STIRPAT, this may be the result of our sample of African nations all having relatively low per capita GDPs on a global scale. In general, the total EF model for the continent tells a story that matches prior work—nearly all the effects are significant, they are all in the expected direction, and most are quite inelastic, thus having a proportionally small effect on the total footprint.

The region models for total footprint (Table 4) provide interesting findings. Except for land area per capita, all of the significant effects are consistent in their direction across the region models and in the continent model. The effect of land per capita is positive in the continent and Western region models, but is negative in the Eastern and Southern models, indicating important differences in the role of population density across different parts of Africa. Rural population density has different-size effects across the Southern and Eastern regions and the African continent, suggesting that there is something different about rural living in these regions compared to the others. The effect of total population is also marginally smaller in the Northern region and larger in the Southern region than in the other regions or for the continent as a whole.

The sub-footprint models are also interesting since they indicate which factors drive particular facets of the total footprint and thus how those factors then drive the total footprint. In examining the sub-footprint models, we find that in many ways the “consumption” of built land and carbon are fundamentally different from the more direct and tangible consumption that the other four sub-footprints measure, as indicated by differences in significance and sign. We can also observe how the ecological context of a country can influence its consumption patterns, presumably through differential availability, ease of access, and/or cost of products that can be produced in-country versus those that must be imported.

For example, countries with more arable land per capita consume more products from cropland (i.e., arable land per capita has a positive effect on the size of the cropland footprint), while countries with a coast consume much more seafood than countries without a coast. This is an unsurprising relation given the cost and effort required to transport seafood to non-coastal areas. Having a coastline is also related to having less productive land covered with built infrastructure, as well as having a smaller carbon footprint, possibly due to population gathering in coastal cities and spill-over effects related to demographic centralization like higher housing density and less energy needed to transport goods over large land areas.

The built land and carbon footprints have notably different drivers from the other four footprints. GDP per capita, total population, and urban percentage all have significant effects on the build land footprint that are in keeping with those on the other non-carbon footprints. However, the built land model is the only sub-footprint model where rural population density has a significant effect, in this case reducing the built land footprint. In addition, arable land per capita has a sizable negative effect on the built land footprint, whereas it has a smaller but positive effect on the cropland footprint. In this case it is thought that, since arable land is of potentially higher value due to its higher biological productivity (Ewing et al. 2009), there is incentive to maintain it in a productive state and cover it as little as possible with built infrastructure, a different kind of consumption from that measured by the cropland footprint.

The carbon footprint is different not only in being significantly influenced by ADR (none of the other sub-footprints are) but also in other effects related to demographic measures like total population and urban percentage. The negative effect of ADR on the carbon footprint means that children and the elderly have smaller carbon footprints than working-age people do. ADR also has a negative effect on the carbon footprint that is larger than the positive effect of total population and urban percentage combined. However, larger ADRs lead to larger populations, thus reducing the potential “benefit” for the carbon footprint of having a relatively large dependent population.

Unlike the other sub-footprints, population and urban percentage have a greater-than-proportional effect on carbon emissions, which does not seem to be offset significantly by increasing affluence. The effect of urban percentage, being larger than in the other sub-footprints and positive, suggests that while urban living is generally related to more efficient consumption this is not the case for the production of carbon emissions. This could be an Africa-specific result, where increasing levels of development (but not of high-tech, energy-efficient development) not only increase carbon emissions but also reduce infant mortality rates and extend life expectancy, leading to a larger population, thus giving population a larger effect on the carbon footprint than expected. More investigation of the potential interactions between development and population characteristics could help clarify the potential trade-offs between ADR and total population regarding the carbon footprint.

CONCLUSION

Prior work on development that called for ensuring environmental sustainability even during social and economic institutional transformation motivated this research. While some progress has been made toward these goals, substantial work remains regarding how to improve the well-being of individuals worldwide while maintaining environmental quality. Of paramount importance for this study were questions of how to model spatial and temporal variability in the environment–development link. By focusing on a set of regions, this research contributes to an important global agenda through empirical investigation of key drivers of environmental sustainability over time in the African continent. The results demonstrate that, over time, environmental footprints for the continent and its regions are shaped by demographic, economic, and environmental factors, but remain consistent with the STIRPAT model. These driving forces do differ, however, across the EF's subcomponents as well as across the regions over time, revealing no definitive driver of smaller EFs. Specifically, we find important differences between the drivers of built land and carbon footprints and the other, more consumptive, footprint measures. We also find notable relations between geographic country characteristics and specific sub-footprint measures (e.g., between arable land per capita and cropland footprints, and between coastline and fishing footprints). Finally, we find some relations that may be Africa-specific—related to a combination of common patterns of demographic characteristics and technological sophistication—such as the large effect of ADR and total population on the carbon footprint.

Our analyses suggest at least three avenues for future work: (1) to compare this with another developing region, like Southeast Asia or Latin America, (2) to incorporate predictors like political institutions and integration in world society/polity into the model, and (3) to consider how this study of EFs relates with work on other topics like well-being, life expectancy, and food security, both comparatively within a region (i.e. between countries) and globally or between regions. The first recommendation addresses a core question of cross-national scholarship in sociology regarding how to construct a comparative study, particularly in terms of what the relevant parameters are for investigation. That is, how focused or broad should the comparison be regarding the appropriate scale as well as the number of countries that comprise the relevant sample? From a comparative standpoint, an intriguing future study would be one that builds a cross-regional component as a central dimension.

As noted above as a second recommendation, with an expanded set of cases, researchers should also consider including additional predictors that derive from explanatory frameworks seeking to describe global processes of diffusion. Such measures include a country's position in the overall world polity, like the INGO Network Country Score (Hughes et al. 2009), and state environmentalism or environmental treaty participation (Roberts, Parks, and Vásquez 2004). These variables may aid in developing models that examine how being involved in different networks of relations may lead to occupying different relative positions in the overall world polity, where even if countries have identical numbers of INGO memberships, they may be involved in different networks that add an additional layer for explaining regional comparisons. Environmental treaty participation, a related measure that could account for engagement with international efforts for environmental protection, might also be included in subsequent studies.

Regarding extending this study beyond environmental sustainability to food security, it is important to remember what the EF can and cannot tell us about quality of life and food security. For example, just how large “should” a footprint be? While a shrinking footprint can indicate increased efficiency of production and use of materials and resources, it seems plausible that a reduced footprint may indicate a reduction in living standards and could result from conditions such as economic recession, natural disasters, epidemics, conflict, and corruption. Thus a reduction in footprint should be considered in comparison with other countries in the region and the global average footprint per capita to see whether it is to be viewed as a positive or negative outcome. Identifying countries that have relatively high well-being or food security with relatively low EFs to learn what combination of factors is associated with that outcome, as Lamb et al. (2014) have done with life expectancy and carbon dioxide emissions, would be a way of trying to identify “target” EFs.

In short, it should be kept in mind that the EF is one measure of impact, and simply living creates an impact. Thresholds may also be important to consider, since what is considered a small footprint may not necessarily be a good thing for people's livelihoods and well-being beyond a certain point. Future work that moves in a quality of life and/or food security direction should consider whether there are key indicators that can be considered in conjunction with EF to tease out the implications of EF change over time. Given the complexity of these issues, which are essential for understanding society-environment-development links, further refinements to the correspondence between our theoretical expectations and empirical models accounting for variability in many forms—including temporally and spatially—remain an integral priority, a fruitful avenue for subsequent work, and ultimately a challenge for future research.

REFERENCES

REFERENCES
Beck, Nathaniel, and Jonathan N. Katz.
1995
.
“What to Do (and Not to Do) with Time-Series Cross-Section Data.”
American Political Science Review
89
(
3
):
634
47
.
Beck, Nathaniel, and Jonathan N. Katz.
1996
.
“Nuisance vs. Substance: Specifying and Estimating Time-Series-Cross-Section Models.”
Political Analysis
6
:
1
36
.
Beck, Nathaniel, and Jonathan N. Katz.
2004
.
“Time-Series–Cross-Section Issues: Dynamics, 2004.”
Retrieved September 13, 2016 (http://www.nyu.edu/gsas/dept/politics/faculty/beck/beckkatz.pdf).
Crenshaw, E. M., and A. Z. Ameen.
1997
.
“Population Dynamics and Economic Development: Age-Specific Population Growth Rates and Economic Growth in Developing Countries, 1965 to 1990.”
American Sociological Review
62
(
6
):
974
84
.
Crenshaw, Edward M., Matthew Christenson, and Doyle Ray Oakey.
2000
.
“Demographic Transition in Ecological Focus.”
American Sociological Review
65
(
3
):
371
91
.
Crenshaw, Edward, and Kristopher Robison.
2010
.
“Socio-Demographic Determinants of Economic Growth: Age-Structure, Preindustrial Heritage and Sociolinguistic Integration.”
Social Forces
88
(
5
):
2217
40
.
Dietz, Thomas.
2013
.
“Bringing Values and Deliberation to Science Communication.”
Proceedings of the National Academy of Sciences
110
(
Supplement 3
):
14081
87
.
Dietz, Thomas, and Andrew Jorgenson, eds.
2013
.
Structural Human Ecology: Risk, Energy and Sustainability
.
Pullman
:
Washington State University Press
.
Dietz, Thomas, and Andrew K. Jorgenson.
2014
.
“Towards a New View of Sustainable Development: Human Well-Being and Environmental Stress.”
Environmental Research Letters
9
(
3
):
31001
.
Dietz, Thomas, and Eugene A. Rosa.
1997
.
“Effects of Population and Affluence on CO2 Emissions.”
Proceedings of the National Academy of Sciences
94
(
1
):
175
79
.
Dietz, Thomas, Eugene A. Rosa, and Richard York.
2007
.
“Driving the Human Ecological Footprint.”
Frontiers in Ecology and the Environment
5
(
1
):
13
18
.
Duncan, Otis Dudley.
1959
. “Human Ecology and Population Studies.” Pp. 678–716 in
The Study of Population: An Inventory and Appraisal
, edited by P. M. Houser and O. D. Duncan.
Chicago, IL
:
University of Chicago Press
.
Ehrhardt-Martinez, Karen, Edward M. Crenshaw, and J. Craig Jenkins.
2002
.
“Deforestation and the Environmental Kuznets Curve: A Cross-National Investigation of Intervening Mechanisms.”
Social Science Quarterly
83
(
1
):
226
43
.
Ehrlich, Paul R., and John P. Holdren.
1971
.
“Impact on Population Growth.”
Science
171
:
1212
17
.
Ewing, Brad, et al.
2009
.
Ecological Footprint Atlas 2009
.
Oakland, CA
:
Global Footprint Network
. Retrieved September 19, 2016 (http://www.footprintnetwork.org/content/images/uploads/Ecological_Footprint_Atlas_2009.pdf).
FAO
.
2016
.
The State of Food Insecurity in the World 2015
.
Rome
:
Food and Agriculture Organization of the United Nations
. Retrieved September 19, 2016 (http://www.fao.org/hunger/en/).
Hughes, Melanie M., Lindsey Peterson, Jill Ann Harrison, and Pamela Paxton.
2009
.
“Power and Relation in the World Polity: The INGO Network Country Score, 1978–1998.”
Social Forces
87
(
4
):
1711
42
.
Jorgenson, Andrew K.
2003
.
“Footprint Consumption and Environmental Degradation: A Cross-National Analysis.”
Social Problems
50
(
3
):
374
94
.
Jorgenson, Andrew K.
2004
.
“Uneven Processes and Environmental Degradation in the World-Economy.”
Human Ecology Review
11
(
2
):
103
17
.
Jorgenson, Andrew K.
2005
.
“Unpacking International Power and the Ecological Footprints of Nations: A Quantitative Cross-National Study.”
Sociological Perspectives
48
(
3
):
383
402
.
Jorgenson, Andrew K.
2009
.
“The Sociology of Unequal Exchange in Ecological Context: A Panel Study of Lower-Income Countries, 1975-2000.”
Sociological Forum
24
(
1
):
22
46
.
Jorgenson, Andrew K., and Thomas J. Burns.
2007a
.
“Effects of Rural and Urban Population Dynamics and National Development on Deforestation in Less-Developed Countries, 1990–2000.”
Sociological Inquiry
77
(
3
):
460
82
.
Jorgenson, Andrew K., and Thomas J. Burns.
2007b
.
“The Political-Economic Causes of Change in the Ecological Footprints of Nations, 1991–2001: A Quantitative Investigation.”
Social Science Research
36
(
2
):
834
53
.
Jorgenson, Andrew K., and Brett Clark.
2011
.
“Societies Consuming Nature: A Panel Study of the Ecological Footprints of Nations, 1960–2003.”
Social Science Research
40
(
1
):
226
44
.
Jorgenson, Andrew K., and James Rice.
2005
.
“Structural Dynamics of International Trade and Material Consumption: A Cross-National Study of the Ecological Footprints of Less-Developed Countries.”
Journal of World-Systems Research
11
(
1
):
57
77
.
Jorgenson, A., J. Rice, and B. Clark.
2012
.
“Assessing the Temporal and Regional Differences in the Relationships between Infant and Child Mortality and Urban Slum Prevalence in Less Developed Countries, 1990–2005.”
Urban Studies
49
(
16
):
3495
3512
.
Jorgenson, Andrew K., James Rice, and Jessica Crowe.
2005
.
“Unpacking the Ecological Footprint of Nations.”
International Journal of Comparative Sociology
46
(
3
):
241
60
.
Jorgenson, Andrew K., Juliet B. Schor, Xiaorui Huang, and Jared Fitzgerald.
2015
.
“Income Inequality and Residential Carbon Emissions in the United States: A Preliminary Analysis.”
Human Ecology Review
22
(
1
):
93
106
.
Knight, Kyle W., Eugene A. Rosa, and Juliet B. Schor.
2013
.
“Could Working Less Reduce Pressures on the Environment? A Cross-National Panel Analysis of OECD Countries, 1970–2007.”
Global Environmental Change
23
(
4
):
691
700
.
Lamb, W. F., et al.
2014
.
“Transitions in Pathways of Human Development and Carbon Emissions.”
Environmental Research Letters
9
(
1
):
14011
.
Liu, Jianguo, Thomas Dietz, Stephen R. Carpenter, Marina Alberti, et al.
2007
.
“Complexity of Coupled Human and Natural Systems.”
Science
317
(
5844
):
1513
16
.
Liu, Jianguo, Thomas Dietz, Stephen R. Carpenter, Carl Folke, et al.
2007
.
“Coupled Human and Natural Systems.”
AMBIO
36
(
8
):
639
49
.
Longo, Stefano B., and Richard York.
2015
.
“How Does Information Communication Technology Affect Energy Use?”
Human Ecology Review
22
(
1
):
55
72
.
Marquart-Pyatt, Sandra T.
2010
.
“Environmental Sustainability: A Closer Look at the Factors Influencing National Ecological Footprints.”
International Journal of Sociology
40
(
2
):
65
84
.
Marquart-Pyatt, Sandra T.
2013
.
“The Implications of Structural Human Ecology for Environmental Concern's Global Reach.”
Pp. 159–86 in
Structural Human Ecology: New Essays in Risk, Energy and Sustainability
, edited by T. Dietz and A. Jorgenson.
Pullman
:
Washington State University Press
.
Marquart-Pyatt, Sandra T.
2015
.
“Environmental Sustainability: The Ecological Footprint in West Africa.”
Human Ecology Review
22
(
1
):
73
92
.
Rice, James.
2007
.
“Ecological Unequal Exchange: International Trade and Uneven Utilization of Environmental Space in the World System.”
Social Forces
85
(
3
):
1369
92
.
Rice, James, and Julie Steinkopf Rice.
2009
.
“The Concentration of Disadvantage and the Rise of an Urban Penalty: Urban Slum Prevalence and the Social Production of Health Inequalities in the Developing Countries.”
International Journal of Health Services
39
(
4
):
749
70
.
Roberts, J. Timmons, Bradley C. Parks, and Alexis A. Vásquez.
2004
.
“Who Ratifies Environmental Treaties and Why? Institutionalism, Structuralism and Participation by 192 Nations in 22 Treaties.”
Global Environmental Politics
4
(
3
):
22
64
.
Rosa, Eugene A., and Thomas Dietz.
2012
.
“Human Drivers of National Greenhouse-Gas Emissions.”
Nature Climate Change
2
(
8
):
581
86
.
Rosa, Eugene A., Richard York, and Thomas Dietz.
2004
.
“Tracking the Anthropogenic Drivers of Ecological Impacts.”
AMBIO
33
(
8
):
509
12
.
UNDP
.
2014
.
Sustaining Human Progress: Reducing Vulnerabilities and Building Resilience (2014 Human Development Report)
.
New York
:
United Nations Development Programme
. Retrieved September 20, 2016 (http://hdr.undp.org/sites/default/files/hdr14-report-en-1.pdf).
World Bank
.
2014
.
Risk and Opportunity: Managing Risk for Development (2014 World Development Report)
.
Washington, DC
:
World Bank
. Retrieved September 20, 2016 (http://siteresources.worldbank.org/EXTNWDR2013/Resources/8258024-1352909193861/8936935-1356011448215/8986901-1380046989056/WDR-2014_Complete_Report.pdf).
World Bank
.
2015
.
World Development Report 2015: Mind, Society, and Behavior
.
Washington, DC
:
World Bank
. Retrieved September 20, 2016 (http://www.worldbank.org/en/publication/wdr2015).
WWF
.
2012
.
Living Planet Report 2012: Biodiversity, Biocapacity and Better Choices
.
Gland, Switzerland
:
World Wildlife Fund
. Retrieved September 20, 2016 (http://assets.worldwildlife.org/publications/333/files/original/Living_Planet_Report_2012_Biodiversity__biocapacity_and_better_choices.pdf?1345733116&_ga=1.125506179.881339897.1474384749).
York, Richard, and Eugene a Rosa.
2012
.
“Choking on Modernity: A Human Ecology of Air Pollution.”
Social Problems
59
(
2
):
282
300
.
York, Richard, Eugene A. Rosa, and Thomas Dietz.
2003a
.
“Footprints on the Earth: The Environmental Consequences of Modernity.”
American Sociological Review
68
(
2
):
279
300
.
York, Richard, Eugene A. Rosa, and Thomas Dietz.
2003b
.
“STIRPAT, IPAT and ImPACT: Analytic Tools for Unpacking the Driving Forces of Environmental Impacts.”
Ecological Economics
46
(
3
):
351
65
.

NOTES

NOTES
1.
Sociological scholarship employs expectations from theoretical traditions including ecological modernization, world-systems approaches, ecologically unequal exchange frameworks, and world polity perspectives to examine how economic, political, ecological, and demographic factors drive change in environmental conditions or environmental sustainability. Initial studies in this vein examined society–environment relations as embodied in sociological work on EF (Dietz and Rosa 1997; Jorgenson 2003, 2005, 2009, Jorgenson and Burns 2007a, 2007b; Jorgenson and Rice 2005; Rice 2007; Rice and Rice 2009; Rosa, York, and Dietz 2004; York, Rosa, and Dietz 2003a, 2003b). More recent sociological work expands the focus to include measures of militarization (physical size of the military and military expenditures, Jorgenson and Clark 2011), income inequality (Jorgenson, Rice, and Crowe 2015), employment (Knight, Rosa, and Schor 2013), and information technology and energy use (Longo and York 2015) as drivers of sustainability.
2.
STIRPAT refers to STochastic Impacts by Regression on Population, Affluence, and Technology (York, Rosa, and Dietz 2003a, 2003b). The baseline STIRPAT equation includes the outcome or dependent variable as the environmental impact along with an intercept term, estimated coefficients linked with population and affluence measures, respectively, “technology” representing everything else in the equation, and an error term. The STIRPAT equation was designed for testing theoretically derived hypotheses from human ecology concerning the effects of population, affluence, and technology, which is now referred to as a core approach of structural human ecology (Dietz and Jorgenson 2013).
3.
See the  appendix for the full list of countries included in our analyses by region. Alphabetically, they are Algeria, Angola, Benin, Botswana, Burkina Faso, Burundi, Cape Verde, Cameroon, Central African Republic, Chad, Democratic Republic of the Congo, Congo, Cote d’Ivoire, Djibouti, Egypt, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Libya, Malawi, Mali, Mauritania, Morocco, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, Sierra Leone, South Africa, Sudan, Tanzania, Togo, Tunisia, Uganda, Zambia, and Zimbabwe.
4.
Our analyses exclude Cape Verde, Comoros, Madagascar, Mauritius, Sao Tome and Principe, and Seychelles, as well as Somalia or South Sudan, due to incomplete or missing data. All years include at least 24 countries with observations.
5.
The limited sample of five West African nations used in Marquart-Pyatt (2015) may explain the differences between our findings.

APPENDIX. LIST OF AFRICAN COUNTRIES (54) BY REGION (5)

Eastern Region (18)

Burundi

Comoros* 

Djibouti

Eritrea

Ethiopia

Kenya

Madagascar* 

Malawi

Mauritius* 

Mozambique

Rwanda

Seychelles* 

Somalia* 

South Sudan* 

Tanzania

Uganda

Zambia

Zimbabwe

Middle Region (9)

Angola

Cameroon

Central African Republic

Chad

Congo

Democratic Republic of the Congo

Equatorial Guinea

Gabon

Sao Tome and Principe* 

Northern Region (6)

Algeria

Egypt

Libya

Morocco

Sudan

Tunisia

Southern Region (5)

Botswana

Lesotho

Namibia

South Africa

Swaziland* 

Western Region (16)

Benin

Burkina Faso

Cape Verde

Cote d’Ivoire

Gambia

Ghana

Guinea

Guinea-Bissau

Liberia

Mali

Mauritania

Niger

Nigeria

Senegal

Sierra Leone

Togo

*
Not included in our analyses.