Essay: How dictatorships can intensify the detrimental impact of executive corruption on forest loss

Previous research finds that executive sector corruption increases forest loss. However, recent political events suggest that repression, largely through dictatorships, may exacerbate the impact of corruption on forest loss. Therefore, this study considers how nations with dictatorships provide a ‘political opportunity structure’ that facilitates the harmful impacts of executive corruption on forest loss. Dictatorships are hypothesized to create opportunities for executive corruption due to the power and lack of checks of the executive branch. To test this claim, I include an interaction term between a dummy variable that represents nations with dictatorships and executive corruption overall, executive bribes, and executive embezzlement using ordinary least squares (OLS) regression for a sample of 81 low and middle income nations from 2000. The dependent variable, forest loss, uses new multi-scalar, remote sensing based data from 2001 to 2014. Results suggest that dictatorships increase forest loss more in nations with high levels of executive corruption than low levels of executive corruption.
Introduction
Recent research on forest loss argues that forest loss is disproportionally high in low and middle income nations compared to high income nations due to corruption (Holland et al. 2008; Laurance et al. 2001; Laurance 2007; Koyuncu and Yilmaz 2009; Sommer 2017). Corruption, or ‘the abuse of entrusted power for private gain’ in the executive sector, is thought to be the most influential on forest loss compared to other sectors for several reasons (Dahlberg et al. 2016; Sundström 2016; Transparency International 2016). For example, the executive sector can grant liberal concessions to logging companies for bribes and profit kickbacks, which tend to result in high levels of forest loss (Palo 2004). Building upon this research, this study considers how dictatorships may exacerbate the effect of executive corruption on forest loss. In nations with dictatorships, executive corruption should be enhanced due to the power, and lack of checks of the executive branch (Palmer 2005; Dahlberg et al. 2016).
Previous research finds that dictatorships often accompany corruption and tend to intensify forest loss (such as the dictatorship under Suharto in Indonesia) (Palo 2004). From this insight, I argue that nations with dictatorships provide a “political opportunity structure” that exacerbates the harmful impacts of executive corruption on forest loss (Meyer 2004; Tilly et al. 2001; Eisinger 1973). However, this association has not been tested cross-nationally. Therefore, I include interaction terms between dictatorships, overall executive corruption, and executive bribes and embezzlement to test whether dictatorships facilitate forest loss in nations with high levels of executive corruption rather than lower levels. The study will give insight as to how political factors, specifically the leadership of nations, relate to corruption in the highest level of government to impact forest loss.
The organization of this study is as followed. Next, I review how dictatorships can intensify the detrimental impact of executive corruption on forest loss. I go on to describe the data, methods, and findings. I conclude by discussing the theoretical and policy implications of the study.
Dictatorships, Executive Corruption, and Forest Loss
Corruption in the executive sector concerns members of the executive branch, or their agents including include presidents, governors, cabinet members, dictators, parliamentary leaders and those who directly follow orders from these leaders (Sundström 2016; Dahlberg et al. 2016). At this level of government, leaders and their agents often grant favors in exchange for bribes, or kickbacks, as well as embezzle or misappropriate state resources for their own use (Agrawal 2007; Halkos et al. 2015; Sundström 2016; Varieties of Democracy 2016; Dahlberg et al. 2016), which limit the amount of environmental protection funds, undermine forestry conservation laws, and encourage unlawful contracts that increase forest loss (Geist and Lambin 2002; Sundström 2016).
The effects of executive corruption on forest loss may be exacerbated in nations with dictatorships (Papaioannou and vanZanden 2015; Møller and Skaaning 2012). A dictatorship is a form of government whereby a nation is ruled directly by one person or political entity (Power 2008). In such nations, the power of this body is reinforced through various mechanisms, such as repression, to ensure that the ruling dictatorship remains strong and uncontested (Coppa 2006). This type of authoritarianism involves politicians regulating basically every aspect of private and public behavior, such as activities in the forestry sector (Papaioannou and vanZanden 2015; Møller and Skaaning 2012). Previous research suggests that dictatorships should have higher levels of forest loss, especially when the executive sector is corrupt (Sundström 2016; Olsen 1993). I review three of these relationships below. Table 1 summarizes the relationships between dictatorships and executive corruption on forest loss.
(Table 1 goes about here)
First, dictatorships have lower levels of activism than democratic nations, which can suppress public dialogue about the environment (Li and Reuveny 2006; Crenshaw and Jenkins 1996; Paxton 2002). The Malaysian dictatorship is one example of this process (Laurance 2004; Speechly and Van Helden 2012). Although corruption is a concern for the people of Malaysia, a lack of a transparent and accountable systems of governance allows corrupt executive politicians to lease out licenses for logging to corporations and private individuals at the expense of Malaysian land owners (Laurance 2004; Speechly and Van Helden 2012; Straumann 2014). Due in part to Malaysia’s system of government, its leaders often engage in money laundering and tax evasion (Straumann 2014). Logging concessions, permits, and contracts are often directly controlled or held by the powerful political and economic ruling class, which support corporate logging for export, and should result in increased deforestation (Laurance 2004; Speechly and Van Helden 2012; Straumann 2014).
Second, dictatorships need not be responsive to political activism, compared to democratic nations, because they often lack electoral accountability to their people (Ehrhardt-Martinez, Crenshaw, and Jenkins 2002; Shandra 2007). Therefore, dictators are often not accountable to laws that reduce corruption, and can exploit them for their own gain (Severin 2010). For example, executive corruption in the dictatorship of the Democratic Republic of the Congo allows deforestation to continue despite laws blocking corrupt actions (Butler 2013; Severin 2010). In 2002, the Democratic Republic of the Congo created a moratorium against commercial logging; however, loggers found their way around the suspension by bribing officials for artisanal permits (these permits are usually only reserved for community logging) (Butler 2013; Debroux et al. 2007). This allows logging companies to target endangered tree species for export to nations such as China and Europe (Butler 2013). Although the government has recently implemented other initiatives such as the Agricultural and Rural Sector Rehabilitation Support Project (PARSAR), executive sector officials may accept bribes so that companies and militiamen can circumvent such laws (Severin 2010; Butler 2013). Moreover, in the Democratic Republic of the Congo, the majority of the parties involved in the unlawful sale of wood are former militiamen, which adds fear that if their operations and products are stopped civil war may resume (Severin 2010). Thus, forest loss may continue unabated in similar political climates for fear of violence and for the dictator’s power to remain unchallenged (Severin 2010).
Third, dictatorships often use propaganda and have tight constraints on the press, which suppresses the diffusion of information and can institutionalize corruption (Li and Reuveny 2006; Akçay 2006; Pellegrini and Gerlagh 2006). Therefore, corruption becomes ‘normalized’ and commonplace, making it more difficult to interrupt (Pellegrini and Gerlagh 2006). Furthermore, such governments also use violence to suppress uprisings of its civilians against issues such as environmental rights and corruption (Sears and Pinedo-Vasquez 2011). One such incident occurred in the dictatorship of Peru (Gibbs 2012). In 2000, while most of the forestland in Peru is held by the state or occupied by indigenous people, several logging concessions have instead given the power to large timber companies (Sears and Pinedo-Vasquez 2011; Urrunaga et al. 2012; Putz and Romero 2015). In an attempt to reform forest policy in Peru, laws were later adopted to incorporate a free trade treaty with the United States that would encourage legal forest trade (Urrunaga et al. 2012; Oliveira et al. 2007; Gibbs 2012). However, in result, several executive decrees created loopholes to convert forestland into agricultural land (Urrunaga et al. 2012). Despite civilian protest, which resulted in violence, the executive sector was able to defend its decrees for their own monetary gain (Oliveira et al. 2007).
In sum, the power and scope of the executive sector in dictatorships allows their authority to go unchecked and difficult to contest (Oliveira et al. 2007). Therefore, the ‘political opportunity structure’ found within strong executive sectors can allow corruption to take place undeterred. Executive corruption in these government systems takes the form of bribes and embezzlement, which in turn allow forest loss to continue unabated (Oliveira et al. 2007). From the above, I hypothesize that nations with dictatorships increase forest loss more in nations with high levels of executive corruption than low levels of executive corruption. Before testing the hypotheses of this study, I review the methodology and data.
Methodology
Sample
The sample consists of 81 low and middle income nations. I define low and middle income nations using the World Bank Atlas method (Shandra et al. 2016). Nations are included in the sample if they have a GNI of 12,475 or less (World Bank 2015). I exclude high income nations because the dynamics of forest loss differ in these countries, and therefore are considered separate from low and middle income nations in cross-national analyses (Shandra et al. 2011a). Moreover, low and middle income nations generally have higher levels of forest loss, corruption, and dictatorships (Crenshaw and Jenkins 1996; Shandra et al. 2010). Therefore, the sample is restricted to low and middle income nations (Shandra et al. 2010, 2016)1.
Statistical Model
I use ordinary least squares regression to analyze the effects of executive corruption and dictatorships on forest loss (Shandra et al. 2016). Researchers use this methodology to estimate the factors that impact deforestation (e.g., Shandra et al. 2010 2016; Jorgenson and Burns 2007).  The formula for this model is below:
yi = a + b1X1 + b2X2 …+ bkXk + ei
where,
yi = dependent variable for each country,
a = the constant,
b1 to bk = unstandardized coefficients for each independent variables,
xk = independent variables for each country, and
ei = error term for each county.
I check to make sure the models adhere to regression assumptions (Shandra et al. 2016; Allison 1999). First, variance inflation factor scores, reported in table 3, reveal no potential problems with multicollinearity (York, Rosa, and Dietz 2003).  Second, I transform variables when appropriate and note it in table 2 (Tabachnick and Fidel 2013). Third, based on standardized residuals, it appears that any potentially extreme cases are not biasing the results (Tabachnick and Fidel 2013). Fourth, there appears to be issues associated with heteroscedasticity based on Breush-Pagan statistics for each model (Juhl and Sosa-Escudero 2014). Therefore, I report robust standard errors (Tabachnick and Fidel 2013). Below I discuss the dependent variable, forest loss.
Dependent Variable
Forest Loss: Cross-national research on forest loss often uses data from the United Nation’s Food and Agriculture Organization’s Global Forest Resources Assessment (e.g., Shandra et al. 2011 2016). However, some of these data are gathered using collection methods that differ by nation (Grainger 2008). Forestry statistics in some nations are more reliable than others that are estimated from remote sensing surveys or extrapolated from outdated forestry inventories (Food and Agriculture Organization 2015; Grainger 2008).
In result, I use newly available data on forest loss from high resolution satellite imagery in order to eliminate this potential source of error (World Resources Institute 2016; Hansen et al. 2010)2. Following Rudel (2013), I use readily available data from the World Resources Institute (2016) Global Forest Watch web page. I calculate the change in forest loss by dividing the provided number of hectares of forest losses in a country from 2001 to 2014 by the country’s total forest size in hectares for 2000, yielding the change of forest loss over this period of time (Rudel et al. 2016).
One limitation of these data is that this variable is only available for the time period of 2001-2014, although this type of measurement is in line with past cross-national research (Shandra 2007; Shandra et al. 2016). Another limitation of these data is that they are not suitable for over time comparisons using panel regression methods (Rudel 2017). I log this variable because it is skewed. In table 2, I provide a bivariate correlation matrix for all the variables used in the analysis. Unless otherwise parenthetically noted, all data may be obtained from the World Bank (2015).
(Table 2)
Independent Variables
Executive Corruption: According to Dahlberg et al. (2016), this variable measures the extent to which members of the executive branch, or their agents grant favors in exchange for bribes, kickbacks, or other material inducements, and how often they steal, embezzle, or misappropriate public funds or other state resources for personal or family use. High numbers of this variable represent high levels of executive corruption (Dahlberg et al. 2016). Based on the above discussion, I expect higher levels of executive corruption are associated with higher levels of forest loss. These data, as well as the data for the following three variables, are from the Varieties of Democracy Dataset (Dahlberg et al. 2016). The creators of this dataset index the following two variables to measure the present variable (Dahlberg et al. 2016).
Executive Embezzlement: This variable measures the extent to which the executive sector and their agents steal, embezzle, or misappropriate public and state funds for family or personal use (Dahlberg et al. 2016). This variable is multiplied by negative 1 so that high numbers represent high levels of executive embezzlement.
Executive Bribes: This variable measures the extent to which the executive sector and their agents grant favors in exchange for bribes, kickbacks, or other material inducements (Dahlberg et al. 2016). Again, this variable is multiplied by negative 1 so that high numbers represent high levels of executive bribery.
Dictatorships: This measure is a dummy variable with a score of one representing nations with dictatorships in the year 2000, and zero representing all other types of governments (democracies, centrist, etc.) (Dahlberg et al. 2016). These data are coded based on regime type data from the Varieties of Democracy Dataset (Dahlberg et al. 2016).
Control Variable Selection and Measurement
The control variables selected, as well as the time period for which they are measured is based on previous cross-national research (Shandra et al. 2016). I note the time period for which each control variable is included for in parentheses, and the variable name is italicized. Rudel (2016 1989), Rudel et al. (2015), and Ehrhardt-Martinez, Crenshaw, and Jenkins (2002) describe the importance of including economic factors of gross domestic product per capita (GDP) (2000), GDP growth (1990-2000), forestry production (available from FAO 2016), and agricultural economic activity (2000) when analyzing forest loss.
There is also a large literature that finds that state factors of democracy (2000) (available from see Vanhanen 2014) and total government expenditures (as a percentage of GDP) to impact deforestation (Jorgenson and Burns 2007; Shandra 2007). Traditional studies of forest loss also find that demographic factors such as population density (2000) are a predictor of forest loss (see Rudel and Roper 1997; Rudel 1989).
Lastly, international factors, including international environmental non-governmental organizations (EINGOs) (as a percentage of total population) (2000) (available from Shandra et al. 2008), International Monetary Fund (IMF) loans (as a percentage of total population) (2000), and debt service (as a percentage of GDP) (2000), have been shown to have an impact on forest loss (see Bunker 1985; Shandra et al. 2009; Timmons and Parks 2007). Therefore, I include the above control variables at their indicated time period.
Findings
Table 3 includes the ordinary least squares regression estimates of forest loss.  The first number presented is the unstandardized coefficient, the second is the standardized coefficient, and the third number in parentheses is the robust standard error (Shandra et al. 2016).  I report one-tailed tests because of the directional nature of the hypotheses (Shandra et al. 2016).  In every equation, I include a dummy variable for dictatorships. I also include the variables that represent GDP per capita, economic growth, forestry production, agricultural domestic economic activity, democracy, total government expenditures, population density, EINGOs, IMF credit, and debt service. Each equation contains one of the three measures of executive corruption. I organize the models in this way because the measures of corruption are highly correlated with each other.
(Table 3)
Let me begin by discussing the main effects of executive corruption on forest loss. Across table 3, I find that the coefficients that represent executive corruption, executive embezzlement, and executive bribes are positive and statistically significant. This suggests that higher levels of executive corruption are associated with increased forest loss. This finding supports previous research that finds executive corruption increases forest loss due to vested interests between illegal loggers and politicians and the granting of liberal concessions (Palo 2004; Currey and Ruwindrijarto 2001; Sommer 2017).
I find a number of other factors are associated with forest loss. First, I find that the coefficients that represent gross domestic product per capita are negative and significant in every equation. This suggests that higher levels of gross domestic product correspond with less forest loss, which is most likely because wealthier nations often import natural resources from poorer nations to offset their internal environmental damage (Burns, Kick, and Davis 2003; Shandra et al. 2016). Second, I find that the coefficients that represent economic growth are positive and significant in every equation. Rudel (1989) and Jorgenson (2006) argue that this is because there are large amounts of capital available for investment in activities that accelerate forest loss during economic expansion (Shandra et al. 2016). Lastly, I find that the coefficients that represent population density are associated with less forest loss. The coefficients that represent this variable are negative and significant in every equation. This may be due to migration to urban areas away from agricultural farms and forests (Jorgenson and Burns 2007; Rudel 2016).
There are also non-significant findings. First, I find that the dummy variable that represents dictatorships is not significant. This suggests that dictatorships alone do not significantly increase forest loss. This non-finding is revisited below. Second, I find a number of economic factors are not related to forest loss. The coefficients that represent forestry production and agricultural economic activity fail to reach levels of statistical significance. Third, I find that democracy and total expenditures are not associated with forest loss3. The coefficients that represent these variables fail to reach levels of statistical significance. Fourth, I find that the coefficients that represent all of the international factors, including EINGOs, IMF credit, and debt service fail to reach levels of significance.
(Table 4 goes about here)
In table 4, I examine the interactive effects of executive corruption and dictatorships to test if nations with dictatorships create a ‘political opportunity structure’ that facilitates the detrimental impact of executive corruption on forest loss. As mentioned above, I hypothesize that dictatorships will increase forest loss more in nations with higher levels of executive corruption than lower levels of executive corruption.
Across table 4, the coefficients that represent each interaction term are positive and statistically significant in every equation. This suggests that nations with dictatorships have higher levels of forest loss in nations with higher levels rather than lower levels of executive corruption. When I calculate and graph the effects of executive corruption on forest loss in dictatorships, I find that nations with dictatorships have higher levels of forest loss at higher levels rather than lower levels of executive corruption.
The predicted effects of these relationships (see figures 1-3) illustrate that dictatorships have different effects on forest loss at different levels of executive corruption. In these figures, I use the coefficients from table 4 to graph the predicted change in dictatorships as executive corruption simultaneously increases, holding all continuous covariates at their mean. I find that the effect of dictatorships on forest loss is relatively low when executive corruption is low. This indicates that in nations with dictatorships and low levels of executive corruption there are lower levels of forest loss. In other words, when executive corruption is high in dictatorships, there are higher levels of forest loss. In nations with dictatorships, initial increases in executive corruption results in an incline forest loss, supporting the hypotheses of this study. In particular, in dictatorships, as executive corruption increases, forest loss steadily inclines (as indicated by the upward sloping line).
(Figures 1-3 go about here)
These predicted effects offer support for the hypothesis that dictatorships facilitate the negative environmental impacts of executive corruption, and that low levels of executive corruption can help absorb the negative environmental effects of dictatorships. The other findings remain stable and consistent across the new model specifications (except for forestry production in equation 4.2). All other results are similar to the results reported in table 4.
Discussion and Conclusion
In this study, I find that dictatorships increase forest loss more at higher levels than at lower levels of executive corruption. Therefore, this study begins to show support for the idea that dictatorships can facilitate the negative environmental impacts of executive corruption on forest loss. The main findings demonstrate the importance of considering the political structure of a nation when accessing how corruption impacts the natural environment (Bradshaw 1987; London and Smith 1988; Smith 1996).
These results have important implications for enhancing government transparency and anti-corruption policies in forestry (Dadge and Thomas 2013; Kishor and Rosenbaum 2012). However, as the present analysis shows, such solutions may not be possible in dictatorships where the executive power goes unchecked and can circumvent anti-corruption policies (Kishor and Damania 2007; Gore et al. 2013; Kolstad and Søreide 2009). Therefore, at the most extreme, state leaders and public officials in low and middle income nations can aim to remove logging concessions, de-incentivize vested interests, and implement internal processes to punish corrupt officials (Bryant and Bailey 1997; Callister 1999; Goncalves, et al. 2012). Going forward, researchers may continue to untangle the relationships between executive corruption and political factors on forest loss. At the very least, corruption in cross-national research should be conceptualized within a framework of government bodies and political environments.