Policy Research Working Paper 10158 Informal Emissions Constantin Burgi Shoghik Hovhannisyan Santosh Ram Joshi Ahmad F. Alkhuzam International Finance Corporation August 2022 Policy Research Working Paper 10158 Abstract Environmental regulations and their enforcement play a added in the informal sector are higher as opposed to in critical role in reducing emissions and their devastating the formal sector. At the sector level, higher informality is effects on humanity and the environment. However, many associated with lower CO2 emissions per value added only developing countries have large informal sectors—account- in manufacturing and other services sectors. In particular, ing for more than 70 percent of total employment, that a one percentage point increase in the share of informal operate outside government control. The presence of the workers in total sector employment reduces the CO2 emis- informal sector could have detrimental consequences on sions per value added by 1.44 percent in manufacturing and the environment as informal firms do not comply with 1.773 percent in services. This implies that the magnitude of regulations, which could jeopardize the effectiveness of emissions per value added in the formal sector relative to the environmental policies. The paper uses reduced form equa- informal sector is ambiguous. Sector-specific estimations tions to estimate the relationship between both CO2 and for non-CO2 emissions yield positive significant coefficients non-CO2 emissions per value added and the informal sector for agriculture, trade, mining, and utilities and a negative measured as the share of informal workers in total across significant coefficient for manufacturing. countries. The estimates indicate that emissions per value This paper is a product of the Development Impact Measurement Department, International Finance Corporation. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/ prwp. The authors may be contacted at shovhannisyan@ifc.org, corsbu@gmail.com, sjoshi5@ifc.org; and aalkhuzam@ifc.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Informal Emissions 1 Constantin Burgi2 Shoghik Hovhannisyan3 Santosh Ram Joshi4 Ahmad F. Alkhuzam5 Keywords: Emissions Accounting, Informal Work JEL: F64, J46, O13, O17 1 The team would like to thank Camilo Mondragon Velez from the International Finance Corporation for his comments, guidance, and support throughout the development of this paper. The paper also benefited from valuable comments by Philip Schellekens, Nancy Lozano Gracia, and Justice Tei Mensah as well as participants of the International Book Workshop on Informal Economy and Environment co-organized by Center for Sustainability Research, Stockholm School of Economics and University of Reading. The views expressed in this paper are those of the authors and do not necessarily represent those of IFC or the World Bank Group. 2 University College Dublin, corsbu@gmail.com 3 International Finance Corporation, shovhannisyan@ifc.org 4 International Finance Corporation, sjoshi5@ifc.org 5 International Finance Corporation, aalkhuzam@ifc.org 1. Introduction The world continues to struggle with high levels of pollution caused by human activities such as fossil fuel consumption, industrial production, and depletion of forests. These produce excessive greenhouse gasses heating the planet and have devastating effects on the environment and humanity especially in developing countries. In fact, the developing countries are responsible for 63 percent of current carbon emissions globally (Center for Global Development 2015). 6 The urgency of addressing the threat of global warming has led countries to combine their efforts to combat climate change and adopt a legally binding international treaty, the Paris Agreement. 7 This became effective in 2020 and aims at limiting global warming to well below 2 (preferably 1.5) degrees Celsius compared to pre-industrial levels. In this regard, environmental regulations and their enforcements have become a critical mechanism for reducing emissions and their detrimental effects on nature and humans as widely discussed in the literature. For example, Danish et al. (2019) find that the current environmental regulations are successful in achieving pollution reduction targets in Brazil, the Russian Federation, India, China, and South Africa and are also responsible for an inverted U-shaped relationship between income and pollution. Similarly, using yearly data for 17 European Union countries over the last two decades, Neves et al. (2020) show that environmental regulations are effective in reducing CO2 emissions in the long-run, while policies supporting renewable energy sources tend to decrease CO2 emissions in both the short- and long-term. Many developing countries have sizable informal sectors operating outside government control: Schneider et al. (2010) show that from 1999 to 2006 the informal sector GDP accounted for more than 50 percent of total in Bolivia, Georgia, Guatemala, Panama, Peru, Tanzania, and Ukraine. According to the World Bank (2021), in emerging markets and developing economies the informal sector contributes about a third of GDP and more than 70 percent of employment of which self-employment accounts for more than half (Ohnsorge et al., 2021). The presence of large informal sectors in developing countries might incur significant risks to the planet, as according to Baksi et al. (2010) informal sectors pose a serious challenge to the enforcement of environmental regulations in these countries. Biswas et al. (2012) argue that the environmental impact of informal sectors can be very significant as they include many pollution intensive activities, such as leather tanning, brick making, metal working, resource extraction, urban transportation with old and inefficient vehicles and production in small scale or family-based factories. Moreover, many developing countries lag behind advanced economies in terms of environmental policies: 6 https://www.cgdev.org/ 7 https://unfccc.int/process-and-meetings/the-paris-agreement/the-paris-agreement 2 the Environmental Performance Index 2020 (Yale Center for Environment Law and Policy, 2020) finds that good policy results are associated with wealth (GDP per capita). Poor environmental regulations combined with large informal sectors might further exacerbate the challenges related to emissions. At the same time, informal firms are excluded from public incentives to reduce emissions such as subsidies to adopt energy efficiency and their workers lack opportunities to access training and skills development programs (International Labor Organization, 2018a). Therefore, understanding the relationship between informality and emissions across countries is critical for informing environmental policies in various countries that aim at reducing emissions both locally and globally. The paper studies the relationship between the informal economy and CO2 and non-CO2 emissions in developing countries using informal employment as a proxy for the informal economy and emissions per value added. This measure of the informal economy is constructed using data collected through labor force surveys and is available for a majority of developing countries. However, it deviates from the literature, which predominantly uses the share of GDP contributed by the informal sector when studying the relationship between informality and emissions. The share of informal workers in total is data-driven as opposed to measurement of the informal sector using GDP, which is based on various estimates offered in the literature and is thus greatly dependent on the chosen estimation approach and assumptions (e.g. see Schneider (2005), Elgin and Oztunali (2014), or Orsi et al. (2012)). In nearly all developing countries, micro surveys – household and labor force surveys – include questions related to informality such as whether a worker has a contract or employment benefits, belongs to a union, contributes social security payments, or receives renumeration. However, these informality-related questions vary considerably across developing countries. To ensure consistency in the measurement of informality across countries, this paper follows Bürgi et al. (2020) and Loayza et al. (2011) and considers as “informal” only those who are self-employed and non-paid workers. While some surveys include additional information that could be used to determine formality, this would reduce the comparability and consistency of how informality is measured across countries. Finally, it chooses to avoid a comparison over time since, according to the data, the informal sector’s share of the economy remains remarkably stable over time. Informality in developing countries can impact emissions in two main ways. On one hand, informal firms in developing countries include mostly self-employed workers or small firms with very low capital, hence their contribution to countries’ emissions could be very low. In contrast, formal firms - which include large firms - are to a great extent responsible for emissions as they rely on capital-intensive technologies that burn fossil fuel. On the other hand, informal firms do not necessarily comply with government regulations 3 and can have practices that are detrimental to the environment as opposed to formal firms which presumably adhere to environmental regulations (e.g. see Basbay et al (2016) or Biswas et al (2012)). To study these effects, the paper builds a theoretical model wherein the economy consists of formal and informal workers and the impact on countries’ total emissions depends on the difference in emissions per worker between the formal and informal sectors and the share of informal workers in total. The literature analyzing the relationship between the informal sector and emissions is very scarce and is limited only to a few papers. For example, Elgin and Oztunali (2014) use cross-country panel data to analyze this relationship by applying fixed effects regressions and system estimations. They find an inverted-U relationship between the size of the informal economy and environmental pollution: small and large sizes of the informal economy are associated with lower environmental pollution and medium levels of informality are associated with higher levels of environmental pollution. This inverted-U shape is also called the environmental Kuznets curve (e.g. see Cole et al. (1997), Dinda (2004), or Stern (2004)). Estimating the relationship between the environmental quality and informality for a sample of 58 countries, Swain et al. (2020) find that the informal sector measured as a share of value added positively influences the emission of local pollutants but has no impact on global pollutants such as CO2 per capita and CO2 intensity. The contribution of our study to the literature is three-fold. First, it adds to the literature by studying the impact of informality on emissions across developing countries using more recent data and helps understand to what extent the differences in production technologies and compliance with the environmental policies drive the differences in the emissions. Second, it uses a new measurement of informal economy – the share of informal workers in total employment which is based on data and not on estimates which are sensitive to approaches and assumptions that could potentially bias the results. Finally, the paper studies the impact of informal employment on emissions using country- and sector-level data and estimates the aforementioned relationship on the sector level. This is important given the considerable differences between sectors in terms of their productivity, capital intensity, and competitiveness, which might affect the aggregate country-level results. The estimation results using country and sector fixed effects show that there is no significant impact from informality on either CO2 or non-CO2 emissions. This implies that emissions per value added in the informal sector are greater than in the formal sector as shown in the theoretical model. 8 Conducting these regressions separately for nine sectors - agriculture, manufacturing, trade, mining, construction, finance, 8 The model uses a decomposition approach to derive the relationship between the estimated coefficients and informal emissions per value added. 4 other services, transportation and communication, and utilities - yields a negative correlation between informality and CO2 emissions only in the manufacturing and other services sectors. In particular, a one percentage point increase in the share of informal workers in total employment reduces emissions per value added by 1.44 percent in manufacturing and 1.773 percent in services. These findings on sectoral differences are also consistent with the environmental Kuznets curve widely discussed in the literature in relation to income and emissions. As countries progressing along their development path transition from agriculture to a manufacturing-dominant economy, CO2 emissions increase, given the capital-intensive production processes typical in manufacturing; a subsequent shift away from manufacturing to a service- dominated economy reduces CO2 emissions. Changes in informality make the inverse-U-shaped relationship more pronounced compared to the scenario with a stable informal sector as manufacturing in the formal sector generates higher emissions than manufacturing in the informal sector. Sector-level regressions for non-CO2 emissions indicate that informality and emissions have positive significant correlations for agriculture, trade, mining, and utilities and a negative significant relationship in manufacturing. 2. Model The model starts with the following identity for an industry s in country c: = + (1) where are the total emissions, are the informal sector emissions and are the formal sector emissions in country c and sector s. Next, both sides of equation (1) are divided by the value added - , to obtain the following: = + (2) Multiplying and dividing each part of the equation by the respective GDP, one can obtain the following: = + = + (1 − ) (3) where is the share of the informal sector in the economy. Intuitively, the emissions per unit of value added in the overall economy are equal to the weighted average of the emissions per unit of value added 5 in the informal and formal sectors of the economy. The weights are the respective shares in total value added that each sector has. For a given level of emissions per unit of value added, one can take the derivative with respect to as shown below: = − (4) This equation shows that the impact of informality on emissions per unit of value added in the overall economy crucially depends on whether the informal or formal sector has more emissions per unit of value added. Two economies with the same overall emissions per unit of value added hence might react differently to a small change in informality, depending on how polluting each sector is. Elgin et al. (2014), for example, explicitly assume that the informal sector has larger emissions per unit of value added, but the above equation allows for the reverse to be true as well. While one can make the argument that formal sector firms have to satisfy stricter environmental regulations, the additional resources such as capital-intensive technologies available to formal sector firms might lead to larger pollution. For example, in the agriculture sector, informal workers might initially burn some forests to gain access to farmland, but the formal sector is more likely to use fertilizers and heavy farm equipment that cause larger emissions. Based on this, it is not necessarily a foregone conclusion that emissions per unit of value added might be larger for the informal sector. One issue with the above equation is that reliable data on informal employment - , are more readily available than the data on informal value added. Instead of multiplying and dividing by formal and informal value added, one can multiply and divide by formal and informal employment and overall employment instead as shown below: = + (5) If is denoted as a share of informally employed workers and = = is the employment per unit of value added equation (5) can be rewritten as follows: = + (1 − ) (6) 6 For a given level of emissions per worker and the number of workers per unit of value added, one can take the derivative with respect to : = � − � (7) As with the informal value added share, the sign of the derivative with respect to the informal employment share is ambiguous. In this case, however, the sign depends on whether emissions per worker in the formal or informal sector are larger. While enters the derivative, it is generally positive and hence does not impact the sign of the derivative. Two similar economies might react differently to changes in informality, depending on their emissions per worker in the two sectors. Comparing the impact of the informal employment share in Equation (7) to the informal value added share in Equation (4), one can distinguish three cases, as value added per worker can reasonably be assumed to be larger in the formal sector: 9 1. − ≤ 0 and − < 0. This means that emissions per unit of value added and per worker in the formal sector are larger than in the informal sector or put differently, the formal sector causes a lot more pollution than the informal sector. Also, there is no difference between the sign of the derivative with respect to the informal employment share and the sign of the derivative with respect to the informal value added share. As the value added per worker is larger for the formal sector, each formal worker produces more units of emissions than informal workers. 2. − > 0 and − ≤ 0. In this case, emissions per unit of value added in the formal sector are smaller than in the informal sector. This is the intermediate case: while the formal sector pollutes less than the informal sector per unit of value added, it is also much more productive and the pollution per worker is larger. At the same time, value added per worker in the formal sector is sufficiently high that the emissions per worker are higher than in the informal sector. 9 The potential fourth case of higher emissions per unit of formal GDP and lower emissions for formal employment can thus be excluded. As each worker produces more output in the formal sector than in the informal one, higher emissions per unit of output imply higher emissions per worker in the formal sector. 7 3. − > 0 and − > 0. This means that emissions per unit of value added in the formal sector are smaller than in the informal sector and that the difference in value added per worker between the formal and informal sectors is sufficiently small. Put differently, this expresses the situation where the informal sector causes a lot more pollution than the formal sector. This way, the higher value added per worker generated in the formal sector is not sufficient to generate emissions that are larger than the emissions per worker in the informal sector. These three cases show that the inferences made about the relationship between the informality share based on employment and emissions per unit of value added cannot always provide information about the relationship of emissions and informality based on value added. For example, if a negative relationship between the informality share based on employment and the emissions per unit of value added is found, this can constitute either case 1 or case 2 and, hence, one cannot infer anything about the relationship between emissions per unit of value added and the informality share based on value added. However, if the relationship for informal employment is positive, one can infer that the relationship for informal value added is also positive (case 3). Conversely, a negative relationship between the emissions per unit of value added and the informal value added share also implies a negative relationship for informal employment (case 1). The above equation can be estimated in a regression and the data allow to test this across countries and industries. 3. Empirical Approach The objective of this study is to estimate the impact of the informal economy, which is a prevalent feature of developing countries, on CO2 and non-CO2 emissions. It regresses the emissions per value added on the share of informal workers in total employment using cross-country and industry data where the industries include agriculture, manufacturing, trade, mining, construction, finance, other services, transportation and communication, and utilities. The choice of these broad sectors is determined by the availability of data on the industry value added and informal employment. This econometric approach estimates the impact of the share of informal employment on emissions per unit of value added using fixed-effects regressions as shown in Equation (8): ln � � = + + + (8) 8 where ln � � is the natural logarithm of emissions per GDP for industry s and country c, is the share of informal employment in that industry and country and are the control variables. Given the potentially large differences across industries, Equation (8) is also estimated separately as a cross-section across countries for a given industry. As mentioned in the model section, > 0 implies that emissions per worker are larger in the informal sector than in the formal sector and < 0 implies the opposite. If = 0 cannot be rejected, this implies that the emissions per worker are the same in both sectors. As the formal sector might produce very different goods from the informal sector, it cannot be determined whether the production of a specific good generates more emissions in either sector. Also, only if it is found that > 0 , there is a direct implication for the informality based on GDP. 4. Data and Descriptive Statistics The study uses data from two different sources: the International Income Distribution Data Set (I2D2) for informal employment shares and the Global Trade Analysis Project (GTAP) for other sector indicators: value added, capital per worker, share of value added in total GDP, and the share of imported intermediate goods in the total intermediate goods (Aguiar et al., 2019). The I2D2 data set compiled by the World Bank is a global harmonized labor force survey database that provides comparable labor market indicators such as employment status, wages, and hours worked at both household and individual levels across countries and time. It contains 600 surveys for 120 countries collected at different points of time. For many countries, especially low-income and fragile or conflict-affected states, it has data only for a single year. The GTAP 10 database combines detailed data on bilateral trade, transport, and protection characterizing economic linkages among regions, together with individual country input-output (I-O) databases which account for inter-sectoral linkages within regions for the most recent year available (2014). The I2D2 database is used to compute informal employment shares in total employment in each of nine sectors, comprising agriculture, mining, manufacturing, transport and communication, finance, construction, trade, utilities, and other services. According to the operational definition used by the International Labor Organization, informal workers include all (i) contributing family workers, (ii) informal employers, own-account workers, members of cooperatives if they own an informal economic unit, and (iii) employees whose employer has no contribution to social security and who are not entitled to and do 9 not effectively receive benefits from paid annual leave (or compensation instead of it) or paid sick leave (ILO, 2018b). Despite a clear definition of informality, its measurement varies substantially across developing countries and depends on the availability of informality-related indicators in labor force surveys. To ensure the comparability of informality measurement across countries, this study considers workers informal if they are self-employed or are non-paid family workers. Similar measures of informal employment have been applied in previous studies: Maloney (2001), and Loayza et al. (2011) used self- employment as a proxy for informal employment when studying the relationship between informality and labor productivity. Burgi et al. (forthcoming) further improve informal employment measurement by including non-paid workers and claim that non-paid workers can account for more than 60 percent of total employment in low -income countries; including them hence substantially improves the informality measurement. For example, a 2017 labor force survey for Bolivia shows that self-employed and non-paid workers accounted for 88 percent of informal employment. 10 The sample of estimation includes 7 high- income countries, 9 low-income countries, 27 lower-middle income countries, and 26 upper-middle income countries. It is important to note that the I2D2 data set includes one or more surveys per country, but each country might have data for different years. In many developing countries the labor force surveys are conducted only once every few years because they are costly, while in others the data are collected on a quarterly basis. The study uses data for the most recent available year for individual countries in I2D2 where the database might have data available for multiple years. 11 This is because the data analysis for countries that have more than one year’s data available in I2D2 shows that informality measured as a share of informal workers in total does not change much over short periods of time, with any changes visible over the long term. Furthermore, according to the World Bank (2019) informality has remained remarkably stable despite economic growth and changes in the nature of work. For example, in Peru informality has been constant at about 75 percent over the last 30 years. In Sub-Saharan Africa, informality remained, on average, at around 75 percent of total employment from 2000 to 2016 and, in South Asia, it increased from an average of 50 percent in the 2000s to 60 percent over the period 2010-16. Hence, cross-section regressions are more relevant estimation methods here given the low variation in informal employment over time compared to time series or panel data methods. 10 Authors’ calculations. 11 See Table A.1 in the Annex for the exact data coverage. For over 80 percent of the countries considered, the latest data is between 2011 and 2017. 10 There is a substantial variation in the total share of informal workers across countries and sectors. According to the World Bank Group (2021), the scale of the informal economy varies substantially among emerging markets, developing economies, and their regions. For example, in 2018, the share contributed by the informal sector ranged from around 10 percent to 68 percent of GDP, while informal employment measured in terms of self-employment ranged from near-zero to 96 percent of total employment. In particular, the highest share of informality is observed in some countries in the Sub-Saharan Africa region, in which informal employment constitutes more than 90 percent of total employment. There is also a considerable variation across industries: among the nine industries in this study, agriculture, trade, and construction have on average the highest levels of informality accounting respectively for 68, 50 and 34 percent of total employment. The lowest informality rates on average are observed in utilities (12 percent), finance (16 percent), and other services (20 percent). Also, due to its capital-intensive production the mining sector employs very few workers and accounts for less than 5 percent of total employment in many countries. Hence, the number of informal workers in this industry is insignificant in the broader picture of total informal employment. Other variables used in the regressions are the capital-labor ratio, share of sectoral value added in total GDP, share of imported intermediate goods in total intermediate goods, and emissions from intermediate usage of imports and domestic products in a metric of carbon dioxide equivalent. The latter is a dependent variable in the regressions which is divided by the value added to take into account the size of the sectors. All these indicators are obtained from the GTAP data set which provides data for 65 sectors and 141 countries and world regions for the year 2014. Country mapping between the I2D2 and GTAP data shows that only 70 developing countries are represented in both databases and thus are included in the sample for estimations. Among the GTAP indicators, the capital-labor ratio is computed by dividing the payment to capital by the payments to labor and emissions include the carbon dioxide (CO2) (mega metric). This CO2 emission data comprises carbon emissions from fossil fuel combustion by sectors. 12 Finally, the GTAP data are aggregated into nine sectors to combine with the I2D2 data set, omitting the public administration and defense sector which is assumed to have no informal employment. 12 Detailed description of CO2 emission data calculations can be found in Lee, H.L. (2008), “The Combustion-based CO2 Emissions Data from GTAP Version 7 Data Base,” Center for Global Trade Analysis, Purdue University: West Lafayette. 11 Figure 1. Sectoral CO2 and non-CO2 Emissions per Value Added across Sectors (metric/million USD). 4,000 20,000 Non-CO2 Emissions per VA CO2 Emissions per VA High Income 15,000 2,000 Lower Middle Income 10,000 Low Upper Middle 5,000 - - Source: Authors’ computations using GTAP data. Utilities are removed from Figure 1 for CO2 emissions for presentational purpose, as they have nearly five times higher emissions per value added (on average) compared to transportation and communication. Figure 1 clearly shows that transportation and communication, mining, and manufacturing are among the highest contributors to CO2 emissions per value added, while other services, finance, and trade have the lowest levels of CO2 emissions per value added. It is important to note that utilities are intentionally removed from the graph as they have nearly five times higher emissions per value added (on average) compared to transportation and communication. There is also a large dispersion within these broad sectors indicating that production structures within industries could be diverse differing among developing countries relative to the different stages of sophistication and the choice of technology. At the same time, the CO2 emissions per value added vary substantially across the different country income groups. Finally, non-CO2 emissions in terms of CO2 equivalent are multiple times higher for agriculture, mining, and utilities with a large variation across income groups. These non-CO2 emissions include methane (CH4), nitrous oxide (N2O) and the group of fluorinated gases (F-gases). 5. Estimation Results The paper estimates the impact of the share of informal workers on emissions per value added using cross country and industry data in the following regressions: OLS, OLS with control variables, industry fixed effects, country fixed effects, and industry and country fixed effects as shown in Table 1. A problem in applying OLS is that regressors can be correlated with the error term and this correlation between regressors and the errors violates an assumption necessary for the consistency of OLS. For example, 12 countries with poor governance systems might have high levels of pollution because of a lack of proper environmental regulations exacerbated with low compliance and, at the same time, have sizable informal sectors. One way to work around this endogeneity is to apply country and industry fixed effects as the data set has both country- and industry-level observations. It is important to acknowledge that estimations in this paper do not claim causal effects despite improving OLS coefficients by using country and industry fixed effects but only discuss correlations between emissions and informality. All these estimations are also conducted using country and sector weights, the results are not reported here as significance and sign of the coefficients remain the same. While regressions with country fixed effects have been used in other studies and produced similar results, this paper adds to the literature by including industry fixed effects and conducting industry-specific estimations. The regressions include several control variables to take into consideration widely documented differences in capital intensity and production technologies across countries and sectors as well as to account for the relative importance of the sectors in the overall economy. At the same time, the choice of control variables is highly driven by the availability of data on a country and sector level, and the selected variables are extracted only from two data sets - GTAP and I2D2 - given the paucity of sector data in general. More specifically, the control variables include the natural logarithm of capital per worker, the share the respective industry value added contributing to the country’s total value added, and the share of imported intermediate goods in total consumption of intermediate goods in each sector. The capital intensity and share of imported intermediate goods in total aim at capturing the underlying differences in production technologies and the extent of the industry’s reliance on global value chains which can be indicative of their competitiveness and productivity. On average, capital-intensive sectors consume more electricity and, hence, are more likely to have high emissions as opposed to sectors with low capital. At the same time, sectors that rely more on imported intermediate goods might be more productive and competitive compared to those that have strong backward linkages with domestic industries only or are self-sufficient in their production processes. Hence, these sectors again might utilize more sophisticated technologies responsible for higher levels of pollution. Finally, the share of an industry’s value added in the country’s total value added accounts for the importance of the respective industry in the economy. The first set of results for CO2 emissions per value added illustrates that there is no difference between emissions per formal workers and per informal workers based on the preferred approach of using both country and industry fixed effects in the regressions. The OLS regression in the first column of Table 1 shows a negative and significant coefficient for the impact of informal employment on CO2 emissions per 13 value added. Indeed, a one percentage point increase in the share of informal workers leads to a 1.4 percent decrease in emissions per unit of value added. Once control variables and industry fixed effects are added to the regressions (columns 2 and 3 of Table 1), the coefficients become smaller but still remain negative and significant. However, when using both country and industry fixed effects (final column in Table 1), the impact of informal employment on emissions becomes insignificant implying that formal sector employees and informal sector workers produce roughly the same amount of CO2 emissions. In turn, this suggests that the formal sector has lower emissions per unit of value added than the informal sector based on the case 2 scenario in the model section as the value added per worker in the formal sector is sufficiently high to reduce its emissions per value added compared to the informal sector. The insignificant coefficient is also in line with the estimates by Basbay et al. (2016), Biswas et al. (2012), and Elgin et al. (2014), who found a positive coefficient for a regression using estimates of informal GDP. Table 1. Regression Results for CO2 Emissions per Output. OLS OLS with Industry Country Industry and No Controls Controls Fixed Effects Fixed Country Fixed Effects Effects -1.406*** -1.075*** -0.640** -0.770* 0.319 Share of Informal Workers (-3.69) (-3.14) (-2.39) (-1.88) (0.92) 0.080 -0.042 0.133 -0.075 Log of Capital per Worker (0.90) (0.69) (1.31) (1.13) Share of Sectoral -6.018*** -7.034*** -5.204*** -6.686*** VA in total VA (5.96) (8.57) (5.18) (8.96) Imported Intermediate 5.000*** 1.568*** 7.437*** 1.635*** Goods-Total Intermediate (9.94) (4.55) (11.30) (3.73) Goods Number of Observations 559 557 557 557 557 Each column shows the estimation results for separate regressions. OLS no Controls and OLS with Controls provide estimated coefficients for OLS regressions with and without control variables which are natural logarithm of capital per worker, share of the respective industry value added in countries’ GDP, and share of imported intermediate goods in total intermediate goods. The Industry and Country Fixed Effects columns add either industry or country fixed effects to the regressions, and the Industry and Country Fixed Effects column includes both industry and country fixed effects. Meanwhile t statistics are in parentheses. * shows significance at the 10% level, ** at the 5% level and *** at the 1% level. 14 The sign and significance of the coefficients for control variables remain unchanged across the various estimation methods. Capital per worker does not appear to have an impact on emissions, while the size of an industry’s value added relative to the total negatively correlates with emissions per unit of value added. The coefficient on capital per worker is somewhat puzzling given that capital-intensive technologies produce more emissions as they consume more power. The negative relationship between emissions per value added and size of the industry is more intuitive when studying each industry case. In the case of agriculture, for instance, the sector’s higher share is associated with low emissions as, in general, agriculture’s contribution to CO2 emissions is very low. For capital-intensive sectors such as manufacturing, the dominance of the sector in the GDP also reflects the high development level of the entire economy which, in turn, implies greater compliance with environmental regulations and low emissions. Finally, the coefficient for imported intermediate goods in total is positive and significant across all regressions with varying magnitudes indicating that a larger share of imported intermediate goods is associated with higher emissions. More research needs to be conducted to better understand the behavior of these variables and their impact on emissions. Next, the regressions are repeated at an industry level for nine industries: agriculture, manufacturing, trade, construction, other services, mining, finance, transport and communication, and utilities (Table 2). Among these sectors only manufacturing and other services exhibit negative and significant coefficients, while other sectors have insignificant results found in country-level regressions. In particular, a one percentage point increase in the share of informal workers in total industry employment reduces the CO2 emissions per value added by 1.44 percent in manufacturing (Table 2, column 5) and 1.773 percent in other services (Table 2, column 7). Manufacturing is one of the sectors with the highest CO2 emissions, ranking fourth after transport and communication, mining, and utilities (Figure 1). Its high level of emissions is primarily driven by its heavy reliance on capital intensive technologies compared to other sectors. Hence, the existence of an informal sector with very low or no capital in developing countries leads to lower CO2 emissions per worker in manufacturing. In fact, according to the World Bank (2019) in many developing countries a large number of workers remain in low-productivity jobs, often in informal sector firms whose access to technology is limited affecting their productivity. Kanbur (2019) shows that in emerging economies informal workers’ productivity constitutes, on average, only 15 percent of the productivity of formal workers. For example, according to Benjamin and Mbaye (2012), in Dakar city in Senegal 87 percent of firms with labor productivity below US$10,000 per worker are in the informal sector. In general, informal firms use little capital, supply goods and services to low-income consumers, and have owners with low levels of education, who are unlikely to move to the formal sector (La Porta 15 and Shleifer, 2014). Contrary to the situation in manufacturing, other services utilize mostly labor- intensive technologies and have very low emissions compared to other sectors (Figure 1). The existence of a large informal sector further reduces the use of capital, resulting in a negative relationship between emissions and informality. Relative to the case with an insignificant coefficient observed in aggregate- level regressions, a negative correlation implies that emissions per unit of value added in the formal sector relative to the informal sector can be either higher or lower depending on the differences in their labor productivities. While there is some variation in the control variables as well, the significant ones are in line with those recorded in Table 1. Table 2. Sectoral Regression Results for CO2 Emissions per Output. (1) (2) (3) (4) (5) (6) (7) (8) (9) Agriculture Trade Construc- Finance Manufac- Mining Services Transport& Utilities tion turing Communication Share of 0.599 -0.599 -1.044 -1.838 -1.440** -0.291 -1.77** 0.386 -1.253 Informal (0.66) (0.75) (1.51) (1.02) (2.45) (0.41) (2.24) (1.33) (0.84) Workers Log of -0.187 0.107 -0.231** 0.004 -0.075 -0.268 0.27 0.036 -0.236 Capital (0.44) (0.63) (2.03) (0.02) (0.43) (1.67) (1.62) (0.38) (0.77) per Worker Share -11.11*** -4.75 -5.39* -21.24** -3.54** -3.41 -6.73*** -4.92** -21.5** of Sectoral (5.16) (1.43) (1.69) (2.36) (-2.25) (1.46) (3.69) (2.51) (2.41) VA in total VA Imported 0.529 3.291** 2.204*** 2.804 -0.634 1.605* 2.39** 1.336*** 0.354 Intermediate (0.25) (2.37) (2.90) (1.32) (0.60) (1.73) (2.22) (4.64) (0.33) Goods/Total Number of 57 68 69 52 69 55 63 68 56 Observations Each column shows the OLS estimation results for separate regressions and different sectors. These regressions include the following control variables: natural logarithm of capital per worker, share of the respective industry value added in countries’ total value added, and share of imported intermediate goods in total intermediate goods. Meanwhile t statistics are in parentheses are in brackets. * shows significance at the 10% level, ** at the 5% level and *** at the 1% level. The findings on sectoral differences can also help explain the environmental Kuznets curve widely discussed in the literature in relation to income and emissions. As shown in Grossman et al. (1991), the transition from an agricultural economy to a manufacturing economy to a service economy can cause an inverse-U shape in emissions in relation to development. An agricultural economy in the early stages of development has low levels of CO2 emissions, which increase as the economy grows and shifts towards manufacturing. With a further transition into a developed economy, the manufacturing sector declines, 16 and the low-emission service sector expands, leading to an overall decline in emissions again. The sectoral differences estimated here add a new component to this transition as informality makes the environmental Kuznets curve more pronounced. Low-income countries have high informality and a low share of manufacturing and services in the total GDP. As they grow, the high capital-intensity formal sector in manufacturing expands, leading to a substantial increase in emissions due to the growth of the manufacturing sector and the decline in informality. If there is no shift in informality, the emissions would increase less, as the formal manufacturing sector has higher emissions than the informal manufacturing sector. With further growth, there is a shift away from manufacturing towards services, and a decline in informal employment, which leads to a broad reduction in emissions. This is not due just to the shift away from emissions-heavy manufacturing, but also due to the decline in informal employment across all other industries. As there is no difference in emissions per worker between the formal and informal sectors in most service sectors, the emissions per unit of value added are lower in the formal sector in those industries (see case 2 in the model section). Due to this, the industry shift in the value added alone would lead to higher emissions than the industry shift in value added together with the informality shift. Finally, these findings are also relevant for the case where countries experience premature deindustrialization. In the absence of growth in the manufacturing sector, the labor surplus is absorbed by informal urban services thus expanding the services sector as the economy transitions from having a dominant agriculture sector to a prevailing service sector, which is the case in many developing countries. In this scenario, the negative environmental effects would be dampened compared to the traditional growth path, albeit at a huge development cost in terms of productivity losses. Table 3 shows regressions results for non-CO2 emissions in terms of CO2 equivalent for OLS without and with control variables, as well as fixed effects regressions with country, industry, and country and industry fixed effects. For all specifications but country and industry fixed effects (Table 3, final column) the relationship between the share of informal employment in total and emissions per value added is significant and positive. However, the preferred specification which accounts for both country and industry unobserved characteristics yields an insignificant coefficient indicating that emissions per informal value added are higher relative to emissions per formal value added across developing countries, as discussed in case 2 of the model. For non-CO2 emissions, the capital per workers matters in three of the four specifications, unlike for CO2 emissions. The other control variables have the same sign and significance for both CO2 and non-CO2 emissions. 17 Table 3. Regression Results for non-CO2 Emissions per Output. OLS OLS with Industry Country Industry and No Controls Controls Fixed Effects Fixed Country Fixed Effects Effects 2.172*** 2.247*** 1.360*** 1.397** -0.657 Share of Informal Workers (4.31) (4.53) (3.90) (2.30) (1.48) -0.302** -0.138* -0.397** -0.126 Log of Capital per Worker (2.22) (1.65) (2.50) (1.42) Share of Sectoral -5.173*** -4.325*** -5.164*** -4.956*** VA in total VA (3.53) (4.13) (3.40) (5.37) Imported Intermediate 3.980*** 2.022*** 4.121*** 1.357** Goods in Total (5.21) (4.29) (4.03) (2.36) Number of Observations 517 515 515 515 515 Each column shows the estimation results for separate regressions. OLS no Controls and OLS with Controls provide estimated coefficients for OLS regressions with and without control variables, which are natural logarithm of capital per worker, share of the respective industry value added in countries’ GDP, and share of imported intermediate goods in total intermediate goods. Industry and Country Fixed Effects columns add either industry or country fixed effects to the regressions and Industry and Country Fixed Effects column includes both industry and country fixed effects. Meanwhile t statistics are in parentheses. * shows significance at the 10% level, ** at the 5% level and *** at the 1% level. The sectoral regressions for non-CO2 emissions produce positive and significant coefficients between the informality and emissions per value added across all sectors with the exception of construction, finance, other services, and transport and communication which yield insignificant results (Table 4). These results vary substantially from the CO2 estimates, which produce negative coefficients for manufacturing and other services only, and further research is needed to better understand the underlying causes. While for some sectors such as agriculture, mining, and utilities, the literature provides intuitive explanations, for other sectors such as finance more research needs to be done to better understand and interpret the results. For example, a positive relationship between informal employment and emissions in agriculture - which also implies a positive relationship between the share of informal value added and emissions per value added - is plausible, as agriculture is the largest source of non-CO2 emissions in the world (mainly methane and N2O), and informal sectors are more likely to emit more because of non-compliance with environmental standards and regulations. In fact, about one-half of global non-CO2 emissions are contributed by the agriculture sectors via the livestock population, fertilizer consumption, and crop production (USEPA, 2019). Within the agriculture sector, livestock accounts for more than half of non-CO2 18 emission because of CH4 from enteric fermentation and manure management. At the same time, land management activities in crop production are responsible for most of the remaining non-CO2 pollution emitted as N2O. Agriculture in developing countries is largely informal and as its share of informal workers increases, it is likely that non-CO2 emissions will increase too due to land use management practices and residues from agricultural production processes. Table 4. Sectoral Regression Results for non-CO2 Emissions per Output. (1) (2) (3) (4) (5) (6) (7) (8) (9) Agricultur Trade Construc- Finance Manufac- Mining Services Transport& Utilities e tion turing Communication Share of 0.894* 3.764** 0.804 -0.255 -4.414*** 4.543** 0.094 -0.214 2.434** Informal (1.98) * (0.64) (0.09) (5.10) * (0.07) (0.33) * Workers (3.41) (4.31) (2.89) Log of 0.337 -0.294 -0.213 -0.060 -0.295 -0.319 0.027 0.260 0.120 Capital (1.58) (1.29) (.0.77) (0.19) (1.12) (1.36) (0.10) (1.22) (0.69) per Worker Share of 0.353 4.341 1.865 -14.622 -6.137** -0.338 -15.8*** -16.96*** -17.5*** Sectoral VA (0.33) (0.96) (0.25) (1.11) (2.61) (0.10) (5.41) (3.85) (3.46) in total VA Imported 1.614 4.889** 6.548*** 5.188 -1.436 -0.728 1.079 2.038*** 0.074 Intermediat (1.48) (2.62) (3.39) (1.40) (0.91) (0.54) (0.63) (3.15) (0.12) e Goods/Total Number of 68 65 26 48 68 54 62 68 56 Observation s Each column shows the OLS estimation results for separate regressions and different sectors. These regressions include the following control variables: natural logarithm of capital per worker, share of the respective industry value added in countries’ total value added, and share of imported intermediate goods in total intermediate goods. Meanwhile t statistics are in parentheses are in brackets. * shows significance at the 10% level, ** at the 5% level and *** at the 1% level. 6. Conclusion The relationship between informality and emissions has not been well studied in the literature despite the existence of large informal sectors in developing countries that operate outside government control and do not comply with regulations including environmental ones. This could to some extent be explained by the lack of data on informal GDP across developing countries as well as by the relatively recent emergence of emissions literature. The informal sectors and their non-compliance with environmental regulations and standards could be a formidable impediment in reducing emissions in developing 19 countries and globally and might substantially undermine the effectiveness of environmental policies. In this regard, studying the relationship between informality and both CO2 and non-CO2 emissions can be critical for understanding whether informal sectors in developing countries are, indeed, a major obstacle in pursuing the climate agenda from the perspectives of international finance institutions and governments in developing countries, as well as for informing environmental policies and their design. In fact, this study directly relates to the United Nations Sustainable Development Goals (SDG) 11 on Sustainable Cities and Communities and SDG 13 on Climate Action. At the same time, it is important to take into account, that in addition to large informal sectors, developing countries also have poor environmental regulations and enforcement compared to more advanced economies, which could further augment these effects. To be able to explain the estimation results, first, it is important to understand the characteristics of informal sectors in developing countries. This paper uses informal employment as a measure of informality and defines it as the share of self-employed and non-paid workers in total employment in each sector and country. While informal employment includes all those (i) who work for the formal sector but are informal as they do not pay taxes and do not receive employment benefits and (ii) who work for informal unregistered firms; the self-employed and non-paid workers account for a large share of total informal workers in many developing countries. The informal workers use no or very low capital in producing goods or providing services especially when compared to formal firms that use capital-intensive production technologies with high levels of CO2 emissions. Hence, the relationship between informality and emissions could be expected to be either positive, if the effects from non-compliance with environmental regulations in countries with large informal sectors prevail, or negative, if the effect of low emissions from using very low capital exceeds the non-compliance effects. Other effects come into play when non-CO2 emissions are considered. Overall, the findings using informal employment shares in developing countries illustrate that emissions per value added in the informal sector are higher than in the formal sector when aggregate regressions are conducted. The sector-level regressions showing that larger informal employment is associated with lower emissions only in manufacturing and other services imply that emissions per value added can be larger or lower in these sectors depending on the differences in labor productivity between the formal and informal firms. Instead, for all other sectors, the non-compliance of informal workers with environmental regulations leads to higher emissions. The latter also holds for non-CO2 emissions when conducting country-level regressions and this positive relationship between informality measured in 20 terms of value added and emissions becomes even stronger when sectoral regressions are considered. Thus, reducing informality or designing environmental policies that reduce non-compliance in the informal sector is critical for tackling climate change across developing countries. 21 References Aguiar, A., Chepeliev, M., Corong, E., McDougall, R., & van der Mensbrugghe, D. (2019). The GTAP data base: Version 10. Journal of Global Economic Analysis 4 (1): 1–27. Baksi, S., & Bose, P. (2010). Environmental regulation in the presence of an informal sector. University of Winnpeg, Department of Economics Paper. Basbay, M. M., Elgin, C., & Torul, O. (2016). Energy consumption and the size of the informal economy. Economics: The Open-Access, Open-Assessment E-Journal 10 (2016–14): 1–28. Blackman, A. (2000). Informal sector pollution control: What policy options do we have? World Development 28 (12): 2067–2082. Benjamin, N., & Mbaye, A. A.. (2012). The informal sector in Francophone Africa: Firm size, productivity, and institutions. With Ibrahima Thione Diop, Stephen S. Golub, Dominique Haughton, and Birahim Bouna Niang. Africa Development Forum Series. Washington, D.C.: Agence Française de Développement and World Bank. Biswas, A. K., Farzanegan, M. R., & Thum, M. (2012). Pollution, shadow economy and corruption: Theory and evidence. Ecological Economics 75: 114–125. Bürgi, C., Hovhannisyan, S., and Mondragon-Velez, C. (Forthcoming), GDP-Employment elasticities across developing countries. Unpublished paper. Cole, M. A., Rayner, A. J., & Bates, J. M. (1997). The environmental Kuznets curve: An empirical analysis. Environment and Development Economics 2 (4): 401–416. Danish, R. U., Ud-Din Khan, S., Awais Baloch, M., & Li, N. (2019). Mitigation pathways toward sustainable development: Is there any trade-off between environmental regulation and carbon emissions reduction? Sustainable Development 28 (4): 813–822. Dinda, S. (2004). Environmental Kuznets curve hypothesis: A survey. Ecological Economics 49 (4): 431– 455. 22 Elgin, C., & Oztunali, O. (2012). Shadow economies around the world: Model based estimates. Bogazici University Department of Economics Working Papers 5, 1–48. Elgin, C., & Oztunali, O. (2014). Pollution and informal economy. Economic Systems 38 (3): 333–349. Yale Center for Environmental Law & Policy. (2020). Environmental Performance Index. Grossman, G. M., & Krueger, A. B. (1991). Environmental impacts of a North American free trade agreement. National Bureau of Economic Research working paper W3914. Cambridge, MA: National Bureau of Economic Research. International Labor Organization. (2018a). World employment social outlook 2018: Greening with jobs. International Labor Organization. (2018b). Women and men in the informal economy: A statistical picture. Third Edition. IPCC. (2018). Summary for policymakers. In Masson-Delmotte, V. et al. (Eds.), Global Warming of 1.5°C: An IPCC Special Report on the impacts of global warming of 1.5°C above pre-industrial levels and related global greenhouse gas emission pathways in the context of strengthening the global response to the threat of climate change, sustainable development, and efforts to eradicate poverty. Geneva, Switzerland: World Meteorological Organization. Kanbur, R. (2017). Informality: Causes, consequences and policy responses. Review of Development Economics 21: 939–61. La Porta, R., and Shleifer, A. (2014). Informality and development. Journal of Economic Perspectives 28 (3): 109–26. Loayza, N. V., & Rigolini, J. (2011). Informal employment: Safety net or growth engine? World Development 39 (9): 1503–1515. Maloney, W. (2001) Self-employment and labor turnover in developing countries: Cross-country evidence. In Devarajan, S., Halsey Rogers, F., & Squire, L. (Eds.), World Bank economists’ forum. Washington, DC: World Bank. 23 Neves, S.A., Marques, A.C., and Patricio, M. (2020). Determinants of CO2 emissions in European Union countries: Does environmental regulation reduce environmental pollution? Economic Analysis and Policy 68: 114–125. Orsi, R., Raggi, D., & Turino, F. (2012). Estimating the size of the underground economy: A DSGE approach. Working Papers wp818, Dipartimento Scienze Economiche, Universita di Bologna. Schneider, F. (2005). Shadow economies around the world: what do we really know?. European Journal of Political Economy 21 (3): 598–642. Stern, D. I. (2004). The rise and fall of the environmental Kuznets curve. World Development 32 (8): 1419–1439. Swain, R. B., Kambhampati, U.S., & Karimu, A. (2020). Regulation, governance and the role of the informal sector in influencing environmental quality? Ecological Economics 173, paper 106649. United Nations Environment Programme. (2019). Global environmental outlook 6. Nairobi: UN Environment. The World Bank. (2019). World development report 2019: The changing nature of work. Washington, D.C.: World Bank. Ohnsorge, F., & Yu, S. (Eds.). (2021). The long shadow of informality: Challenges and policies. Washington, D.C.: World Bank Publications. United States Environmental Protection Agency (2019). Global non-CO2 greenhouse gas emission projections & mitigation 2015–2050. Washington, D.C: US EPA. 24 Annex Table A1. Country List and Years for the I2D2 Data Set Country Year Country Year Albania 2008 Lao PDR 2007 Argentina 2014 Madagascar 2012 Armenia 2016 Malawi 2016 Azerbaijan 2015 Mauritius 2012 Belarus 2016 Mexico 2012 Benin 2015 Mongolia 2014 Bolivia 2015 Morocco 1998 Botswana 2009 Mozambique 2014 Brazil 2015 Nicaragua 2014 Bulgaria 2007 Nigeria 2009 Burkina Faso 2014 Pakistan 2015 Cambodia 2012 Panama 2015 Cameroon 2014 Paraguay 2017 Chile 2017 Peru 2015 China 2013 Philippines 2014 Colombia 2017 Romania 2013 Costa Rica 2015 Rwanda 2013 Côte d'Ivoire 2015 Senegal 2011 Dominican Republic 2015 Slovenia 2004 Ecuador 2015 South Africa 2017 Egypt, Arab Rep. 2005 Sri Lanka 2016 El Salvador 2014 Tajikistan 2013 Ethiopia 2016 Tanzania 2014 Georgia 2013 Thailand 2011 Ghana 2016 Trinidad and Tobago 2000 Guatemala 2014 Tunisia 2011 Guinea 2012 Türkiye 2015 Honduras 2016 Uganda 2016 India 2011 Ukraine 2013 Indonesia 2014 Uruguay 2015 Jamaica 2002 Venezuela, RB 2006 Jordan 2016 Vietnam 2010 Kazakhstan 2013 Zambia 2015 Kenya 2005 Zimbabwe 2011 Kyrgyzstan 2011 25