Environmental consequences of the shadow economy

The underground economy is generally defined as a set of economic activities that take place outside the framework of bureaucratic public and private sector establishments (Hart, 2008). Nowadays, it is impossible to conceive an economy free from the underground or the informal economic sector . According to estimates , 54.3% of firms around the world are competing against informal or unrecorded firms while 68.3% of formal firms in Sub-Saharan Africa (SSA) are competing against informal firms.

The average size of the shadow economy in 2015 was estimated at 35.76% in SSA. The largest size of the underground sector (as a % of GDP) were observed in Benin (48.28%), Nigeria (52.49%) and Zimbabwe (67%) and the lowest observed in Mauritius (19.23%), Namibia (21.78%) and South Africa (21.99%) (Medina and Schneider, 2018). Figure 1 shows the size of the shadow economy (in percentage of GDP) in selected Sub-Saharan Africa countries and the average rate of the shadow economy in the world.

A voluminous literature has been concentrated on what drives the shadow sector (Krakowski, 2005; Maddah and Sobhani, 2014; Njangang et al., 2014). Other have studies had examined the consequences of the shadow sector on economic development (Abid and Salha, 2013; Chatterjee and Turnovsky, 2018; Kumar, 2013), on internet usage (Elgin, 2013), energy consumption (Karanfil, 2008; Mustafa et al., 2016), on financial development (Berdiev and Saunoris, 2016; Capasso and Jappelli, 2013), on corruption and institutional quality (Dreher and Schneider, 2009; Dreher et al., 2009).

However, relatively little is known on the environmental consequences of the shadow economy. Theoretically, the underground sector affects environmental quality through two opposite channels (Elgin and Oztunali, 2014): the scale effect and the deregulation effect. The scale effect presents the mechanism through which a larger (smaller) informal sector size is associated with a lower (higher) level of environmental pollution. As argued by these authors, the scale effect is motivated by the fact that the informal economy operates on a small scale (especially compared to the formal sector) with a highly labor-intensive and less capital intensive production technology, and thus generate lower levels of pollution.

Thus, countries with a large size of the informal sector are characterized by low level of pollution. Contrary to the scale effect, the deregulation effect suggest that a larger size of the informal sector is associated to higher levels of pollution since informal activities pose a serious hindrance to environmental regulations (Baksi and Bose, 2010). Thus, the informal

economy’s share of the overall economy could have a positive or negative impact on overall emissions, depending on which of these factors dominate.

Till date, empirical studies examining the informal economy- pollution nexus are infrequent. The seminal work of Biswas et al. (2012) show that increases in the size of the shadow economy increases carbon emissions and sulfur dioxide emissions in more than 100 countries. Moreover, they found that the negative effect of the shadow economy is exacerbated in the presence of corrupt governments. Imamoglu (2018) finds that an increase in the size of the informal sector activity significantly increases carbon emissions in Turkey. Chen et al. (2018) find that both the shadow economy and the regulatory quality increase carbon emissions in China. Their results also indicate that tighter environmental control would help curbing pollution.

Elgin and Oztunali (2014) detect an inverted U-shaped relationship between the informal economy and environmental pollution in a panel of 52 countries between 1999 and 2009 period. This result suggests that lower and higher levels of informality are associated to lower levels of environmental pollution, while medium level of informality is associated to higher levels of environmental pollution.

To the best of our knowledge, no previous study has been conducted on the environmental effect of the shadow economy in the African context. Existing studies (see for example Nkengfack and Kaffo, 2019; Zoundi, 2016; Abid, 2016) concentrate on the effect of official economic activities of environmental outcomes. Doing so, they ignore the potential effects of the informal sector on the environment. Thus, there is need to address the issue of growing shadow activities and sustainable development.

Indeed, in the presence of unofficial economic activities, the official GDP is not accurately measured and the conclusions on the linkages between economic growth and environmental quality may be partially valid. To fill this gap, we propose in this paper to investigate the effect of the shadow economy on the environmental quality on a sample of 22 SSA countries over the period spanning from 1991 to 2015. The motivations to conduct such a study are threefold. Firstly, to the best of our knowledge, no paper has yet investigated the environmental effects of the informal sector in Sub-Saharan Africa (SSA), whereas these countries are severely confronted with the informal economy.

In this context, previous studies have probably provided an incomplete picture of the effect of economic activities on the environment. Secondly, SSA is one of the most vulnerable regions to the adverse effect of climate change, as their economies rely on agriculture and the exploitation of natural resources that generate substantive revenues to government budget. In this context, increases in greenhouse gas emissions, mainly attributed to CO2 emissions from fossil fuels burning and the conversion of forest land, will have severe impacts on future economic growth and undermine gains in development and poverty reduction (Hogarth et al., 2015).

Thirdly and lastly, we use panel Autoregressive Distributed Lag (ARDL) to investigate the long run effect of the size of the shadow economy on carbon emissions. The practical advantage of the panel ARDL over the conventional panel estimators (i.e., fixed-effects, random effects) is that it allows the determination of the long-run effects of the parameters of interest and revelation of their speed of adjustment to long-run equilibrium, while allowing for the intercepts, short-run coefficients and error variances to differ freely across cross-sections (Baek, 2016).

The robustness of our results is checked through the panel ordinary least squares (OLS); fully-modified least squares (FMOLS) and dynamic ordinary least squares (DOLS) estimation techniques. The outline of our paper is organized as follows: section 2 describes the data, presents the descriptive statistics and the correlation between our variables. Section 3 presents the econometric model and the estimation techniques. Section 4 presents the empirical results and the discussions, and the last section, section 6, concludes the article.

Data and descriptive statistics

Our data cover a sample of twenty-two Sub-Saharan Africa countries over the period from 1991 to 2015. Data are gathered from the World Bank (2018) and Medina and Schneider (2018). The sample and time period are primarily dictated by data availability. The list of countries included in the study, classified per income group is presented in the appendix.

The dependent variable is CO2 emissions (in metric ton), which has been used as proxy for environmental quality by the relevant literature (e.g. Ahmad et al., 2016; Baek, 2016). The key explanatory variable is the shadow economy as a percentage of GDP. This variable is obtained from Medina and Schneider (2018). They used the Multiple Indicators Multiple Causes (MIMIC) approach to compute the size of the shadow economy (as a percentage of GDP) in 158 countries in world over 1991 to 2015.

According to the authors, the shadow economy reflects mostly legal economic and productive activities that, if recorded, would contribute to national GDP. This definition of the shadow economy tries to avoid illegal or criminal activities, do-it-yourself, or other household activities. In order to overcome omitted-variables bias as suggested by Ang (2007), we consider other relevant determinants of CO2 emissions. They include energy use, trade openness and urbanization rate (Boutabba, 2014, Alam et al., 2015; Zhang et al., 2017).

The variable descriptions, data definitions and data sources are reported in Table 1. [Insert table 1 here] Descriptive statistics and the correlation between our variables are reported in Table 2. It should be noted that the variables are taken before the natural logarithm is introduced. In addition, the correlation matrix did not reveal severe correlations between our variables. Indeed, the highest correlation coefficient between our variables (0.5917) is lowest than 0.7 used as the rule of thumb for high correlation. This suggests that neither collinearity nor multicollinearity is considered as a concern in our data.

Model to be estimated and the econometric strategy

Following Biswas et al. (2012) and Chen et al. (2018), we rely on the following simplified model to examine the effects of the underground sector on environmental quality in SSA countries: Where CO2it is the amount of CO2 emissions from fossil fuel (in metric ton) in country i for period t; undergroundit is the underground sector (as a share of GDP) in country i in time t; enerit is the energy use per capita in country i in period t; tradeit is the trade openness in country i at time t; urbanit is the rate of urbanization in country i at time t; ԑit is the error term. Our coefficient of interest is α1 since it captures the long run effect of the underground sector on carbon emissions. All the variables are transformed with natural logarithm in order to eliminate heteroscedasticity.

Theoretically, the effect of the shadow economy is undetermined. The increase in the size of the underground economy will be coupled to more carbon emissions if the deregulation effect holds. Conversely, the expansion in the size of the shadow economy will moderate CO2 if the scale effect holds. In addition, it is expected that α2 is positive since higher energy consumption results in greater economic activity and stimulates carbon emissions. Finally, the effect of trade openness is not determined a priori as suggested by Halicioglu (2009) and is expected to be positive or negative depending on the level of development of countries.

If α3 is positive, trade openness is harmful for environmental quality. Otherwise, if α3 is negative, trade openness improves environmental quality through technology transfer and the adoption of environmental friendly technologies. Our econometric strategy follows four stages. In the first stage, we examine the stationarity of our variables in order to avoid fallacious regressions. To this end, we use three unit root tests that account for common as well as individual unit root processes. They include the Levin, Lin and Chu (LLC) test, the Breitung test and the Im, Pesaran and Shin (IPS) test.

In the second one, we apply panel cointegration tests to explore the existence of the long-run equilibrium relationship among CO2 emissions, shadow economy, energy use, trade openness and urbanization. Cointegration among variables is checked through Predoni’s (2004) and Kao’s (1999) cointegration tests. In the third stage, we employ the panel Autoregressive Distributed Lag (ARDL) or the Pooled mean group (PMG) estimator to investigate the short-run and the long-run effects of the shadow economy on carbon emissions. The PMG estimator, proposed by Pesaran et al. (2001), is helpful in investigating the long-run and short-run effect of the main drivers of carbon emissions.

The main advantage of this method is that it provides consistent estimators, irrespective of whether the variables are stationarity at level or at first difference (Baek, 2016). However, the PMG ceases to be applicable when the order of integration of variables is greater than 1. Following Pesaran et al. (2001), the ARDL (p, q, q, q) is derived from equation (1): Where Xit is a (3×1) vector of explanatory variables, (p, q, q, q) represent the maximum order of integration of our variables; μi represents the fixed effects; α_it^\’  are (3×1) vector of coefficients, λit are scalars, Δ is the first difference operator and ∑ is the sum operator.

The error correction model is derived from equation (2) as follows: Where:  ;  ;   j=1,2,…, p-1 and    j = 1,2,…,q-1. The parameter φi is the convergence coefficient or the error-correcting speed of adjustment. It measures the speed of adjustment toward equilibrium and should be negative and significant to allow for the existence of a cointegration relationship between CO2 emissions and it determinants. In addition to the PMG estimator, results obtained from the mean group (MG) are reported to compare our results.

As stated by Pesaran et al. (2001), the PMG imposes pooled long run effects; the MG is obtained by estimating independent regressions and then averaging the (unweighted) coefficients. Finally, we rely on the ordinary least squares (OLS), the fully-modified least squares (FMOLS) and the dynamic ordinary least squares (DOLS) as robustness check to the PMG estimator. In the context of a panel model, the OLS estimator is asymptotically biased and its distribution depends on nuisance parameters. To address such a bias, we refer to FMOLS and DOLS proposed by Pedroni (2001, 2004).

Empirical results and discussions

Panel unit root tests

The unit root tests results are reported in Table 3. The results indicate that the null hypothesis of the absence of the unit root process is rejected (at 1% for LLC and IPS but at 5% for Breitung) for the variable shadow, suggesting that the variable is I(0). However, the results indicate that the remaining variables (CO2, energy use per capita, trade openness and urbanization rate) are stationary at first difference. This indicates that the maximum order of integration of our variables is 1. [Insert table 3 here]

Panel cointegration results

Given the results of the stationary test, the next step consists of testing the existence of a long run relationship between CO2 emissions and its determinants. In this paper, we use panel Pedroni’s (2004) residual cointegration test. This test relies on two categories of cointegration test and seven statistics. These tests are classified based on the within-dimension and the between dimension of the autoregressive coefficients for each country in the panel.

The null hypothesis is that there is no cointegration and the alternative is that there is cointegration between variables. The results of the Pedroni (2004) cointegration test are presented in Table 4, and suggest that there are three statistics for the within-dimension and two statistics for the between-dimension that are significant at 1% significance level. This reveals the existence of the cointegration between the variables.

To check for the robustness of the cointegration result, we refer to the Kao’s (1999) residual unit root test. This test is constructed on the same approach as the Pedroni’s (2004) test but specifies cross-section intercepts and homogenous coefficients on the first stage regression. The null hypothesis assumes the absence of cointegration, whereas the alternative hypothesis implies the existence of the cointegration relation between the variables. Table 5 provides the result of the cointegration test suggested by Kao (1999). The result rejects the null hypothesis of the absence of cointegration at 5% significance level.

Long and short run estimates

Table 6 presents the results obtained from the alternative estimators, including the PMG, and the MG. First of all, the convergence coefficients are negative and significant for the estimators, supporting the evidence of a stable long-run relationship among the variables. The coefficients of -0.242 and -0.493, respectively for the PMG and the MG estimators suggest that a deviation from the long-run equilibrium level of carbon emissions in one year is corrected, respectively, by 24.6%; 49.3%  over the following year. However, we refer to the Hausman specification test to discriminate between the PMG and the MG estimator. The calculated Hausman statistics is 1.08 and the p-value is 0.8970, indicating that the PMG estimator, the efficient estimator under the null hypothesis, is preferred.

Concerning our variable of interest (i.e. the underground sector), our results indicate that an increase in the size of the shadow economy will result in lower levels of pollution both in the short-run and in the long-run; even if the coefficient associated to the shadow economy is statistically significant only at long-run. At the long run, a 1% increase in the shadow sector (as percentage of GDP) will lead to 0.479% abatement in carbon emissions. This result is contrary to Imamoglu (2018), Chen et al. (2018) and Biswas et al. (2012) who find that an increase in the size of the shadow sector is associated to more pollution.

A possible explanation of our result is that the underground sector in SSA is dominated by a large number of small size enterprises or family businesses including small scale agriculture and fishing, trading and other services. In addition, due to problems with credit access and less secured investment conditions, the underground sector in SSA is less capital intensive and more labor intensive than the formal sector, leading to a reduction in carbon emissions from the informal sector.

This result is in line with the scale effect described by Elgin and Oztunali (2014).According to this effect, when the informal sector is mainly constituted of firms engaged in labor-intensive activities and these activities do not have adverse effects on the environmental quality. The results also indicate that an increase in the energy use per capita has a positive and significant effect on carbon emissions both in the short-run and long-run, the effect of energy variable being more important in the long-run. An increase by 1% of energy use per capita results to an increase in carbon emissions by 1.096% in the short-run and 1.995% in the long-run.

This result is consistent with Boutabba (2014) for India and Baek (2016) in a panel of five countries of the Association of Southeast Asian Nations (ASEAN). This result is probably explained by the fact that the energy demand in SSA in mainly met by fossil fuels. According to United Nations Environment Programme (2017) report, oil is the main source of energy consumed in Africa (42 % of its total energy consumption), followed by gas (28%), coal (22%) and hydroelectricity (6%). [Insert table 6 here]

Concerning the impact of trade openness on the environmental quality, our results indicate that two forms of relationship exist. In the long-run, a 1% increase in the trade openness ratio (as percentage to GDP) will significantly reduce carbon emissions by 0.305%. In the short run, the effect of trade openness on carbon emissions is positive but not statistically significant.

This result suggests that all things being equal, trade openness is not a threat to environmental quality. An implication of this result is that trade openness is coupled to the reduction of high pollution-oriented activities, the increase of high tech imports in the exchange of goods and services, and the transfer of environmental friendly technologies. This result supports the findings of Charfeddine and Khediri (2016) in a federation of seven emirates and contrasts with the findings of Farhani and Ozturk (2015). In addition, Zhang et al. (2017) found that the effect of trade openness on carbon emissions differs, depending on the group of countries considered. In fact, they observed that trade openness augments carbon intensity in non-OECD countries but insignificantly reduces it in OECD countries.

Finally, an increase in the urbanization rate has a mixed impact on CO2 emissions. The findings suggest that at the early stage of the urbanization process, rapid urbanization is compatible to reductions in carbon emissions, while the opposite effect is observed in the long-run. This suggests that the impact of urbanization on CO2 emissions might not be linear (Poumanyvong and Kaneko, 2010, Shahbaz et al., 2016) since the sign and magnitude of its coefficients change between the short and the long run.

Robustness check

The results of the alternative estimation techniques are reported in Table 7. Results obtained from the OLS show that the effect of the underground sector on carbon emissions is negative and statistically non-significant. However, the results obtained from FMOLS and DOLS estimators are quite similar in terms of sign, magnitude and statistical significance. Indeed, a 1% increase in the size of the shadow sector decreases carbon emissions by 1.34% and 1.30% respectively for FMOLS and DOLS.

These findings are in line with results obtained from the PMG estimator and indicate that the expansion in the size of the shadow sector is not associated to more pollution in SSA. In addition, an increase in energy use per capita is associated to the increase in carbon emissions. This result is valid for OLS, FMOLS and DOLS. Trade openness has mixed effects on carbon emissions, the effect being negative and significant on carbon emissions with the OLS, positive and statistically significant with the FMOLS and non-significant when the DOLS is considered. Finally, an increase in the urbanization rate significantly increases CO2 emissions in Africa in the long run.

Results of estimations based on the level of development of countries: to better understand the effect of informality on the environment among the selected countries, we estimate equation (1) separately for different groups of countries. Based on the World Bank classification (2018), we have nine low income countries (LIC), eight lower-middle-income countries (LMIC) and five upper-middle-income countries (UMIC). Table 8 summarizes the results of the estimations.

The Hausman specification test suggests that the PMG estimator is preferred. The convergence coefficients (ϕi) are negative and statistically significant for all the estimated models. This indicates the existence of a long-run cointegration relationship between carbon emissions and its determinants for all the income groups. The estimated coefficients of the shadow economy are negative but statistically significant only in the long-run for LMIC and significant in the short run for the UMIC.

This result implies that an increase in the size of the shadow economy in LIC does not necessarily lead to increase in carbon emissions. The estimated elasticities of energy consumption are positive and highly significant in the long-run in all the countries groups. For the LIC, LMIC and UMIC, a 1% increase in the energy use per capita increases CO2 emissions by 3.573%, 0.888% and 1.861% respectively. This confirms the crucial role of energy consumption in the explanation of increases in carbon emissions in SSA countries.

Concluding remarks

This paper has explored the effect of the underground economy on carbon emissions for a selected sample of 22 Sub-Saharan Africa countries over the period 1991-2015. The novel feature of our study relies on the utilization of the informal economy (as a percentage of GDP) to account for the effect of the unrecorded economy on the environmental quality. The Pedroni’s (2004) and Kao’s (1999) tests suggest the existence of a long-run cointegrated relationship between the carbon emissions, informal economy, energy use, trade openness and urbanization rate.

Our empirical findings, obtained from the PMG estimators, reveal that expansion in the size of the underground sector tend to reduce carbon emissions both in the short and the long run, but its effect is statistically significant only in the long run. In addition, the results show that, both in the short-run and the long-run, the estimated coefficient of energy consumption are positive and highly significant, while trade openness reduces carbon emissions only in the long-run. These results are supported by the OLS, FMOLS and DOLS.

In addition, our results show that the effect of the informal sector on CO2 emissions is negative in lower, lower-middle and upper-middle income countries in Sub-Saharan Africa. However, the effect is statistically significant only in the lower-middle income countries. Our findings suggest that policies designed for curbing informality in SSA countries are likely to have a negative impact on the environment through increasing emissions of carbon dioxide.

Thus, such policies should be accompanied by appropriates measures to facilitate the transition and the inclusion of formalized enterprises in green-based activities. Some of these measures include: fiscal incentives in the form of green-subsidies for new technologies adoption in the formalized enterprises; Increase public awareness and raise the public interest on the consequence of their activities on the environmental quality; Enforcement of environmental regulation in the formalized enterprises.

However, this study concentrates only on the effect of the underground sector on carbon dioxide emissions as environmental indicator, though it stills possible to investigate its effect on other environmental variables such as water pollution, deforestation and municipal solid waste. In addition, because of differences in the composition of the shadow sector in African economies, if the relevant data exist, a possible extension of this work could be the investigation of the effects of such differences on the level of CO2 emissions. All these might provide new and more interesting findings and more specific policy recommendations.