UNIVERSITY OF GRONINGEN
FACULTY OF ECONOMICS AND BUSINESS
Master of Science in Economic Development and Globalization (MSc ED&G)
Master’s Thesis
THE IMPACT OF DEINDUSTRIALISATION ON INFORMAL EMPLOYMENT IN DEVELOPING COUNTRIES
Submitted by: Sara Alhola S4144295
S.M.Alhola.1@student.rug.nl Supervisor: Prof. Dr. H.H. van Ark
16th of June, 2020 Co-assessor: Dr. A. Minasyan
ABSTRACT
The persistent informal employment in developing countries continues to shape and define the life of billions, while simultaneously puzzling economists and policy makers. This paper studies the trends in informal employment and links them together with industrialisation in Sub-Saharan Africa and Latin America. These regions provide an interesting setting, as the recent decades have shaped their industrialisation paths abnormal from the standard, namely getting them into unfavourable positions in global manufacturing race. By using a sample of 37 countries, I find that contraction in the industry sector employment increases urban informal employment in relation to formal employment. Furthermore, this effect diminishes in the presence of better institutions.
Table of contents
Abstract ... i
List of figures ... iii
List of tables ... iii
Abbreviations ... iii
1. Introduction ... 1
2. Literature review ... 3
2.1. Structural change then and now ... 3
2.2. The role of institutions in structural change processes... 5
2.3. Informal economies ... 6
2.3.1. What is informality? ... 6
2.3.2. What causes informality? ... 8
2.4. Structural change and informality ... 11
2.5. Hypotheses ... 13
3. Data and methodology ... 14
LIST OF FIGURES
Figure 1: Urban self-employment and GDP per capita 1991-2018 ... 22
LIST OF TABLES Table 1: Expected signs of the explanatory variable coefficients ... 19
Table 2: Average annual employment growth decomposition for 2006-2018. ... 23
Table 3: Determinants of informal employment ... 24
Table 4: Determinants of informal employment in different regions ... 26
Table 5: Determinants of informal employment including joint effects ... 29
Table 6: Robustness: Determinants of total informal employment ... 30
Table 7: Robustness: Causality ... 32
ABBREVIATIONS
FE Fixed Effects
GGDC Groningen Growth and Development Centre
GLS Generalised Least Squares
ICLS International Conference of Labour Statisticians ILO International Labour Organization
ISIC International Standard Industrial Classification JoIn Global Jobs Indicators Database
LAC Latin America and Caribbean
MIMIC Multiple Indicators – Multiple Causes
OLS Ordinary Least Squares
RE Random Effects
SSA Sub-Saharan Africa
WGI World Governance Indicators
1. INTRODUCTION
Looking at the growth paths of Latin American and Sub-Saharan African economies, there are two features that make these regions stand out from others and which have shaped their economies during the last decades. Firstly, their paths of structural change are different from those we have learned from the books. Globalisation and heavy trade competition have dropped these regions further from the top of the industrialisation race, as the Asian economies have become the dominant power. Economies in Latin America have stagnated in the middle ground with their manufacturing sectors: the promising growth stopped before reaching peaks high enough, leaving economies in the middle-income trap. Poor countries in Sub-Saharan Africa continue to be dominated by the agricultural sector and the performance in manufacturing has not been good enough to succeed in the global competition. As the economies of these regions have been growing and developing, some faster than others, the losses in industrialisation have been partly compensated by the growing service sector. This is not the kind of structural change that we read about in the history books, which tend to bring up the Industrial Revolution as the starting engine for the development of the West, giving us the comfort of today’s living. The question is, how much do we actually know of the consequences of different structural change paths? What we know for sure is that manufacturing sector has historically been a great tool to absorb labour from the fields, increasing the overall productivity. We do also know that economies around the globe are tilting towards services1, which seem to be providing endless opportunities to earn income and create consumers needs that did not exist before. Even though the route to a services-driven economy has historically gone through industrialisation, today we have economies that bypassed this phase and still became thriving service economies. But there are also one billion people living in Sub-Saharan Africa (World Bank, 2020b), who are neither enjoying the benefits of thriving industrialisation nor seem to be getting out of their low-income status with service sector growth miracles. These trends have not favoured the volatile economies in Latin America either, who keep searching ways to stabilise their growth paths in order to redeem their checks as high-income countries. What is the role of lacking industrialisation in explaining how these economies function today?
The second common characteristic of these regions requires us to look beyond the official GDP statistics and more fundamentally at the structures of the economy. Large part of the economic activity in these two regions, takes place in the shadows of the official economy. Whereas in Western high-income countries, this kind of business is considered strictly illegal thus must include something sketchy, in Latin America and especially in Sub-Saharan Africa, this is a common way to make a living. The informal sector consists of businesses that by their nature could be part of the formal sector, but for a reason or another, are not. What is puzzling is that
the informal sector has persisted, or even grown despite the economic growth of these countries. This is contradictory to a common perception: informality will fade as income grows.
informal employment while deindustrialisation increases it. Second, employment creation in the service sector does not have an effect on the informality of urban employment. Third, the joint effect of institutions and structural change is ambiguous. While better accountability to law does not have an effect on the deindustrialisation-informality nexus, better quality of regulation removes the informality-increasing effect of deindustrialisation. This implies that by improving institutions, countries can dampen the negative effects of deindustrialisation on labour markets. Additionally, the drivers for informality differ between LAC and SSA, which is most likely due to different nature of informal employment between these regions. Deindustrialisation has a stronger informality increasing effect in Sub-Saharan Africa than in Latin America.
This thesis is structured as following. The second chapter gives an overview of the literature on deindustrialisation and informal employment, and the definitions for informal employment. In the third chapter, I describe how informality can be measured, introduce the constructed dataset and the used data sources as well as the methodology in which the empirics will be based on. Statistical overview of the key trends and the regression results are presented in the fourth chapter. In chapter five, I discuss the results and their policy implications and finally, chapter six concludes.
2. LITERATURE REVIEW
2.1.Structural change then and now
One of the core theories of explaining economic development has been the mechanism of structural change: economic growth driven by resource allocation from the less productive traditional sector into the more productive modern sector (Lewis, 1954). The traditional sector is often defined as the rural agriculture sector whereas urban activities in manufacturing and services have been defined as the modern sector. The process of structural change allows the overall productivity to rise, leading to higher levels of output and income. The contribution of this process to economic growth is largely dependent on the productivity differences of the two sectors. Productivity enhancing growth, where structural change occurs from less productive to more productive sectors, boosts overall economic growth. This is what happened to the Western countries during the industrial revolution. The growing manufacturing sector put economic growth on a new track, as the manufacturing sector was able to absorb large amounts of labour from less productive agriculture without large investments into human capital.
Above all, the possibilities for industrialisation are not as they were 200, 100 or even 50 years ago, and secondly, these possibilities seem not to be distributed evenly. More recent developers show wide heterogeneity in their structural transformation patterns. Not only does this transformation in developing countries differ in terms of industrialisation, but also on its contribution to overall productivity. Since 1990, the structural change in Asia has been productivity enhancing while both in Africa and Latin America, the structural change has been productivity reducing (McMillan & Rodrik, 2011). This implies that in the latter two regions, labour has shifted from higher productivity sectors into lower productivity sectors.
The possible underlying reasons behind the negative contribution of structural change to overall productivity, differ between Latin America and Sub Saharan Africa as well as within those regions. Latin American economies have experienced losses in their shares of manufacturing employment while other modern activities in the service sector have gained more employment and agriculture has continued to shrink (ILO, 2020). While many Latin American countries did show promising starts of rapid industrialisation during the second half of 20th century, the process has stagnated over the last decades before reaching the levels as high like the former industrialisers did. Sub-Sahara African countries show wide heterogeneity when it comes to structural change, but as a region, it has lacked behind the rest of the world in industrialisation. Agriculture remains the largest provider of employment. Some countries have succeeded in reaching higher levels of manufacturing, but those countries have been more an exception than the rule. In SSA together, the service sector has continued to grow, but rather slowly (ILO, 2020). The results by McMillan et al. (2011) on productivity reducing structural change in these two regions and these trends on lacking industrialisation suggest a positive relationship. The victory of industrialisation has derived from its ability to effectively rise the overall productivity. For instance, those countries in SSA where structural change has been productivity-enhancing, also had an increasing manufacturing employment share (McMillan & Rodrik, 2011).
Latin America (World Bank, 2020b). This indicates that even though the nature of manufacturing has changed, some countries have been able to grow through industrialisation. However, for the late industrialisers it has become increasingly difficult to exploit advantages from manufacturing. In this sense, the East Asian miracle has been claimed to be more like an exception to the rule (Rodrik, 2014).
Industrialisation is therefore not quite the same story as it was in the past centuries. Globalisation and trade liberalisation resulted in increased competition, pushing the manufacturing sector to become more productive. Several studies have shown that this results in the survival of the most productive firms, while the less productive firms exit the market (Pavcnik, 2000; Paus et al., 2003). During the modern era of globalisation, Asian countries have been the winners of this industrialisation race. The different regional consequences of globalisation on structural change can also be partly explained by policy responses (Kay, 2002). For instance, some Asian countries have publicly supported industries most exposed by import competition. Another notable factor has been the technological progress in manufacturing. It has resulted in less labour to be required for producing the same amount of output and in the increasing capital and skill intensity of this sector (McMillan et al., 2017). Thus, capital in relation to labour, has become an increasingly important resource for the success in the global competition in manufacturing. This also means that whereas traditionally, the manufacturing sector has been able to absorb large bulks of labour, this is not necessarily the case anymore. In Latin America especially, the increasing capital intensity of the manufacturing sector has been a big player behind the decreasing employment shares. As capital resources differ between regions, the consequences of the increasing capital intensity in manufacturing has diverse effects. SSA has its comparative advantage mainly in labour intensive sectors, which implies increasing difficulties to keep up and succeed in global manufacturing race.
What is important to remember is that structural change through large scale industrialisation is not the only path for sustained growth. For instance, India has been growing in the last decades largely due to service sector growth and seems to have skipped the industrialisation phase before it (World Bank, 2020b). It is evident that industrialisation is not a mandatory condition for reaching higher levels of development, as the whole world is tilting towards dominating service sectors. But what is not clear, is the consequences of different structural change paths as the development stories we are most familiar of, have been fuelled by industrialisation.
2.2.The role of institutions in structural change processes
through structural change can be rapid, but without investments in the fundamentals, it diminishes in the longer run (McMillan et al., 2017). The quality of institutions has been found to have a crucial role in explaining the global difference in income levels. One of the pioneer studies of this research field is a study by Acemoglu, Johnson and Robinson (2001), who find that global income differences source by large extent from the history of institutional development. There is an ongoing debate in the economic literature on whether it is the institutions or human capital that has been the main driver for the divergence of the economies (Glaeser et al., 2004). But that question is not critical for the analysis in this thesis nor it was in the work of McMillan et al. (2017), who simply placed both under the term “fundamentals”. To be precise, in this paper, by institutions, I refer to the whole institutional system of a country, i.e. the “fundamentals”, which is therefore also linked to concepts like human capital.
Literature has shown that the structural change itself is insufficient source of sustained economic growth if the investments in the fundamentals are lacking. Both LAC and SSA countries have struggled with building high quality institutions which has been constraining the economic convergence of these regions with respect to the rest of the world (Knack & Keefer, 1995; Coatsworth, 2008). Additionally, Sub-Saharan Africa is an international outlier when it comes to the levels of human capital. When analysing the structural change patterns of these regions, we also need to acknowledge the coexisting trends and characteristics that potentially influence on the development and outcome. Specifically, in this thesis, I will study the role of institutions together with sectoral employment changes, in an attempt to explain some of the current characteristics of how these economies have evolved and how they are currently functioning.
2.3. Informal economies
Aside from presenting divergent growth paths in terms of industrialisation, Latin America and Sub-Saharan Africa also share the common feature that much of private sector growth occurs in the extra-legal sector. In order to bring this aspect into the analysis, it is important to define what is meant by informal economy and what are the pivotal findings on what fosters its surge and persistence.
2.3.1. What is informality?
population growth and modern employment opportunities and another between people’s skills and the skills that are actually needed in the modern economic occupations (Chen, 2012). Many of the early concepts of informal work did not consider the possible positive impacts of these activities but rather tied informality together with criminal activities and tax evasion. Later, the concept has evolved more towards to an alternative way to organise markets, namely in the fast-growing cities of developing countries, and this sector has been increasingly linked together with the formal sector and state regulation (Hart, 2006). The concept of formal and informal markets as two separate entities has also been challenged. For instance, Maloney (2003) finds that the organisation of the labour markets in many developing countries does not fit into the dualistic framework. Informal sector should also not be considered purely as a safety net for those who end up unemployed in the formal sector, as it also includes workers voluntarily choosing this sector over the formal one (Maloney, 2003). The interaction between these two sectors has been acknowledged and increasingly emphasised in literature as well as in policy discussions. Where the traditional dualistic view saw connections between formal regulations and informal sector, more recent research also emphasises the economic linkages between informal and formal companies and considers the respective markets to be connected.
Due to continuous research and the heterogeneous characteristic of informality, its definition continues to transform. Conceptually, informal activities can be described as happening outside the controls of the state, which in that case are often rather weak (Hart, 2006). Informality has also been uniformized by the International Conference of Labour Statisticians (ICLS), which has provided criteria for defining and measuring informality. The latest publications have emphasised the heterogeneity of informal sector, since in different countries, informality appears in very different ways especially when comparing countries with different development levels (ICLS, 2018).
(or the absence of direct social payments) or a written contract. A drawback of the definition based on work contract, however, is that it can only be applied to wage workers. Although the definitions of informal sector and informal employment differ, estimates based on both measures correlate strongly with each other. With the exception of Eastern Europe, Central Asia and urban China, larger shares of informal employment happen in the informal sector than outside of it (Vanek et al., 2014). The part of informal employment occurring outside the informal sector consist of unprotected workers in the formal sector and domestic workers in the households (Charmes, 2012).
Overall, a fundamental shift in the definition of informality can be identified. The old definition approaches the topic by an enterprise point of view and emphasises that informality is mainly due to employees avoiding legal obligations. The new thinking on the other hand focuses on the employment status and highlights the involuntary characteristic of informality (Chen, 2005). In this thesis, I make use of the informality approach based on more recent definitions. Specifically, informality will be considered from a worker perspective rather than the perspective of the informal sector or informal firms. This approach is chosen as the interest is to explain on how sectoral employment opportunities and changes over time influence individual’s decision to be employed in the formal versus in the informal sector. This approach also provides advantages for measuring the informality of an economy, which will be further discussed in chapter 3.
2.3.2. What causes informality?
Various estimates show that in LAC and SSA, the informal sector accounts for larger shares of employment compared to the rest of the world (ILO, 2018). During the last decades, this sector has been growing in these economies and accounted for a significant part of the economic growth. In SSA, the informal employment has been estimated to account over three quarters of the total employment and in Latin America, the corresponding share is approximately one-third (ILO, 2018). Especially for the poorest, this sector plays a crucial role in providing employment and opportunities for higher levels of income.
Historically, informality has been associated with low levels of income. This still seems to hold in a cross-country comparison: it is the poorest parts of the world who have the largest informal sectors. But informality has also been considered as a passing stage of early development, which then fades away as a country reaches higher levels of income. The support for a negative relationship of income and informality through time is however less evident. This has also been the case for LAC and SSA: despite a growth in GDP, informal sectors in many countries have been growing or persisted at their levels. Whereas the cross-country comparison clearly indicates the negative relationship between income level and informality, time-series of individual countries show less support. More recent literature has addressed that informality is perhaps not just a transitory stage, but instead it has become a norm of many economies (Aryeetey, 2015). Yet, these findings do not reject the possibility that large-scale informality has hindered faster economic growth and higher levels of GDP. Nevertheless, it is evident that income growth does not fully explain the regional trends in informality, and therefore scholars have attempted to capture more specific determinants for informality.
Similarly to the discussion of economic growth, the role of institutional quality emerges again. Empirics provide strong evidence of the negative relationship between the quality of institutions and informality of the economy. There are several mechanisms of how the institutions of a country can affect the existence and size of the shadow economy. Institutions influence the behaviour of economic agents regarding informality through for instance by determining the quality of public services, the existence of property rights, rule of law, political regime, and tax compliance (Benjamin et al., 2014). Better institutions have been found to reduce informality by enforcing the compliance of rules and regulations and by improving educational attainment (Dabla-Norris et al., 2008), by providing a stable environment for businesses (Enste, 2010) and by reducing uncertainty and transaction costs (Aryeetey, 2015). In contrast, poor institutions can encourage informality through the lack of social security or other protections, poor law enforcement or control of corruption and by worsening the financial institutions and therefore preventing the access to credits.
levels and duration, and minimum wages (Lehmann & Muravyev, 2012). Labour market rigidities have found to be linked to the lack of job creation in the formal sector which has further led into higher levels of unemployment and informality (Khera, 2016). Product market regulations on the other hand often affect through formal sector entry costs and requirements as well as property rights (La Porta & Shleifer, 2014). But it is not only the quantity of regulation that matters, but also its quality. In the presence of poor institutions, strict regulations result in higher levels of informality (Loayza et al., 2005). Dabla-Norris et al. (2008) argue based on their findings, that it is indeed the legal system and quality of institutions which fundamentally determine the size of an informal sector. Similar regulations can therefore have very diverse effects in different institutional environments.
The concept of informality cannot be analysed without discussing the role of capital. What is interesting, is that although the informal sector is considered with low capital intensity and the development failures of the poor countries are associated with lacking capital, still, there are large amounts of unrecognised assets circulating in the informal economy. De Soto (2000) estimated that in the whole developing world and former communist countries, ”85 % of the urban parcels in these nations, and 40-55 % of rural parcels, are held in such a way that they cannot be used to create capital” (p. 34). The amount of assets owned informally, outside the capitalistic system, is approximately at least as large as 9.3 trillion US dollars (De Soto, 2000, p.35). These workers of the informal sector are constrained from accessing the formal economy, as their assets simply would not be considered there. This is a result of lacking property rights. Without the recognition of their assets, people cannot use them as collateral to get credit and thus lack the required capital to enter the formal markets. The accessibility of credit is therefore crucial if a country wants to increase formal employment.
2.4. Structural change and informality
A question of interest in this thesis is that to which extent the puzzling trends of structural change (or the lack of it) and informality in LAC and SSA regions are connected. I have now presented two kind of dual characteristics of economies: one that divides employment into agriculture and modern activities and another that divides it into formal and informal sector. Economic development models state that the process of structural change implies a shift of labour from the agriculture into the modern sector. But if the modern formal sector is unable to absorb all the labour willing to work in that sector, the excess labour could potentially be pushed into informality. As a result, structural change could potentially raise non-agricultural informal employment. But the causality between these two trends is likely to be difficult to predict. One can rather easily notice that these are not separate phenomena since the informality changes generally occur simultaneously with changes in the sectoral structure of the economy. The question that I am addressing is to which extent these different growth patterns have affected the informality levels. More specifically, what is the role of industrialisation and could the lack or decreases of industry employment explain some of the high informality rates of economies in LAC and SSA?
an internationally competitive manufacturing sector has been closed before this sector has even boomed. One reason why structural change has not increased total productivity could be that it has implied labour moving from agriculture into low productive informal activities, namely in the service sector.
It is important to remind that similar labour movements between sectors does not imply similar effects on productivity levels. Whether changes in the employment composition have an impact on productivity, always depends on the sectoral productivity levels of that particular economy. As manufacturing sectors around the world are found to be unconditionally converging i.e. the productivity levels should equalise in the long run (Rodrik, 2013), outcomes of structural change depend the productivity levels of other modern activities. These findings give support for the setting, that followed by decreasing manufacturing, labour has shifted into small-scale informal activities, which further has played its part on the negative structural change. However, to the best of my knowledge, no empirical evidence of this mechanism exists. The question that follows, is that what happens on the informality of a country when changes in the industrial sector occur? Estimates of the non-agricultural informal employment indicate that informality does not only occur in the service sector, but also in the industry sector (ILO, 2018). If informality appeared to the same extent in the industry sector and service sector, deindustrialisation in the expense of service sector would not change the informal-formal balance. This would imply that negative structural change is not due to shift into more informal activities, as suggested by McMillan et al (2011), but simply due to lower productivity levels of the service sector. Similarly, labour shifting from agriculture to either of these two sectors would lead to same increase (or decrease) in the informal employment levels. I raise a suggestion, that in the regions where deindustrialisation or the lack of industrialisation, and negative structural change has appeared while informality has persisted (or even grown), informality of the service sector in relation to industry sector differs.
reinforcing one another. History therefore offers support for the argument that manufacturing fosters the job creation in the formal sector and improves labour market institutions, but the question of how valid this process is in today’s global world, remains open.
2.5. Hypotheses
Motivated by the literature and its gaps, I suggest that different changes in the aggregate sectoral employment composition of the economy could explain some of the differences in the informal employment levels. More specifically, my interest is in testing whether industrialisation, and deindustrialisation, has an impact on informal employment. Due to larger scale of units and higher capital intensity, manufacturing sectors are more likely to grow formally than service sectors. I expect that institutions play a big role in determining this effect. In the case for decreasing industry employment, the displaced workers are more likely to end up in the informal sector if the institutional environment is poor. In another scenario, more applicable to SSA countries, the lack of industrialisation together with migration from rural to urban areas is expected to increase informality. In this case, if the urban (formal) labour markets are unable to absorb the increased labour supply due to institutional constraints, the displaced workers from agriculture end up in the urban informal jobs. Thus, deindustrialisation as well as the lack of industrialisation combined with poor institutional environment is likely to cause increases in informal employment. Vice versa, deindustrialisation is not expected to impact informality in the presence of good institutions where the transition into services would occur in the formal sector. To provide answers on how these structural change patterns and informal employment trends are connected, the following hypotheses will be tested.
H1: Deindustrialisation leads to higher informal employment.
H2: If structural change implies service sector growth without industrialisation, this leads to increased informal employment.
H3: The informality-increasing effect of deindustrialisation diminishes in the presence of strong institutions.
3. DATA AND METHODOLOGY
3.1. Measuring informality
One of the biggest challenges in the research around informal economic activity has been how to measure it. As this activity is not disclosed in the national accounts, alternative methods to measure it have been developed. One way to categorise these methods is to divide them into direct and indirect approaches. Direct approaches use data from micro surveys, such as the national labour force surveys, to capture the size of informality. They provide more detailed information of the size of the informal economy, but the challenge with this approach is the comparability of the results across countries. Although estimates based on these micro surveys has been harmonized to some extent, by the ILO for instance, the data coverage remains scarce. Therefore, this approach is generally more suitable for micro-level empirical analyses concerning fewer number of countries.
are known, the size of the informal economy can be estimated (Schneider & Enste, 2000). This method has become widely used in the analysis of informality at the macro-level.
The most common proxy for informal employment used in the literature, has been self-employment (see Fiess et al., 2010; Loayza et al., 2011). It has been found to highly correlate with informal employment in the developing and emerging economies, as indeed the majority of the workers of this sector are self-employed. Since it also fits into my interest of informality, the employment side, this proxy is used in this thesis to capture the informal employment in LAC and SSA countries. Using self-employment as a proxy for informal employment provides some key advantages. Firstly, it allows me to construct a larger dataset compared to the use of existing estimates of informal employment. Secondly, self-employment is a well-defined concept and highly comparable between countries. Therefore, a panel of data can be obtained, allowing analysis with both cross section and time dimension. Furthermore, a correlation with estimates of informal employment provided by the ILO for those countries that they exist is tested: share of informal employment (ILO estimate) correlates with the share of total self-employment by 0.82 and with the share of estimated urban self-self-employment by 0.67. Hence, self-employment is considered as an appropriate proxy for informal employment.
Since I am using structural change measures based on sectoral employment shares as explanatory variables, the use of the total self-employment share is likely to cause an issue of endogeneity. This is because large part of the self-employment in less developed countries occurs in the agricultural sector, which implies that total self-employment share is partly determined by the structure of the economy and vice versa, sectoral employment composition by the total employment share. To solve this issue, I use measure for urban self-employment instead of total self-self-employment. An implication that emerges, is that any non-agricultural informal employment in the rural areas will be excluded from the measure. However, due to data limitations on agricultural self-employment on macro level, this approach is considered as the second-best option to detect the non-agricultural self-employment. Another caveat is that formal self-employment will be included in the measure. Additionally, if it appears disproportionately in services versus industry sector, this could bias the regression results. This will be further discussed and taken into account when interpreting the results.
3.2. Dataset
also located in LAC and SSA, will be used. A list of countries included in the main sample can be found in the table A1 and the extended sample in the table A6 in the appendix. The self-employment measures and structural change measures cover the time period from 1991 to 2018, while the set of other independent variables together span the period from 2006 until 2018. Regressions therefore only cover this latter period, but the mentioned employment data prior that will be used for describing the trends informal employment and to test for causality. This period is not only chosen because of data availability, but also as it covers most recent trends of the latest phase of globalisation, also known as the second unbundling (Baldwin, 2006), where international trade patterns have become increasingly complex and the production processes have been divided into multiple locations. The chosen time period also captures most of the recent deindustrialisation trend.
3.2.1. Dependent variables
The International Labour Organisation (ILO) has collected data from national labour surveys to estimate the share of self-employment from the total employment. In ILO estimates, total employment is divided into two categories, those who are self-employed and those who work for salary & wage (employees). Self-employed workers are those who are “working for their own account or with one or a few partners or in cooperative and hold the type of jobs defines as a self-employment job” (ILO, 2020). The sub-categories of self-employed are employers, own-account workers, members of producers’ cooperatives, and contributing family workers. Self-employment is measured as a share (%) of the total employment and it will be used as an additional dependent variable InformalTotal to proximate total informal employment. To further calculate urban self-employment, I use the Global Jobs Indicator Database (JoIn) provided by the World Bank, which includes harmonized data from national surveys and subnational microdata for standardized labour indicators. The advantage of this dataset is that it disaggregates indicators by rural and urban individuals and divides employment into wage workers (employees), self-employed and employers, similarly to ILO estimation. By combining data from these two data sources, an estimation for urban self-employment, the variable
Informal, is constructed. It gives a measure of the share (%) of urban self-employment of the
total urban employment.3 Alongside, a growth rate of urban self-employment, InformalGrowth, will be used as a dependent variable. More detailed description of the calculations of the urban self-employment variable based on the two data sources can be found in the appendix I. 3.2.2. Independent variables
To calculate structural change in terms of aggregate labour movements between sectors, I use sectoral employment shares provided by ILO. The ILO estimations disaggregate employment
into three sectors, industry, agriculture and services, by their share of total employment. ILO uses the International Standard Industrial Classification (ISIC) to divide the employment into sectors. Because the ISIC classification is based on the workplace, and not on the type of work, sectoral employment includes all those workers who are employed in a company or employer classified with that specific sector, regardless of the occupation of the worker. Since my interest is in the expansion and contradiction of the industrial and service sector, using this classification provides the required measure for my analysis. More specifically, the interest here is not in the type of work, for example industry-related tasks occurring outside the industry sector (such as small manufacturing tasks in the rural farm) but rather on the amount of labour employed in the respective sector. Therefore, the estimations provided by the ILO are suitable for this purpose. An implication of ILO data is that I cannot measure solely manufacturing, but instead the industry sector as a whole has to be used. Although this limits the analysis to only three sectors, it allows me to cover more countries and years which further allows the use of highly comparable data on institutional quality and regulations.4
Using the sectoral employment data, I decompose the total annual employment growth by the three sectors. Each sector’s contribution to total employment growth is equal to the sectoral employment growth rate multiplied by the sector’s share of total employment. For mathematical expression of the decomposition, see Appendix I. To measure structural change in terms of industrialisation and service sector growth, the contributions of industry and service sector to total employment growth as described, will be used as independent variables Industry and
Services. These variables can be interpreted as how largely the respective sector accounted for
annual job creation in an economy.
To measure institutional quality, I use data from the Worldwide Governance Indicators (WGI), developed by Kaufmann, Kraay and Mastruzzi (2010). This dataset consists of six indicators measuring perception of institutional quality: control of corruption, government effectiveness, political stability & absence of violence/terrorism, regulatory quality, rule of law and voice & accountability. Countries are given a yearly score in all these areas, which value ranges from -2.5 (weak institutional quality) to -2.5 (strong institutional quality). I use the scores for rule of law and regulatory quality in order to capture the effect of institutions. Rule of Law score measures perceptions of confidence and obedience for the rules of society and how the quality of contract enforcement, property rights, the police and other public authorities is perceived. By choosing this indicator, I assume that better rule of law creates less incentives to find ways to do business outside the law. The score for Regulatory Quality indicates perceptions of the government’s ability to create and implement regulations that assure and encourage private sector development. As shown in the literature, not only the strictness of the regulation but also 4 An alternative would be to use the 10-sector database provided by the Groningen Growth and Development Centre
its quality affects informality. I therefore suggest that these variables, RuleOfLaw and
RegQuality, are comprehensive for capturing the link between institutions and the way business
is done, i.e. either formally or informally. To further measure regulations and possible entry costs for the formal sector, I use indicators from the Doing Business data. More specifically, I will use data for the number of procedures required to start a business (Procedures), labour tax rate (LabourTax) and the score for access to getting credit (Credit). More detailed description of these variables and the computation of variable Credit can be found in the appendix.
3.2.3. Control variables
The following set of control variables will be used. To control for the effect of income and general development level, I use GDP per capita (measured by PPP in current international dollars). The variables used are lnGDP, which is the annual GDP per capita in logarithmic scale, and GDPGrowth, which is the annual growth rate of GDP per capita. Unemployment rate has been found to be linked to informal employment, although the direction of this effect seems ambiguous.5 To control for the effect of unemployment in the model, a variable Unemployment will be added. It is measured as a percentage of unemployed of the total labour force, and the data is provided by ILO. The use of urban self-employment is likely to be influenced by the urbanisation level of a country. To control for this, I add UrbanPop as a control variable, which measures the share of total population that lives in the urban areas. Finally, to detect the effect of trade openness on informal employment, I include a variable Trade, which measures the sum of annual exports and imports of total GDP. The variables GDPGrowth, Unemployment,
UrbanPop and Trade are all expressed as percentage. The set of control variables are obtained
from the World Development Indicators database provided by the World Bank.
3.3. Model
The methodology for explaining the size of informal employment is based on Loayza and Rigolini (2006). The model assumes the duality of the economy: individuals work in either in the formal sector or in the informal sector, depending on the benefits and costs they receive, regarding their own skills. Given the environment individuals work in, there are fixed costs as well as benefits for working in the formal sector which are equal regardless of the individual characteristics. The model suggests that regulation increases the costs of staying formal but also creates benefits depending on the efficiency in transforming regulation (quality). This implies that regulation can have a dual impact by providing benefits of formal economic activity but also creating costs and that the costs and benefits also depend on the quality of the regulatory
system and institutional environment. Strong institutions create incentives to enter the formal markets, thus the effect on informal employment is expected to be negative. These aspects and assumptions form the base setting for the models used.
In order to test the first hypothesis, I include a variable for industrialisation into the model. The effect of industrialisation on the level and growth rate of informal employment, will be separately tested by using equations (1) and (2):
(1) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡+ 𝛽2𝑅𝑢𝑙𝑒𝑂𝑓𝐿𝑎𝑤𝑐𝑡+ 𝛽3𝑅𝑒𝑔𝑄𝑢𝑎𝑙𝑖𝑡𝑦𝑐𝑡+ 𝛽4𝐶𝑟𝑒𝑑𝑖𝑡𝑐𝑡+
𝛽5𝑃𝑟𝑜𝑐𝑒𝑑𝑢𝑟𝑒𝑠𝑐𝑡+ 𝛽6𝐿𝑎𝑏𝑜𝑢𝑟𝑇𝑎𝑥𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡
(2) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝐺𝑟𝑜𝑤𝑡ℎ𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡+ 𝛽2𝑅𝑢𝑙𝑒𝑂𝑓𝐿𝑎𝑤𝑐𝑡+ 𝛽3𝑅𝑒𝑔𝑄𝑢𝑎𝑙𝑖𝑡𝑦𝑐𝑡+ 𝛽4𝐶𝑟𝑒𝑑𝑖𝑡𝑐𝑡+
𝛽5𝑃𝑟𝑜𝑐𝑒𝑑𝑢𝑟𝑒𝑠𝑐𝑡+ 𝛽6𝐿𝑎𝑏𝑜𝑢𝑟𝑇𝑎𝑥𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡
where informal and InformalGrowth, represent the approximated urban informal employment share and its growth rate, for country c in year t. In line with the first hypothesis, I expect deindustrialisation to increase informal employment and vice versa industrialisation to decrease it. Therefore, I expect the sign of the coefficient for 𝛽1 to be negative, when using both equations
(1) and (2). I expect institutional and regulatory quality as well as access to credit to have a negative impact on informal employment, i.e. the sign of 𝛽2, 𝛽3 and 𝛽4 to be negative. The
costs and constraints of formal employment are expected to increase informal employment, so that the signs of 𝛽4 and 𝛽5 are positive. To avoid repetition, these five institutional variables
will be included jointly by the denominator Institution in the equations that follow. The term 𝑋 includes the set of control variables. The term 𝛼𝑐 accounts for unobserved time-invariant
individual effects, 𝜇𝑡 controls for the effect of common (worldwide) shocks and 𝜀𝑐𝑡 represents
the error term. Table 1 displays the list and the expected signs of the set of institutional and structural change variables.
Table 1: Expected signs of the explanatory variable coefficients
The main change that I will make regarding the estimation of the model in comparison to the original study by Loayza and Rigolini (2006), is to use a fixed effects estimator instead of pooled OLS estimator. The data shows violation of the homoskedasticity and uncorrelation of the error term assumptions of OLS, causing regression efficiency problems with using the OLS estimator. I opt to use the fixed effects estimator which captures the individual specific
Independent variable Expected sign of the coefficient
characteristics that are correlated with the independent variables, by the intercepts. Applying the Hausman test indicates that due to unobserved heterogeneity, the random-effects estimator cannot be used. The heteroskedasticity is controlled by using robust standard errors. This change applies to the estimation of the following equations as well.
To test the second hypothesis, the variable for service employment contribution will be added: (3) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡 + 𝛽2𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡+ 𝛽3𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡
(4) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝐺𝑟𝑜𝑤𝑡ℎ𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡 + 𝛽2𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡+ 𝛽3𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡 As the growth of service sector is expected to increase informal employment, the sign of the coefficient 𝛽2 is expected to be positive. This implies that a positive contribution of service sector on total employment growth would increase informal employment in relation to wage employment. In addition, for testing the hypotheses H1-H2 with the models presented, regional differences will be captured by using sub samples of LAC and SSA separately as well as interaction with regional dummy variable.
To test the role of institutions in the structural change-informal employment relation, interaction variables will be used. Those variables capture the possible joint effect of structural change and institutional quality on informal employment. As institutional quality is originally (in equations (1)-(4)) measured by the continuous variable of RuleOfLaw and RegQuality, both ranging from -2.5 to 2.5, I first construct an additional categorial dummy variable to ease the comprehension and interpretability of the interaction. The dummy variable takes the value of 1, when the score for RuleOfLaw or RegQuality, respectively, is above or equal to zero and the value of 0 when the score is below zero. I therefore define the institutions or the quality of regulations to be
strong, if RuleOfLaw or RegQuality score is above or equal to 0. Thereafter, an interaction
variable between this dummy variable and structural change is constructed, measuring the effect of the respective structural change measure in the presence of better institutions in equations (5) and (6), and regulatory quality in equations (7) and (8):
(5) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡 + 𝛽2𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡+ 𝛽3𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑐𝑡+ 𝛽4𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡∗ 𝑅𝑢𝑙𝑒𝑜𝑓𝐿𝑎𝑤𝑐𝑡+ 𝛽5𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡∗ 𝑅𝑢𝑙𝑒𝑜𝑓𝐿𝑎𝑤𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡 (6) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝐺𝑟𝑜𝑤𝑡ℎ𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡 + 𝛽2𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡+ 𝛽3𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑐𝑡+ 𝛽4𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡∗ 𝑅𝑢𝑙𝑒𝑜𝑓𝐿𝑎𝑤𝑐𝑡+ 𝛽5𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡∗ 𝑅𝑢𝑙𝑒𝑜𝑓𝐿𝑎𝑤𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡 (7) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡 + 𝛽2𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡+ 𝛽3𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑐𝑡+ 𝛽6𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡∗ 𝑅𝑒𝑔𝑄𝑢𝑎𝑙𝑖𝑡𝑦𝑐𝑡+ 𝛽7𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡∗ 𝑅𝑒𝑔𝑄𝑢𝑎𝑙𝑖𝑡𝑦𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡 (8) 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑙𝐺𝑟𝑜𝑤𝑡ℎ𝑐𝑡= 𝛽0+ 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡 + 𝛽2𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡+ 𝛽3𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑐𝑡+ 𝛽6𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑐𝑡∗ 𝑅𝑒𝑔𝑄𝑢𝑎𝑙𝑖𝑡𝑦𝑐𝑡+ 𝛽7𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠𝑐𝑡∗ 𝑅𝑒𝑔𝑄𝑢𝑎𝑙𝑖𝑡𝑦𝑐𝑡+ 𝛽𝑛𝑋𝑐𝑡+ 𝛼𝑐+ 𝜇𝑡+ 𝜀𝑐𝑡
The signs of the coefficients 𝛽4 and 𝛽6 are expected to be positive, since the presence of strong
employment, the presence of strong institutions is expected to decrease this effect. Thus, the expected sign for the interaction terms takes the opposite sign than the coefficient for the original Industry variable. The same logic holds for the expected joint effect of institutions and services, the expected signs for 𝛽5 and 𝛽7 are negative, opposite to 𝛽2. This implies that better institutions are expected to decrease the informality increasing effects of service sector growth.
4. EMPIRICS
4.1. Statistical overview
4.1.1. Estimations of the size of informal employment
Using total self-employment as a proxy, the regional comparison shows that informal employment appears the largest in SSA with a share of total employment above 70 percent through the whole period while in LAC the respective share has remained between 35 and 40 percent. However, there are large differences across countries as self-employment ranges between 10 percent and 90 percent of total employment. Specifically, among the poorest countries in SSA, total self-employment reaches very high levels. Overall, the approximated total informal employment has reduced in both regions during the time period, but the change has been fairly moderate. Full list of the average values for InformalTotal per country can be found in the table A6 in the appendix.
Figure 1: Urban self-employment and GDP per capita 1991-2018
in both regions and no conclusions of decreasing trends of informality within the last years can be made.
4.1.2. Employment growth decomposition
Table 2 shows the average values for the employment growth decomposition over the period 2006-2018 for all the countries separately as well as on regional level. The average contribution of the service sector is largest in both regions, which signals the transition into more service-based economies. The most notable regional difference is that in SSA, agriculture has had a large contribution to the total employment growth. But as agriculture still captures a large share of total employment, it is reasonable that the total employment growth remains mainly driven by this sector. The industrial sector, on the other hand, has not been able to provide new employment opportunities to the same extent as the other two sectors. The average annual contribution of the industry sector has still been positive, which implies that over the period between 2006 and 2018, more jobs have been created in this sector than deconstructed. A closer look in the yearly data, however, does disclose years of negative employment growth thus negative values for Industry variable. It is also worth reminding, that deindustrialisation in general refers to decreasing share of industry sector, not necessarily a shrinking sector in absolute terms. Nevertheless, the remarkably lower contribution of industry sector in relation to services (and in relation to agriculture in the case for SSA), tells that the industry sector’s role as an employment provider has contracted. The average annual contribution of industry sector does not remarkably differ when the years 1991-2005 are included into the decomposition. The results for the employment growth decomposition for the whole period 1991-2018 can be found in the appendix table A8. Furthermore, cross checking the respective
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 10 15 20 25 30 35 40 45 50 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 G D P per cap it a (PP P, $US ) Urban in formal em pl oy m en t (% )
GDP per capita Entire sample GDP per capita SSA
GDP per capita LAC Urban self-employment Entire sample
employment growth decomposition on sub-sector level for a smaller group of countries shows that especially in the case for LAC, a notable part of employment growth of the industry sector occurs outside the manufacturing sector6. This implies that if industrialisation is considered mainly by the manufacturing sector, the values presented in the table 2 for IND thus the variable
Industry, might underestimate the recent deindustrialisation trends.
Table 2: Average annual employment growth decomposition for 2006-2018.
Latin America and Caribbean Sub-Saharan Africa
Country IND SER AGR Total (%) Country IND SER AGR Total (%) Belize 0.469 2.893 0.535 3.897 Benin 0.496 1.680 0.695 2.871 Bolivia 0.423 1.603 0.105 2.131 Botswana 0.437 2.217 0.644 3.298 Brazil 0.092 1.366 -0.433 1.024 Burkina Faso 2.252 2.285 -2.318 2.219 Chile 0.534 2.004 -0.037 2.502 Cabo Verde 0.648 2.035 -0.182 2.501 Colombia 0.575 1.764 0.139 2.478 Cameroon 0.652 1.960 -0.111 2.501 Costa Rica 0.157 1.447 -0.032 1.573 Congo, Rep. 0.791 2.092 1.052 3.935 Dominican Rep. 0.311 2.435 -0.027 2.719 Cote d'Ivoire 0.348 1.479 0.320 2.147 Ecuador 0.491 1.250 0.500 2.241 Ethiopia 0.491 1.332 1.547 3.371 Guatemala 0.464 1.951 0.812 3.227 Gambia 0.482 2.398 0.614 3.495 Honduras 0.489 2.104 0.998 3.591 Guinea 0.198 1.229 1.168 2.595 Mexico 0.423 1.454 0.098 1.974 Guinea-Bissau 0.169 0.963 1.739 2.871 Nicaragua 0.242 1.419 0.869 2.530 Malawi 0.460 1.800 0.868 3.128 Panama 0.571 1.539 0.264 2.374 Mauritania 0.445 1.437 1.006 2.887 Paraguay 0.629 1.997 -0.373 2.253 Mozambique 0.500 0.764 0.991 2.255 Peru 0.486 1.764 0.124 2.374 Namibia 0.525 2.014 0.237 2.776 Average 0.424 1.799 0.236 2.459 Niger 0.189 0.755 2.299 3.243 Nigeria 0.211 1.317 -0.155 1.372 Rwanda 0.577 1.833 0.362 2.772 Senegal 0.345 2.005 -0.098 2.251 South Africa 0.261 1.833 -0.029 2.065 Tanzania 0.310 1.218 1.213 2.740 Uganda 0.194 0.600 3.003 3.796 Average 0.499 1.602 0.676 2.777 Notes: IND= Industry, SER= Services, AGR= Agriculture. For more detailed methodology and equations, see Appendix I. Source: Own calculations based on sectoral employment data from the ILO.
4.2.Regression results
Regression results for equations (1)-(4) are presented in the table 3. With the share of informal employment as a dependent variable (left panel), the signs of the coefficients for institutional variables are as predicted and in line with the existing findings. Better regulatory quality and access to credit is associated with lower levels of informal employment, while costs of entering and staying in the formal sector lead to higher informal employment. Rule of law, surprisingly, does not explain informal employment within the sample. Higher share of urban population increases informal employment and trade openness decreases it, both with statistically significant coefficients. The right panel of table 3 shows the results when the model is tested
Table 3: Determinants of informal employment
with the growth rate of informal employment share as a dependent variable. Although income level does not explain the growth of informal employment, higher GDP growth rate reduced the growth of informal employment. These results together therefore do not give strong support for higher income levels resulting in lower informal employment but do indicate that higher annual growth rate of GDP affects negatively on the informal employment growth rate.
Turning into the effect of industrialisation, a positive contribution of industry sector on employment growth is associated with negative effects on both the level and growth rate informal employment, which is captured by the negative coefficients for Industry in columns (1) and (3). This is in line with hypothesis one. To give a perspective of the size of this effect, in an economy where the industry sector captures one third of the employment, 3 percent annual
Dependent variable: Informal Dependent variable: InformalGrowth
VARIABLES (1) (2) VARIABLES (3) (4) Industry -0.363** -0.353** Industry -1.060** -1.095** (0.157) (0.159) (0.477) (0.482) Services -0.0315 Services 0.104 (0.0915) (0.228) RuleOfLaw -0.0478 -0.0391 RuleOfLaw -1.387 -1.416 (1.943) (1.939) (1.145) (1.156) RegQuality -3.882* -3.862* RegQuality 3.133 3.066 (2.220) (2.207) (1.919) (1.918) Credit -0.219** -0.218** Credit 0.0590 0.0552 (0.0988) (0.0978) (0.0763) (0.0768) Procedures 0.424** 0.425** Procedures 0.238*** 0.234** (0.178) (0.180) (0.0858) (0.0881) LabourTax 0.143** 0.144** LabourTax 0.0438 0.0421 (0.0619) (0.0625) (0.0926) (0.0911) lnGDP -7.117 -7.153 lnGDP 2.498 2.617 (4.715) (4.678) (3.242) (3.202) GDPGrowth -0.0244 -0.0237 GDPGrowth -0.153* -0.155* (0.0638) (0.0632) (0.0785) (0.0799) Unemployment -0.342 -0.342 Unemployment 0.249* 0.248* (0.258) (0.258) (0.136) (0.137) UrbanPop 0.539* 0.536* UrbanPop -0.101 -0.0913 (0.270) (0.272) (0.163) (0.171) Trade -0.0489** -0.0490** Trade -0.0340** -0.0337** (0.0183) (0.0183) (0.0138) (0.0137) Constant 70.94 71.41* Constant -16.91 -18.45 (42.38) (41.87) (27.06) (26.99)
County Fixed Effects Yes Yes County Fixed Effects Yes Yes
Year Fixed Effects Yes Yes Year Fixed Effects Yes Yes
Observations 449 449 Observations 449 449
R-squared 0.326 0.326 R-squared 0.174 0.175
Number of countries 37 37 Number of countries 37 37
F-test 3.77 5.50 F-test 8.45 13.35
Prob > F 0.0002 0.000 Prob > F 0.000 0.0000
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
growth rate of industry employment lowers the share of informal employment by 0.36 percentage points the and the growth rate of urban self-employment by 1.06 percentage points. If the share of industry employment is larger, this effect becomes bigger and vice versa. These findings are statistically significant on a 5-percentage significance level. It can be therefore concluded that industrialisation generates more employment in the formal sector than in the informal sector, leading to reduction in the informal employment share. Vice versa, a negative contribution of industry sector to employment growth increases informal employment relative to formal employment. Deindustrialisation therefore leads to increased urban informality. The larger the industry sector is (as a share of total employment), the bigger the effect of its growth rate will be. This also means that in a country with larger industry sector and negative industry employment growth, the increase in the informal employment relative to formal will be bigger than in a country with smaller industry sector.
To test the second hypothesis, a variable Services is added to regressions as presented in equation (3) and (4). The results are displayed in the table columns (2) and (4) of table 3. The coefficients are positive, although small and statistically insignificant. Thus, these results do not indicate that employment growth driven by the service sector has an effect on the informal employment level or growth rate. The coefficients for Industry remain negative and statistically significant. This implies that whereas deindustrialisation has an increasing effect on informal employment, an employment expansion in the service sector itself does not reinforce the effect. But due to negative coefficient for Industry, if deindustrialisation occurs in the expense of service sector, overall, it will result in higher informality. Thus, the second hypothesis can only be confirmed with an if condition. If service sector grows simultaneously with contraction in the industry employment, overall informal employment will increase. However, if deindustrialisation is defined as an absence of industrialisation, service sector employment expansion does not have an effect on informality.
It is worth reminding that some of the difference between coefficients for Industry and Services is likely to be driven by the fact that in services, formal self-employment might appear more than it does in the industry sector, causing the coefficient for Services to be overestimated (towards positive). However, the negative coefficients for the institutional quality variables and the positive effect of regulatory burden indicate that the measure for urban self-employment in this sample is not a representative of entrepreneurial formal self-employment but instead, more driven by the businesses in extra-legal sector. Nevertheless, given the constraints with data and the use of the chosen proxy, results need to be interpreted with care.
Table 4: Determinants of informal employment in different regions
model’s goodness of fit and the coefficients when testing for regions separately, should be discussed. By the difference in R-squared between regions, it can be concluded that the model has more explanatory power for the variation in informality in SSA than for the variation LAC.In SSA sample, the variables explain 45 percent of the variation in informal employment share and 38 percent in the growth rate whereas the respective numbers for LAC are 36 and 24. The sub sampling in reveals that income does not to explain the level of informal employment within either of the regions, which is captured by the statistically insignificant coefficients for GDP variables in the left panel. As discussed, it is commonly accepted in the research that informal employment is negatively correlated with income. On a cross section comparison with large income level variation this could be the case, but within groups with more similar development levels, as well as over time, this effect is ambiguous. Although the coefficients of
lnGDP are statistically insignificant, they have the opposite sign for regions. The SSA sample
Dependent variable: Informal Dependent variable: InformalGrowth
LAC SSA LAC SSA
(1) (2) (3) (4) (5) (6)
Industry -0.0263 -0.641* -0.303* Industry -0.417 -3.906** -0.745**
(0.145) (0.364) (0.162) (0.488) (1.549) (0.353)
Services -0.0363 -0.00927 -0.00943 Services 0.123 0.552 0.0991
(0.0805) (0.230) (0.0888) (0.383) (0.454) (0.279)
Industry * SSA -0.267 Industry * SSA -2.253
(0.324) (1.577)
Services * SSA -0.0386 Services * SSA 0.382
(0.265) (0.561) RuleOfLaw -5.507 3.296 -0.0574 RuleOfLaw -1.558 -1.164 -1.491 (3.437) (2.059) (1.953) (1.402) (1.155) (1.102) RegQuality 7.287*** -7.989*** -3.814* RegQuality 6.670 1.432 2.920 (2.210) (2.505) (2.153) (4.473) (1.423) (1.811) Credit -0.203 -0.0924 -0.216** Credit -0.182 0.0238 0.0421 (0.216) (0.0829) (0.0973) (0.198) (0.0850) (0.0723) Procedures 0.467** 0.283 0.422** Procedures 0.245 0.301** 0.214** (0.213) (0.188) (0.178) (0.155) (0.138) (0.0924) TaxLabour 0.0648 0.322*** 0.144** TaxLabour 0.0185 0.0312 0.0352 (0.128) (0.109) (0.0621) (0.231) (0.0871) (0.0863) lnGDP 1.221 -7.386 -7.024 lnGDP -8.494 7.612* 3.832 (8.516) (5.082) (4.768) (6.073) (3.800) (2.830) GDPGrowth 0.00858 -0.0257 -0.0170 GDPGrowth -0.324 0.0772 -0.104 (0.0503) (0.0728) (0.0615) (0.250) (0.0567) (0.0824) Unemployment 0.130 -0.397 -0.334 Unemployment -0.723 0.503*** 0.302** (0.255) (0.320) (0.261) (0.418) (0.163) (0.149) UrbanPop -0.0434 0.970** 0.532* UrbanPop 0.630 -0.206 -0.114 (0.303) (0.383) (0.271) (0.561) (0.222) (0.171) Trade -0.0608** -0.0335 -0.0491** Trade -0.0738* -0.0304** -0.0353** (0.0252) (0.0279) (0.0184) (0.0362) (0.0142) (0.0135) Constant 20.52 61.96 70.54 Constant 45.56 -55.90 -27.49 (82.03) (48.43) (42.63) (53.73) (34.10) (24.89)
Country Fixed Effects Yes Yes Yes Country Fixed Effects Yes Yes Yes
Year fixed effects Yes Yes Yes Year fixed effects Yes Yes Yes
Observations 178 271 449 Observations 178 271 449
R-squared 0.360 0.454 0.327 R-squared 0.237 0.384 0.197
Number of countries 15 22 37 Number of countries 15 22 37
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
of countries has a lower average income level than that of LAC sample, suggesting that on lower levels of development income is more likely to negatively correlate with informality. The opposite signs could also signal of the different nature of informality between these regions. The right panel shows that income increases the informality growth rate in SSA. One percent increase in GDP per capita increases informal employment growth rate by 0.08 percentage points i.e. leads to 0.08 percent increase in informal employment share. The scale of this effect does not seem too large, but perhaps more interesting is the direction of this effect. A possible explanation is that even though overall, higher development level does not explain the differences in informal employment within SSA, in the short run, fluctuations in country’s GDP can correlate positively with informal employment changes. This could also indicate of reverse causality, implying that informal employment growth has contributed to increased annual GDP. As several studies have shown the sizeable impact of informal economic activity on GDP in developing countries, it is reasonable that informal employment growth and GDP correlate positively to some extent.7 Furthermore, column (4) shows that unemployment rate increases the growth of informal employment in SSA: 1 percentage point increase in unemployment rate increases the share of informal employment by 0.5 percentage. This is reasonable since unemployment and wage employment are negatively correlated (causing an increase in informal employment in relation to wage employment) and also as unemployed search for income opportunities in the informal sector. As this result only appears in the SSA sample group, where social protection is often weak, it gives support for the view that urban informal employment works for some extent as a safety net for those who are unemployment and with no guaranteed government support for survival.
There are also differences in the effect of institutions on the share of informal employment when using different region samples. Again, RuleOfLaw does not explain informal employment. But an important remark is that stronger regulatory quality increases informal employment within LAC but decreases it in SSA. This could again indicate of the different character of urban informal employment between these regions. It is also possible that there is more formal self-employment in LAC than in SSA, causing the used proxy to be less representative of informal employment in LAC than in SSA.
Turning into regional differences in the effect of structural change, the sub-sampling reveals that the negative effect of industrialisation on informal employment remains significant only in the SSA sample. Thus, within LAC countries, industrialisation does not explain differences in the informal employment share or growth rate. Comparison between the original coefficients for the whole sample (Table 3) and the coefficients for SSA show that within SSA, the effect of industry employment growth on urban self-employment growth is more than 3 times higher. The effect on the informal employment share is also notably times larger when located in SSA.
