Urbanization and vulnerable employment: Empirical evidence from 163 countries in 1991–2019

Abstract

Many urban residents have recently lost their jobs due to the COVID-19 pandemic, which has made employment vulnerability in cities attained attention. It is thus important to explore the relationship between urbanization and employment. This study quantitatively analyzes spatiotemporal evolution and data correlation of urbanization and vulnerable employment, and explores the role urbanization plays in vulnerable employment by using historical data on 163 countries in the period 1991–2019 to test the theoretical hypothesis. The results show: It's clearly observed that there is a high correlation between the rate of urbanization and that of vulnerable employment, and the examples of G7 and BRICs are for it. The estimated urbanization yields a negative and statistically significant regression coefficient (−0.168), indicating that urbanization has a negative effect on vulnerable employment. If the urbanization rate increased by 1 %, the rate of vulnerable employment decreased by 0.168 %. The rural–urban sector conversion and changes in employment relationship driven by urbanization account for this. Countries with different income groups or populations have reacted differently to the rise in urbanization. Vulnerable employment in higher-income countries has been more significantly affected by the rise in urbanization, and more populous countries are more sensitive to it as well. These findings provide evidence for how urbanization promotes employment and decent work.

1. Introduction

Urbanization and poverty reduction are two vital aspects for global sustainable and healthy development (Chen et al., 2019; Srinivas, 2008) The global rate of urbanization (the share of urban population in total population) has increased from 43.4 % in 1991 to 55.7 % in 2019, which calls for insights into the complexity of urban development, social inequalities, economics and politics (Krellenberg et al., 2017). Furthermore, employment that is associated with social equalities, has substantial influences on global sustainability and human well-being. The inequality brought about by urbanization become more prominent because of COVID-19 (Acuto et al., 2020). The COVID-19 pandemic has led to a large number of deaths, damaged the global economy, and disrupted people's lives. The health crisis has affected all industries, and has restrained consumption and disturbed supply chains. Its economic impact on industry has affected the livelihoods of employers and workers alike (ILO, 2020a). In the face of this crisis, The vulnerable workers bear the brunt heavily. The global unemployment rate is about 5.4 %, but salaried workers account for only 52.8 % of all employed people.1 The difference is made up primarily by vulnerable employment, which accounts for 44.6 % of all employed people (about 1.46 billion people) in 2019. Vulnerable employment groups are poor in terms of economic power, knowledge and skills, income and remuneration, degree of unionization, social status, and social security, and thus are the first to bear the brunt of any given crisis (Laß & Wooden, 2019). The self-employed group must assume the risk of loss due to unstable employment relationships (Kirchhoff, 1996). Influenced by sudden environmental changes, reductions in production and sales may directly affect household income and livelihood for vulnerable employment groups. A growing number of researchers have realized that the challenge of employment in the context of global urbanization involves not only solving the problem of unemployment, but also treating and dealing with the problem of vulnerable employment. Vulnerable employment is generally observed in declining sectors (such as agriculture) and growth sectors (such as market services) (Hipple, 2004). In rural and feudal society, Vulnerable agriculture-dependent groups2 (e.g., farmers) are likely to suffer heavy losses under the impact of sudden natural disasters. Driven by urban and industrial power, people typically migrate from rural to urban areas. However, not all migrant workers have access to opportunity of formality, and a considerable number of them may choose to be self-employed or employed informally (Bederman & Adams, 1974). Taking migrant workers in China in 2017 as an example, three main forms of employment were available to them in cities: informal employment (42 %), self-employment (39 %), and formal employment (19 %) (Li & Zhang, 2020). Compared with the informally-employed, self-employed groups tend to engage in market services with lower entry barriers, but vulnerable groups in the services (e.g., shoemakers) are easily affected by poor economic and social conditions. Gindling and Newhouse (2014) find that as per capita income increases across countries, the structure of employment shifts rapidly, first out of agriculture into unsuccessful non-agricultural self-employment, and then mainly into non-agricultural wage employment.

The concerning situation of vulnerable employment has prompted international attention on the issue. It is worth noting that vulnerable employment has been observed in countries regardless of their levels of development, and is especially pronounced in developing countries, for which some international organizations have issued policies to promote economic equality and reasonable access to employment. “Decent work for all” is the principle guiding the International Labor Organization (ILO)'s work. It defines “decent” work as jobs covered and protected by formal labor institutions (Kucera & Roncolato, 2008). In September 2015, the United Nations approved “Transforming our World: The 2030 Agenda for Sustainable Development,” which listed “Promoting inclusive and sustainable economic growth, employment, and decent work for all” as the eighth of its sustainable development goals (UN-Habitat, 2009).

This paper tries to explore the impact of urbanization on vulnerable employment through long-term, transnational data. Data on 163 countries across the world for 29 years, from 1991 to 2019, were used to identify the relationship between urbanization and vulnerable employment, and test different urbanization effects among countries classified by wealth level and labor supply. The remainder of this paper is organized as follows: We first attempt to go deep into the relationship between urbanization and vulnerable employment based on literature review and phenomenon analysis, and then put forward a theoretical framework to explain the relationship. We next conduct an empirical analysis to establish a robust linear relationship between urbanization rate and vulnerable employment rate, after controlling for variables that might be associated with employment vulnerability. Further, we test heterogeneity of countries in income level and population size. Finally, we offer the conclusions of this study. The work here helps understand the trend of development of vulnerable employment and its significance for urbanization, and look deep into the relationship between them. This provides a theoretical basis for the relevant governmental policies.

2. The discussion about impacts of urbanization on employment

2.1. Literature review

Vulnerable employment, including own-account workers and contributing family workers, is the vulnerable part of self-employment, and is the main source of labor for the informal economy. The share of vulnerable employment in total employment (vulnerable employment rate) is an important indicator for measuring the vulnerability of the labor market. According to the International Classification of Status in Employment (ICSE-93) in 1993, workers are divided into six groups: namely, wage and salaried workers (employees), employers, own-account workers, members of producers' cooperatives, contributing family workers, and workers not classifiable by status (Fig. 1 ). Of them, wage and salaried workers are classed as paid employment and the rest as self-employment. Paid employment is stable and reasonable, employees are provided basic compensation, and their wages are not indexed to the work done. By contrast, vulnerable employment, own–account workers and contributing family workers included, is a highly heterogeneous group, often consisting of a mix of subsistence and unformed entrepreneurial activities (Osmani, 2004). Vulnerable employment is characterized by insufficient income, low productivity, hard working conditions, lack of guarantee of basic rights, and poor emergency capability (Gangopadhyay & Shankar, 2016; ILO, 2010). In particular in developing countries, owing to the lack of decent work opportunities, conventional indicators such as the unemployment rate cannot adequately describe the vulnerability of the labor market (ILO, 2010, ILO, 2018), and vulnerable employment rate explains more. Own-account workers and family workers in developing countries contribute a lot on informal sector. They are generally engaged in self-sufficient agriculture or other low value-added activities with low and variable incomes, and they usually produce goods or services meant for sale or barter, such as self-employed street vendors, taxi drivers, and home-based workers (ILO, 2020b). Different from formal sector, informal sector is characterized by (a) ease of entry; (b) reliance on indigenous resources; (c) family ownership of enterprises; (d) small scale of operation; (e) labor-intensive and adapted technology; (f) skilled acquired outside the formal school system; and(g) unregulated and competitive markets.

Fig. 1.

International classification of status in employment.

It's argued that vulnerable employment is not a perfect indicator to describe the vulnerability of labor market because part of own-account workers and contributing family workers are not in a precarious or vulnerable situation (Martinez et al., 2017). Labor market segmentation, insecurity, and inequality transcend simple citizen–noncitizen binaries due to deregulating sectors and informal economy (DeFilippis et al., 2009; Rogaly, 2009). These workers in vulnerable employment lack basic elements of decent work (such as not being covered by social security and/or social dialogue) as a matter of fact (Brown et al., 2015; ILO, 2018), and it's hard for them to gain equal rights in the labor market. For example, many of untrained construction migrants in Dubai were living in barrack-style, mass-worker accommodations and were routinely subject to abuses including breaches of contract, illegal fee extractions, and wage theft by both labor intermediaries and employers (Buckley, 2014). Statistics on vulnerable employment provide valuable information on the quality of employment and are crucial to a comprehensive understanding of the labor market, in both developing and developed countries.

The literature on the relationship between urbanization and employment has focused on the impact of urbanization on the industrial structure and employment structure. Classical structural theory (such as Lewis' Dual Economic Structure Theory) and population migration theory (such as the Petty-Clark Theorem) take the respective perspectives of economic structural change and migration to explain that urbanization results in an increase in employment in the secondary and tertiary industries, and a decrease in the primary industry (Timberlake, 1985). However, the relationship between urbanization and employment is not only a simple transformation of the structure of employment, but is also closely related to such social problems as poverty and inequality, as it goes, labor research is not just about labor, it is for labor (Castree, 2007; Peck, 2003). Therefore, a growing literature has underlined the role of employment vulnerability and more researchers have been attending the relevant researches in the context of urbanization.

There exists a fierce debate when taking employment vulnerability into consideration. Some researchers hold the points that more urbanized areas have suffered from greater unemployment problems in the period of the post-2008 crisis than have less urbanized areas – for instance, Palaskas et al. (2015) note a positive relationship between population size and unemployment expansion, calling this ‘a crisis of urbanization’. It is suggested that areas' connectedness (or lack thereof) to other places is more significant than urbanization level in determining employment vulnerability (Psycharis et al., 2014). Equally, Mavroudeas (2014) has shown that in Greece's two largest metropolitan regions, hosting the cities of Athens and Thessaloniki, there has been a disproportionate expansion of precarious work (specifically part-time work) after the crisis, a rate that far exceeds the national rate. But other scholars show different view, and they think that in the relationship between urbanization and vulnerable employment, the influx of vulnerable immigrants is considered a major economic and social factor affecting labor vulnerability (Weil, 2009) – in other words, it is level of urbanization or city size that may influence employment vulnerability. A high ratio of wage and salaried workers in a country signifies advanced economic development, while a large ratio of self-employed workers indicates high risk in the labor market, a large agriculture sector, and low growth in the formal economy. Hence these scholars have researched the relationship between urbanization and informal employment by using self-employment data from Latin American countries (Rauch, 1993), global cross-country data (Elgin & Oyvat, 2013; Wu & Schneider, 2019), and provincial data (Huang et al., 2016; Huang et al., 2019). They have verified an inverted U-shaped relationship between urbanization and informal employment, that is, if the urbanization rate of a country reaches a certain level, informal employment in it as a share of overall employment decreases instead of increasing. Baklouti and Boujelbene (2020) hold that informality does not necessarily decline with increases in GDP per capita in countries with poor institutional quality with structural equation modeling. And past levels of inequality are the most salient factors explaining the size of the informal economy (Gutiérrez-Romero, 2021).

2.2. Phenomenon analysis

The evolutions of urbanization and vulnerable employment showed different patterns, with urbanization rate increasing and vulnerable employment rate dropping, respectively.3 The global rate of urbanization was on the rise on the whole, and its kernel density curve showed the characteristics of temporal variation in it (Fig. 2-a). The kernel density curve moved upward and to the right, showing that the distribution of the low rate of urbanization decreased significantly while that of a high rate increased significantly. The curve of kernel density of vulnerable employment showed that the global rate of vulnerable employment exhibited a slow downward trend (Fig. 2-b). The density of the higher vulnerable employment decreased year by year while that of lower vulnerable employment increased. The density of a moderate vulnerable employment changed little.

Fig. 2.

The kernel density curve of urbanization rate and vulnerable employment rate.

a. Urbanization rate.

b. Vulnerable employment rate.

According to the spatial pattern of urbanization from 163 countries from 1991 to 2019 (Fig. 3 ), the urbanization rates of countries in Europe, America, and Oceania were higher than those in Asia and Africa. This was related to the level of economic development and industrialization of each country, but was not completely consistent. We noted countries where urbanization had lagged behind economic development and ones where urbanization was ahead of economic development that, nonetheless, had high rates of urbanization. From 1991 to 2019, many African and Asian countries driven by industrialization recorded a rapid increase in their rates of urbanization. China, Indonesia, and Malaysia are typical countries of this kind. Some small-island countries like Dominica, significantly increased their rates of urbanization due to tourism. Others, such as Equatorial Guinea, relied on exports and primary product processing.

Fig. 3.

The global spatial pattern of urbanization rate in 1991 and 2019.

It's clear form spatial patterns of vulnerable employment that the rates in Europe, North America, and Oceania were generally lower than those in Asia, Africa, and South America (Fig. 4 ). From 1991 to 2019, the rate of vulnerable employment of Asian and African countries generally decreased significantly, especially in Cambodia, Vietnam, China, Thailand, and Myanmar in East Asia, India in South Asia, Turkey in West Asia, Kyrgyzstan and Kazakhstan in Central Asia, and Rwanda, Senegal, Gabon, and Cape Verde in Africa. In total, the rate of vulnerable employment in South American countries was consistently lower than that in Asian and African countries, but the change of the rate over 29-year period was less significant.

Fig. 4.

The global spatial pattern of vulnerable employment rate in 1991 and 2019.

Evidence of the negative relationship between urbanization and vulnerable employment was well presented in the separate years 1991 and 2019. The vulnerable employment rate and urbanization rate of various countries in 1991 and 2019 are drawn into scatter plot diagrams, with urbanization rate as the abscissa, and vulnerable employment rate as the ordinate. The value of each point in the coordinate system was coded (U, VE). It can be seen in the Fig. 5 that the square point representing the world average moved towards bottom-right from 1991 to 2019, reflecting the rise of the world average urbanization rate and the decline of vulnerable employment rate. The findings (Fig. 5) supported the existence of a negative relationship between the country rate of urbanization and vulnerable employment, suggesting that agricultural countries were faced with the most serious vulnerable employment problems, and vice versa. The black full line was a linear fitting of the trend in that year. The two equations of linear fitting were y = −0.9567x + 91.033 (a,R² = 0.593,F = 234.78, Prob > F = 0.000) and y = −0.802x + 84.962 (b,R² = 0.448,F = 128.94, Prob > F = 0.000),4 respectively. The absolute value of the slope in equation (b), −0.9567, was less than that in equation (a), −0.802, and the former fitting line in 1991 was steeper than that in 2019, suggesting that the decline of vulnerable employment rate caused by the increase of urbanization rate of a unit became smaller from 1991 to 2019.

Fig. 5.

Scatter plot of urbanization rate and vulnerable employment rate 1991 and 2019.

The rate of vulnerable employment was negatively correlated with that of urbanization. Fig. 6 showed the rates of urbanization and vulnerable employment of countries divided by income groups from 1991 to 2019. It's clear that the rate of vulnerable employment exhibited a downward trend with an increase in the urbanization rate on the whole. From the perspective of income groups, countries could be divided into 4 distinguished groups which were different in level of urbanization and vulnerable employment (referring to United Nations' meta data Country Income Group List). The urbanization rate of countries in more income groups was higher while that of vulnerable employment was lower correspondingly. As scatter plot in Fig. 6 shown, it could be concluded that a low-income country was characterized by low urbanization and high vulnerable employment, whereas a high-income country was characterized by high urbanization and low vulnerable employment. Meanwhile, the distribution of dots belonging to middle-income countries was typical. Lower-middle-income country dots were closer to low-income country dots, and upper-middle-income country dots were near high-income country dots in Fig. 6. In order to further understand the relationship between the urbanization rate and vulnerable employment rate, on the basis of the cross-country data, a fitting line with a slope of about −0.9 obtained statistically (Fig. 6). By using the coefficient and constant of the fitting line, the equation was drawn such that when the rate of urbanization was lower than 10 %, the rate of vulnerable employment was higher than 81 %, and when the rate of urbanization was above 90 %, that of vulnerable employment was below 9 %. If the urbanization rate was 50 %, the rate of vulnerable employment was about 45 %. Taking the BRICs and G7 countries as an example, the change in the rates of urbanization and vulnerable employment in each country had a negative correlation (Fig. 7 ). In relative terms, developed countries (G7) had entered a steady state, with high urbanization and low vulnerable employment, while emerging industrialized countries (BRICs) differ widely from one another in the degree and range of these two indicators. There might exist a deeper cause for the different pattern, and this is what we turn to next.

Fig. 6.

The inverse relationship between urbanization rate and vulnerable employment rate.

Fig. 7.

The urbanization rate and vulnerable employment rate in BRICS and G7.

3. Theoretical hypothesis

Urbanization denotes population flow from rural to urban areas as well as structural transformation in the economy and society. The influx of vulnerable migrants is among the important factors affecting the vulnerability of the labor force (Weil, 2009). As shown in Fig. 8 , large agricultural population has come into urban areas, and has turned to industry and services. However, new jobs in the formal sector are far fewer than the number of new migrants into cities (Oh, 2008). Due to the imbalance between supply and demand, these agricultural migrants are at a disadvantage when competing for jobs, and are compelled to be self-employed or informally employed (Huang et al., 2018). Structural transformation in urbanization necessitates alternative arrangements in dedicated production units that allow for economies of scale, and organized production in line with an increasing specialization of the workforce. Accordingly, transformation brings a reduction of own-account work of the subsistence type. In other words, falling proportions of the share of vulnerable workers, can be expected to accompany structural transformation from a low-income situation with a large informal or rural sector to a higher-income situation (Sparreboom & Gier, 2008).

Fig. 8.

The mechanism of urbanization affecting vulnerable employment.

In Fig. 8, two mechanisms of urbanization and vulnerable employment are described. The first mechanism is industry sector conversion in initial urbanization. Industrialization develops rapidly and labor is in short supply. In this process, a large number of rural population flow into cities, changing from agriculture to secondary or tertiary industry, updating their employment status. The second mechanism is employment relationship changes when urbanization is developing in depth. The speed of urbanization slows down, and the flow of population between urban and rural areas has slowed down, resulting in less opportunities in industry sector conversion. But those, who are vulnerable employment in cities, are still trying to get rid of vulnerability, transforming from being self-employed vulnerably to be employed formally. Although the first mechanism mostly occurs in the early stage of urbanization, the second mechanism is more likely to happen in the late stage (Sato & Zenou, 2015). In fact, these two processes are interrelated and complementary in the process of urbanization development. In the transition processes associated with urbanization, the government's concern has shifted from the scale of urban employment to the quality of urban employment.

3.1. Initial urbanization and industry sector conversion

Urbanization changes the urban–rural structure as well as the structure of employment. A large number of rural migrants have transferred from the fragile agricultural economy to non-agricultural work to raise real income (Guy et al., 2012). The ratio of self-employment farmers has thus declined in rural areas, leading to a drop in the rate of vulnerable employment. In low-income and lower-middle-income countries, although the agricultural output is not a major contributor to the economy, a considerable number of people are still engaged in agricultural production and most of them live in poor conditions, especially in Uganda, Madagascar, Mozambique, and other African countries. Contemporary urban areas remain the growth poles of economic progress and the lightning rods of political and social unrest (Todaro, 1997). Technological progress and large-scale management of farms have led agricultural labor to flows to cities in expectation of higher income in cities (Todaro, 1977). The informal sector developed very fast with migrant workers crowing in (Aguilar, 1997). At this stage, the informal employment acts primarily as a safety net and the size of informal employment is determined by opportunity of formality and the distribution of migrant workers' skills (Loayza & Rigolini, 2011; Shaw & Pandit, 2001). This transformation in industry contributed to the decline of vulnerable employment in the early stages of urbanization.

3.2. In-depth urbanization and changes in employment relationship

With continuing urbanization, many urban workers engaged in vulnerable employment have transferred to the formal sector for a stable employment relationship. This has altered their status from that of vulnerable self-employment, characterized by low and volatile earnings, to relatively low-risk and stable paid employment, thus reducing the rate of vulnerable employment (Gindling & Newhouse, 2014). The informal sector houses a large number of involuntary workers (Guenther & Launov, 2012). For them, it is difficult for migrant workers to engage in the limited formal sector or find stable jobs because they are at a disadvantage in the labor market in terms of human capital in the beginning (Okpara, 1986). Actually, urban informal employment not only acts as a safety net, but also can be seen as a springboard for migrant workers. Rather than being stuck in a weak and marginalized labor market, they can accumulate human capital and finally move into the high-level labor market, or develop into employers of micro-enterprises by utilizing informal and formal training during their informal employment (Huang et al., 2020). At the same time, the domestic formal sector has expanded in scale due to industrialization and urbanization, providing more opportunities for formal employment. Moreover, population agglomeration, industrial development, and knowledge sharing due to urbanization have brought about large-scale operations, and vulnerable employment groups that are self-employed have therefore suffered. For example, the fragmented self-employed retail mode is challenged by the large-scale retail mode characterized by chain operation and the supermarket, and has been forced to seek new jobs in enterprises.

3.3. Cross-country differences of urbanization effects

The response of employment vulnerability to urbanization changes is different in different countries. In addition to the general theoretical framework of urbanization and vulnerable employment, Societal institutions and functioning play an important role in the decline of vulnerable employment, but the welfare systems of the sample countries are hard to fully classified (Esping-Andersen, 1990). Labor supply and wealth related to welfare systems will affect the reflection of vulnerable employment on urbanization in different country groups (Bloom & Freeman, 1986). The labor supply affects employment size reacting to the improvement of urbanization level (Gundogan & Bicerli, 2009). Population size is an important indicator of labor supply. Populous country, such as China, India, the United States, Indonesia, Brazil have hundreds of millions of people, are rich in labor supply, while the Vatican, Tuvalu, Nauru and the island countries of the ocean are sparsely populated. The great gap between the rich and the poor among different countries is associated with employment quality, and there are great differences in national wealth levels between developed and developing countries all over the word (Bloom et al., 2008). The benign impacts of urbanization on urban-rural income gap and an out-of-box strategy—containing income inequality by promoting well-managed urbanization—is proposed (Wan et al., 2022). Therefore, countries need to be classified according to labor supply (population size) and wealth level (income level) before testing urbanization effects.

According to the above analysis, the ratio of agricultural own-account workers decreases, and then the ratio of urban own-account workers and contributing family workers in vulnerable employment gradually drops due to urbanization. Therefore, the following theoretical hypothesis is proposed: Urbanization reduces vulnerable employment on the whole, but its influence on different countries is different.

4. Empirical analysis

4.1. The response of vulnerable employment to urbanization

The rate of urbanization was estimated by the report “World Urbanization Prospects: 2018 Revision” by the United Nations Population Division. The rate of vulnerable employment was derived from data from the International Labor Organization: the ILOSTAT database. The economic and social data concerning the control variables were taken from the World Development Index (WDI) database for 224 countries or regions5 in total, covering 29 years from 1991 to 2019. Excluding samples with missing data, there were a total of 163 samples. A total of 4727 data items were finally obtained. As it was supported by large data samples, this study considered extreme cases, such as those of countries with particularly high or low levels of urbanization.

This paper constructs a model of the quantitative relationship between urbanization and vulnerable employment. The F-test and Hausman test are used to prove that fixed effect regression is better than pooled OLS regression and random effect regression, and is thus used here. Owing to the large number of samples, individual heterogeneity was strong and the time series was long. Hence, we control individual effects and temporal effects in the model. This chief innovation of our study is the modeling of the impact of urbanization and other influential factors on vulnerable employment. The panel regression model is as follows:

$$\mathrm{v}{\mathrm{e}}_{i,t}={\beta }_0+{\beta }_1{u}_{i,t}+{\beta }_2\text{lnagd}{\mathrm{p}}_{i,t}+{\beta }_3\mathrm{a}{\mathrm{p}}_{i,t}+{\beta }_4\mathrm{s}{\mathrm{p}}_{i,t}+{\beta }_5\mathrm{u}{\mathrm{e}}_{i,t}+{\beta }_5\mathrm{lns}{\mathrm{l}}_{i,t}+{\theta }_i+{\lambda }_t+{\epsilon }_{i,t}$$ (1)

where ve_{i, t} denotes the vulnerable employment rate in each country i observed in year t, and u _{i, t}denotes the urbanization rate in each country i as observed in year t. β ₁∼β ₅ are all estimation coefficients, and β ₁ is most important and denotes the impact of urbanization on vulnerable employment. θ _i and λ _t denote country-fixed effects and time dummies that account for time-invariant unobservable factors at the country level and common shocks to countries, respectively. ε _{i, t} is an error term.

The explained variable is the vulnerable employment rate (ve), that is, vulnerable employment as a ratio of the total employment. The core explanatory variable is the urbanization rate (u), which is the ratio of the urban population to the total population. We consider other important factors as well: economic development, non-agriculturalization, modernization, employment environment, and securing loans convenience. The control variables were the GDP per capita (lnagdp), agriculture value added as a ratio of GDP (ap), services value added as a ratio of GDP (sp), unemployment rate (ue), and strength of the legal rights index (lnsl). For the above variables, the ratio data were the original values, and the logarithm of the other data was taken. Most of the control variables used to identify the impact of urbanization on vulnerable employment are widely used in the related empirical literature on informal employment as dependent variables (Elgin & Oyvat, 2013; Rauch, 1993). The GDP per capita denotes economic development. The higher the level of economic development is, the smaller is the share of vulnerable employment. Agriculture value added as a ratio of GDP denotes non-agriculturalization. It is expected that non-agricultural industrialization prospers if non-agriculturalization level declines, and the share of self-employed people engaged in agriculture decreases correspondingly. Services value added as a ratio of GDP denotes the level of modernization, and its expected coefficient is negative. An improvement in services value added as a ratio of GDP implies the optimization of the economic structure and a decrease in vulnerable employment rate. For unemployment rate (ue), the expected coefficient was positive. The informal sector had the potential to absorb the unemployed (Henry, 1982), and unemployed workers were more likely to become self-employed workers and informal wage workers, which led to high vulnerable employment rate. The strength of the legal rights index (lnsl) measures the degree to which collateral and bankruptcy laws protect the rights of borrowers and lenders, and thus facilitates lending. A higher value indicates that the relevant laws in the given country are better designed to expand access to credit. The expected coefficient was positive, that is, the higher the index was, if loaning was encouraged, more workers were self-employed (Murdoch, 2006).

For panel data, we first tested for the existence of the unit roots of all variables. The result implies that all variables were stationary in the first difference, and that a long-term equilibrium could be sustained among all variables. This analysis next proceeded with the cointegration test. The result shows a constant, long-term equilibrium relationship obtained among urbanization, vulnerable employment, economic development, agricultural level, modernization level, employment environment, and loan convenience from 1990 to 2019. The consequences from these tests for the specification are delegated to Appendix.

To avoid multi-collinearity, heteroscedasticity, and autocorrelation in panel regression because of a large number of dummy variables, we used a number of methods. In view of potential autocorrelation problems, the fixed-effect model was used according to the Hausman test, and regression was conducted after the elimination of autocorrelation. Finally, the OLS method was used (Stock & Watson, 2019). According to Eq. (1), we used regression analysis, adding economic development (lnagdp), non-agriculturalization (ap), modernization (sp), employment environment (ue), and securing loan convenience (lnsl) in the model, in order. The results of regression are presented in Table 1 . In the table, The F test is significant, showing explanatory variables have a significant impact on dependent variables. And the R² fitting degree is >0.3, reflecting that variable can explain >30 % of the variation. The main result was that the impact of urbanization on vulnerable employment was significantly negative. Column 6 shows that if the urbanization rate increased by 1 %, the rate of vulnerable employment decreased by 0.168 %. This implies that the scale of modern enterprises has expanded, and more jobs are available due to urbanization. Former own-account workers and family workers have turned to the formal sector, which has reduced the rate of vulnerable employment (Tomal & Johnson, 2008). Overall, our main hypothesis is supported by the data.

Table 1.

Results of panel regression from full-sample regressions.

	(1)	(2)	(3)	(4)	(5)	(6)
Urbanization rate (u)	−0.197***	−0.168**	−0.172***	−0.169***	−0.168***	−0.168***
	(0.0640)	(0.0646)	(0.0603)	(0.0601)	(0.0599)	(0.0599)
The GDP per capita (lnagdp)		−0.020***	−0.014**	−0.014**	−0.015**	−0.015**
		(0.0065)	(0.0060)	(0.0060)	(0.0061)	(0.0061)
Agriculture value percentage (ap)			0.146***	0.171***	0.168***	0.168***
			(0.0449)	(0.0499)	(0.0503)	(0.0504)
Services value percentage (sp)				0.042	0.044	0.044
				(0.0288)	(0.0289)	(0.0290)
Unemployment rate (ue)					−0.039	−0.040
					(0.0383)	(0.0385)
Strength of the legal rights index (lnsl)						0.004
						(0.0084)
Constants	0.522***	0.659***	0.592***	0.563***	0.568***	0.562***
	(0.0332)	(0.0630)	(0.0603)	(0.0656)	(0.0677)	(0.0690)
N	4727	4727	4727	4727	4727	4727
R2	0.312	0.340	0.366	0.368	0.369	0.369
F	8.57***	10.25***	7.51***	7.54***	7.54***	7.24***

Note: Standard deviation in parentheses. Significance of standard errors ***p < 0.01, **p < 0.05, *p < 0.1.

Statistically, note that the GDP per capita is significantly negative at a level of 1 % in columns 1–6 in Table 1, and the ratios of agriculture, forestry, and fishing, value added, were significantly positive at a 1 % significance, which is consistent with expectation. The unemployment rate was significantly negative at the 1 % level, which was not in line with expectation. The higher the unemployment rate was, the more likely it was to be affected by an inclement environment. It's illustrated that own-account workers and contributing family workers were more willing to choose stable employment relationships at low risk in the context of tough employment environment (high unemployment rate). The strength of the legal rights index was significant at the level of 10 % in model 6, which is consistent with expectation. The ratios of services value added, unemployment rate, and the strength of the legal rights index are not significant in columns 4 and 5, and may need further group discussion. To deal with common shocks, time fixed effects and individual fixed effects are considered. Columns 1–4 in Table 2 report the benchmark results of the impact of urbanization on vulnerable employment under different effects. Time fixed effect model, which capture common time trends to all countries over the years, solves the problem of missing variables that do not change with individuals but vary with time (individual invariant). Similarly, the introduction of individual fixed effect can solve the problem of missing variables that do not change with time but change with individuals (time invariant). In this study, institutional factors may both affect the explained variables and explanatory variables. These factors may change with time and national individuals, so it is necessary to control individual effects and time effects simultaneously. All algebraic symbols of the estimated coefficients remained unchanged, and the coefficients changed relatively little.

Table 2.

The results of full-sample regression under different effects.

	(1)	(2)	(3)	(4)
	Time fixed effects	Individual fixed effects	Two-way fixed effects	Two-way fixed effects
Urbanization rate	−0.219***	−0.196***	−0.168***	−0.172***
	(0.0565)	(0.0522)	(0.0599)	(0.0603)
The GDP per capita	−0.019***	−0.018***	−0.015**	−0.014**
	(0.0064)	(0.0038)	(0.0061)	(0.0060)
Agriculture value percentage	0.175***	0.157***	0.168***	0.146***
	(0.0520)	(0.0496)	(0.0504)	(0.0449)
Services value percentage	0.036	0.027	0.044
	(0.0294)	(0.0257)	(0.0290)
Unemployment rate	−0.054	−0.031	−0.040
	(0.0392)	(0.0378)	(0.0385)
Strength of the legal rights index	0.005	−0.003	0.004
	(0.0085)	(0.0075)	(0.0084)
Constants	0.622***	0.619***	0.562***	0.592***
	(0.0721)	(0.0423)	(0.0690)	(0.0603)
N	4727	4727	4727	4727
R2	0.366	0.359	0.369	0.366
F	16.49***	16.49***	7.242***	7.508***

Note: Standard deviation in parentheses. Significance of standard errors ***p < 0.01, **p < 0.05, *p < 0.1.

4.2. The urbanization effect in classified countries

To test different impact of different countries' urbanization on vulnerable employment, countries should be divided into different groups. A single model may contain a number of linear equations due to group dividing. In such a model it is often unrealistic to expect that the equation errors would be uncorrelated. If OLS estimation is performed respectively, the standard errors of coefficients are too high. That is, only when we use GLS to estimate the equations and take the correlation of interference into consideration, can we get a more effective estimate. Generally, three methods are used to test the coefficient difference between groups, and they are Chow test, Permutation test, and test based on seemingly uncorrelated regression (SUR) model. The assumptions of SUR are looser: (1) it is not pre-defined that the coefficients of each variable in different groups must be the same; (2) the interference terms of the groups can have different distributions. We conducted a test based on the seemingly unrelated regression model of equations to make estimated coefficient of different country groups comparable. The SUR system is a set of equations with contemporaneous cross-equation error correlation (i.e., the error terms in the regression equations are correlated) (Zellner, 1962). The equations are apparently or “seemingly” unrelated regressions rather than independent relationships. Supposing that there are N equations (N explanatory variables), each has T observations, where T > N. In the ith equation are K_i explanatory variables. The ith equation can be expressed as y _i = X _i β _i + ε _i. These n equations can be compactly expressed as

$$y=\left[\begin{array}{c}{y}_1\\ y_2\\ ⋮\\ y_{\mathrm{N}}\end{array}\right]=\left[\begin{array}{cccc}{X}_1& 0& \dots & 0\\ 0& X_2& \dots & 0\\ ⋮& ⋮& \ddots & ⋮\\ 0& 0& ⋮& X_N\end{array}\right]\times \left[\begin{array}{c}{\beta }_1\\ {\beta }_2\\ ⋮\\ {\beta }_N\end{array}\right]+\left[\begin{array}{c}{\epsilon }_1\\ {\epsilon }_2\\ ⋮\\ {\epsilon }_N\end{array}\right]=\mathit{X\beta }+\epsilon$$ (2)

where β ₁ ∼ β _n are observations on an explanatory variable in the ith equation. The coefficients of each variable are not expected to be the same. The error terms of different groups can have different distributions, and can be correlated. In case of two equations, we find that ε ₁∼N(0,  σ ₁ ²), ε ₂∼N(0,  σ ₂ ²), and corr(ε ₁,  ε ₂) ≠ 0.

4.2.1. Results by country income group

To investigate the varying impacts of urbanization on vulnerable employment in different countries, we classified the 163 countries considered into four categories in accordance with income groups suggested by the World Bank. The four categories consisted of 53 high-income countries, 22 upper–middle-income countries, 45 lower–middle-income country, and 43 low-income countries. According to Eq. (2), in regression by group, the individual effect was first removed, and the seemingly unrelated model estimation and inter-group coefficient test were then performed on the transformed data.

Table 3 reports the results of the impact of urbanization on vulnerable employment by country income group. The F test is significant, and the R² fitting degree is >0.3. All results had negative and statistically significant coefficients of urbanization rate, which supports the main hypothesis. The influence of urbanization on vulnerable employment in countries with different income levels was remarkable. In particular for upper–middle-income and high-income countries, the negative effect of urbanization on vulnerable employment was pronounced. If the urbanization rate increased by 1 %, the vulnerable employment rate decreased by 0.181 % in upper–middle-income countries and 0.173 % in high-income countries, respectively. By contrast, for lower–middle-income and low-income countries, the increase in the urbanization rate had a smaller impact on the rate of vulnerable employment. It can be explained that the time consumption of 1 % urbanization rate does not equal in high-income countries and low-income countries. Because the urbanization rate not only denotes the proportion of urban residents, but also represents quality of development. In the urbanization process, with the increase of urbanization rate, there is a higher demand for urbanization quality, and the urbanization progress is more difficult. Therefore, high-income countries are likely to use more time to reach the 1 % increased urbanization rate than low-income countries. During the period, the effect of urbanization on vulnerable employment is more significant in high-income countries as is shown in the results. Besides, it also shows more significant improvement of employment quality in high-income countries, and implies that the reduction of the vulnerable employment rate in depth of urbanization is attributed to employment relationship changes rather than rural to urban migration only. To be more specific, in cases with the same degree of urbanization increasing, the mobility of employment in higher-income countries was greater, for their employment relationships were more sensitive to minor changes in urbanization, and the transformation between the employees and employers is more frequent. While, it is worth noting that although the quality of urbanization in low-income countries may not high, the rapid urbanization indeed drove plenty of vulnerable employment groups from rural areas in low-income countries to find more decent work (Grant, 2012; Sato & Zenou, 2015). Some of them were likely to get access to urban informal sector, and made employment relationship unchanged (e. g. from self-employed farmers to self-employed workers), which made absolute value of the coefficient lower and smaller negative impact than high-income countries. The ratio of services value added is significant in columns 2 and 3, the unemployment rate is significant in columns 2 and 4, and the strength of the legal rights index is significant in column 1 (Table 3). The algebraic symbols for the estimated coefficient of such control variables as the GDP per capita, ratio of agricultural employment, ratio of services value added, unemployment rate, and strength of the legal rights index remained constant both in the results of regression by country income group and by full-sample coefficients.

Table 3.

Results of panel regression by country income group.

	(1)	(2)	(3)	(4)
	High-income country	Upper–middle-income country	Lower–middle-income country	Low-income country
Urbanization rate	−0.173***	−0.181***	−0.099**	−0.090
	(0.0395)	(0.0261)	(0.0424)	(0.0546)
The GDP per capita	−0.004	−0.012***	−0.006	−0.027***
	(0.0036)	(0.0033)	(0.0045)	(0.0043)
Agriculture value percentage	0.708***	0.065*	0.259***	0.101***
	(0.1111)	(0.0338)	(0.0282)	(0.0158)
Services value percentage	0.015	0.068***	0.094***	0.011
	(0.0145)	(0.0204)	(0.0223)	(0.0190)
Unemployment rate	−0.0120	−0.107***	−0.026	−0.133**
	(0.0250)	(0.0272)	(0.0327)	(0.0608)
Strength of the legal rights index	0.037***	0.008	0.001	−0.012
	(0.0079)	(0.0061)	(0.0063)	(0.0086)
Constants	−0.002	0.009	0.015**	0.006
	(0.0046)	(0.0058)	(0.0063)	(0.0060)
N	1537	638	1305	1247
R2	0.307	0.395	0.427	0.447
F	19.56***	24.36***	26.55***	14.34***

Note: Standard deviation in parentheses. Significance of standard errors ***p < 0.01, **p < 0.05, *p < 0.1.

4.2.2. Results by country population

Population has an impact on the effect of scale for any given analysis. The path of development of a large country is often different from that of a small country (Chen et al., 2015). Based on a standard of 25 million, countries were divided into two categories: countries with a large population and those with a small population. There were 1228 samples with large population and 3499 samples with small population. According to Eq. (2), the regression by group was conducted by using SUR model estimation and the inter-group coefficient test to explore a general law of impact on different countries.

Table 4 compares the results by the population of the country. The F test is significant, and the R² fitting degree is 0.369, 0.580, 0.297, respectively. For two country categories, the estimated coefficient of the rate of urbanization relative of the rate of vulnerable employment remained negative, which indicates that urbanization had a negative effect on vulnerable employment. Comparing the four coefficients, it is clear that the influence of urbanization on vulnerable employment for populations of different sizes was different. The two coefficients of different categories were −0.246 and −0.127, respectively. Although the two coefficients were negative, their absolute values for countries with large ratios were larger. Vulnerable employment in countries with large populations was more responsive to urbanization, that is, the change in vulnerable employment in countries with smaller population was less affected by the same increment in the rate of urbanization. It illustrated those countries with different population size need to discuss separately due to their unequal impact. Other regression results of variables remained virtually unchanged.

Table 4.

Results of panel regression of countries by population.

	(1)	(2)	(3)
	Full-sample country	Country with large population	Country with small population
Urbanization rate	−0.168***	−0.246***	−0.127***
	(0.0159)	(0.0325)	(0.0160)
The GDP per capita	−0.015***	−0.027***	−0.010***
	(0.0018)	(0.0033)	(0.0019)
Agriculture value percentage	0.168***	0.141***	0.188***
	(0.0153)	(0.0201)	(0.0212)
Services value percentage	0.044***	0.128***	0.035***
	(0.0098)	(0.0244)	(0.0109)
Unemployment rate	−0.040***	−0.014	−0.051***
	(0.0150)	(0.0275)	(0.0175)
Strength of the legal rights index	0.004	0.002	0.005
	(0.0040)	(0.0067)	(0.0048)
Constants	0.003	0.009	0.004
	(0.0028)	(0.0058)	(0.0031)
N	4727	1228	3499
R2	0.369	0.580	0.297
F	80.77***	48.39***	42.98***

Note: Standard deviation in parentheses. Significance of standard errors ***p < 0.01, **p < 0.05, *p < 0.1.

4.3. Robustness checks

Robustness checks were carried out to ensure the reliability of the results of analysis, and to avoid the adverse effects of endogenous factors (Table 5 ). The interval of samples was appropriately determined to avoid the impact of external economic shocks. By taking the outbreak of the global financial crisis in 2008 as the dividing point, the sample time was divided into two periods: 1991–2008 (column 1) and 2009–2019 (column 2). Considering the impact of industry on vulnerable employment, we used variables to replace the ratios of employment of agriculture, forestry, and fishing, with value added, with the proportion of employment in agriculture (column 3), and replaced the ratio of services value added with the ratio of industry value added (column 4). Few changes were noted in the results in terms of the significance of the regression results, when setting time interval or alternative variables. Considering initial conditions issue or convergence issue, we randomly selected 80 % of the samples for data exploration in the robustness test, the results is shown in Table 5 (column 5).

Table 5.

Results of robustness checks.

	(1)	(2)	(3)	(4)	(5)	(6)		(7)
Urbanization rate	−0.222***	−0.249***	−0.082	−0.168***	−0.162***	−1.113*	Urbanization rate	−0.161***
	(0.0643)	(0.0832)	(0.0567)	(0.0599)	(0.0596)	(0.6362)		(0.0610)
The GDP per capita	−0.006	−0.018***	−0.006	−0.015**	−0.013**	0.000	L.lnagdp	−0.014**
	(0.0046)	(0.0070)	(0.0059)	(0.0061)	(0.0060)	(0.0165)		(0.0061)
Agriculture value percentage	0.057*	0.169**		0.124***	0.174***	0.1707*	L.ap	0.131***
	(0.0335)	(0.0785)		(0.0442)	(0.0502)	(0.0885)		(0.0437)
Services value percentage	0.025	0.039	0.003		0.038	0.014	L.ip	−0.037
	(0.0229)	(0.0271)	(0.0219)		(0.0276)	(0.0590)		(0.0294)
Unemployment rate	0.089**	−0.002	−0.040	−0.041	−0.040	0.201	L.ue	−0.033
	(0.0442)	(0.0472)	(0.0352)	(0.0385)	(0.0402)	(0.0874)		(0.0379)
Strength of the legal rights index			0.005	0.004	0.003	0.010	L.lnsl	0.005
			(0.0068)	(0.0084)	(0.0080)	(0.0160)		(0.0093)
Agriculture employment rate			0.387***
			(0.1004)
Industry value percentage				−0.045
				(0.0292)
Constants	0.548***	0.641***	0.373***	0.607***	0.548***	1.240***	Constants	0.592***
	(0.0538)	(0. 0746)	(0.0836)	(0.0610)	(0.0686)	(0.3232)		(0.0614)
R2	0.281	0.251	0.499	0.369	0.366	–	R2	0.363
N	2934	1793	4727	4727	3782	4720	N	4564
F	6.19***	9.03***	11.45***	7.24***	5.39***	731.92***	F	6.65***

Note: Standard deviation in parentheses. Significance of standard errors ***p < 0.01, **p < 0.05, *p < 0.1. L.lnagdp, L.ap, L.ip, L.ue, L.lnsl are representing corresponding variables lagged by a period, respectively.

This paper mainly explores the impact of urbanization on vulnerable employment, however, using the above regression model to identify causality may fall into endogenous problems. In terms of the study, endogeneity mainly comes from the following sources: (1) reverse causality. Countries with low vulnerable employment rate have experienced formalization and provide more formal jobs, attracting agricultural population to urban areas, resulting in the underestimate of the regression coefficient. (2) Missing variable. There may be factors that are difficult to observe and are related to vulnerable employment rate and urbanization rate. For example, the country's institutional environment leads to inaccurate estimation of regression coefficient. Generally speaking, the direction of error in regression coefficient estimation is not clear in theory. In order to solve the above endogenous problems, this study uses life expectation at birth as instrumental variable. The random physiological character is related to urbanization and has nothing to do with the vulnerable employment for more health services and caring centers in urban systems. Besides, the urbanization can be postulated to have resultant returns on the economic growth, fostering better income, and better income can be linked to improved health outcomes (Donald et al., 2019). In first-stage regression, the coefficient of life expectation at birth is significantly 0.002 at 1 % level, and the standard error is 0.0002. In Stock and Yogo types of testing, minimum eigenvalue statistic is equal to 127.807, and the original hypothesis is rejected. So, it's considered that there exists no weak instrument variable. Column 6 in Table 5 reports the regression results of the two-stage least squares method of instrumental variables. The lagging variables are common way to alleviate endogenous problems in robust test. When all control variables lagged by a period, the algebraic symbol of their coefficients in the results of regression remained unchanged (column 7). The coefficients and symbols of all variables are stable compared with results in Table 1, and the estimation results (based on two-way fixed effect model) conform to expectation, which do support our main hypothesis.

5. Discussion of findings

It's proved that urbanization has positive impact on employment from macro data. Urbanization can reduce the rate of vulnerable employment and improve the quality of employment. However, there are more details for further discussion. One is the understanding of micro spatial scale. It is difficult to see the employment space competition within the city only from the comparison on macroeconomic data, but gentrification and employment vulnerability in urban centers are worth discussing. The other is the awareness of new trend. In recent years, the flexible employment has been redefined along with the “Internet+” new economy and new business format. It is necessary for us to discuss the relationship between new economy and flexible employment, so as to better understand the impact of urbanization on employment.

5.1. Promoting effect of urbanization

Overall, urbanization has a positive impact on reducing the vulnerable employment rate and helps to achieve the sustainable development goal of decent work and economic growth. Although rapid urbanization is associated productivity barriers such as congestion and environmental degradation, research has shown that workers in urban areas have higher productivity and higher income than those in rural areas (Bloom et al., 2008). There remains much potential for rural-urban migration in developing countries, and these countries should make good use of the potential of labor supply and continue to promote urbanization. While, the organization and structure of urban industry need to be improved to create more formal opportunities for workers in developed countries, which benefits employment quality.

5.2. Gentrification and employment vulnerability

From the perspective of the city development in the world, there will be social differentiation in the central area of the city. The reinvestment of inner neighborhoods makes the transition from lower to higher social-economic status residents in the city center (Shaw, 2008). Although this study does not focus on the space and land use, it's necessary to know that the impact of urbanization on vulnerable employment groups is not only the downward trend, but also the spatial competition within the city. The higher expel and replace the lower in the central urban area at the stage of in-depth urbanization, thus changing the urban land use and forming social spatial differentiation (Bates, 2013). The spatial competition and differentiation are also considered the effects of urbanization on vulnerable groups in microscopic view.

5.3. New economy and flexible employment

Along with the “Internet+” new economy and new business format, the flexible employment, which is considered to be informal employment and non-standard employment, has been redefined (Wan et al., 2022; Wei, 2019). For example, the emergence of freelance, short-term employment contracts results in vulnerable employments in service sector in urban contexts. Although it is sorted as “ vulnerable employment” in data statistics, it is actually an important form of flexible employment. The study has proved the overall relationship between urbanization and vulnerable employment, but failed to subdivide vulnerable employment into industries due to data limitations. There is not enough evidence to support that the flexible employment mode under the new economy is a temporary phenomenon or trend of urbanization.

6. Conclusions

Humanity is undergoing a dramatic shift from rural to urban living, where the global rate of urbanization is above 50 %, and will continue to rise. It's questioned that vulnerable employment would consistently go down and the social equality would come true along the way. Based on an analysis of data from 163 countries from 1991 to 2019, this study first examined the joint spatial and temporal evolution of urbanization and vulnerable employment, and then discussed about the correlation of both. A panel regression model was formulated to explore the general law for the relationship, and further group regression by country-based income groups and population sizes was carried out. The theoretical hypothesis proposed, was finally verified – urbanization reduces vulnerable employment on the whole, but its influence on different countries is different.

A high correlation has been observed between the rates of urbanization and vulnerable employment, and a clear quantitative relationship has been noted. From a temporal trend, as urbanization rate rises, vulnerable employment falls. Since 1991, global urbanization has exhibited a general upward trend whereas vulnerable employment has shown an overall downward trend. From the perspective of spatial patterns, the rates of urbanization in Europe, America, and Oceania have been higher than those in Asian and African countries. Vulnerable employment shows a downward trend with an increase in the urbanization. It's not difficult to find a roughly negative relationship between urbanization and vulnerable employment when comparing the figures in separate years 1991 and 2019. Using linear fitting, correspondence between urbanization rate and vulnerable employment rate could be presented clearly. When the urbanization rate is 50 %, the rate of vulnerable employment is about 45 %.

Based on the above theoretical analysis and empirical tests, we conclude that urbanization has a negative effect on vulnerable employment. If the urbanization rate increases by 1 %, the rate of vulnerable employment decreased by 0.168 %. Urbanization affects the rate of vulnerable employment by promoting industrial transformation between the urban and rural sectors as well as changes in employment relationships in the urban sector. Countries with different income groups or populations have reacted differently to the rise of urbanization. Specifically, vulnerable employment in higher-income countries or larger population size is much more acutely affected by urbanization. Higher-income countries are more sensitive to the rise of urbanization, and vulnerable employment in countries with larger populations is more clearly influenced by it.

We look forward to collecting more detailed statistics on employment worldwide, which is required to be comparable among different countries. More detailed statistics, such as informal employment rate and vulnerable employment rate of various industries in all countries will provide support for the study of world development. In the present situation, vulnerable employment data available cannot distinguish between urban and rural sectors and is not exactly equal to employment vulnerability. However, the paper, which makes full use of data available and keeps looking for theoretical support in the literature, balances scientificity and practicability in the study. In the sense, the paper provides evidence for the conclusion and sights into the relationship between urbanization and vulnerable employment as far as possible. In the future, official international organizations, such as ILO, can further consider collecting these detailed employment data to support further research, while relevant individuals and union can also share data on the online sharing platform to improve the utilization of research data.

Although this study provides information to relate urbanization to vulnerable employment, it has a few limitations. The relationship between the rates of urbanization and vulnerable employment showed a negative correlation here, implying that the rate of vulnerable employment decreases as urbanization increases. On the whole, high-income countries show high urbanization rates and low vulnerable employment rates, while low-income countries show low urbanization rates and high rates of vulnerable employment. Note that urbanization and vulnerable employment for some countries are not coordinated. The first typical case featured countries with high urbanization rates and high vulnerable employment rates, where the latter did not decrease with an attendant rise in the former. This might have occurred for the following reasons: First, a large number of people are engaged in agriculture owing to intensive farming. Second, urbanization has occurred more quickly than the transformation of the employment structure. Third, in some cases, low rates of urbanization and vulnerable employment may be related to the traditional agricultural management of the given country. It would be a challenge for latecomers to do more detailed work such as subdividing vulnerable workers into categories according to countries in this manner for further research. Furthermore, it is noticeable that the global three-year pandemic has caused very serious impact on the vulnerable employment. So this paper has done the basic research before the pandemic, which is meaningful for related research during the pandemic, and also an important research direction in the future.

CRediT authorship contribution statement

Chen Mingxing: Conceptualization, Methodology, Supervision, Funding acquisition, Writing - Review & Editing. Huang Xinrong: Data curation, Visualization, Formal analysis, Methodology, Writing-Original draft preparation, Writing-Review & Editing. Cheng Jiafan: Methodology, Writing-Review & Editing. Tang Zhipeng: Methodology, Writing - Review & Editing, Supervision. Huang Gengzhi: Supervision.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A.

Panel unit root and panel co-integration checks. To avoid spurious regression, we tested for the existence of the unit roots of all variables. The LLC, IPS, Fisher-ADF, and Fisher-PP panel unit root tests were used to determine the stability of the variables. Table 6 shows that three variables were non-stationary but that all variables became stationary at a 1 % significance level, which led us to reject the null hypothesis of an existing unit root after considering the first differences. This implies that all variables were stationary in the first difference, and that a long-term equilibrium could be sustained among all variables.

Table 6.

Results of panel unit root test.

	LLC	IPS	Fisher-ADF	Fisher-PP
vulnerable employment rate (ve)	−5.483***	−2.902***	−1.906	−2.409
Urbanization rate (ue)	−14.110***	11.039	3.478***	72.182***
The GDP per capita (lnagdp)	−4.761***	−9.463**	−5.424	−5.984
Agriculture value percentage (ap)	−6.819***	−9.683***	13.895***	12.984***
Services value percentage (sp)	−8.483***	−10.383***	7.775***	6.551***
Unemployment rate (ue)	−11.958***	−5.532***	9.374***	−0.508
Strength of the legal rights index (lnsl)	58.167	42.557	−11.063	−11.409***
∆ve	−16.014***	−27.999***	52.307***	98.269***
∆u	−7.661***	−4.365***	3.129***	7.786***
∆lnagdp	−22.239***	−32.756***	57.698***	93.047***
∆ap	−25.094***	−38.176***	91.835***	168.674***
∆sp	−24.740***	−37.807***	81.858***	153.696***
∆ue	−20.523***	−26.196***	48.316***	72.728***
∆lnsl	3.072	−18.801***	10.103***	28.351***

Note: Significance of standard errors ***p < 0.01, **p < 0.05, *p < 0.1.

If all variables remained non-stationary until the first-order difference, this analysis proceeded with the cointegration test. We used the cointegration proposed by the Kao, Pedroni, and Westerlund panel to determine whether the panel data had the cointegration relationship. The results in Table 7 show that all statistics in all methods rejected the null hypothesis of no cointegration at a 5 % significance level, which means that a constant, long-term equilibrium relationship obtained among urbanization, vulnerable employment, economic development, agricultural level, modernization level, employment environment, and loan convenience from 1990 to 2019.

Table 7.

Results of panel co-integration test.

		Statistic	p-Value
Kao	Modified Dickey–Fuller t	6.2631	0.0000
	Dickey–Fuller t	6.2073	0.0000
	Augmented Dickey–Fuller t	4.187	0.0000
	Unadjusted modified Dickey–Fuller t	6.4344	0.0000
	Unadjusted Dickey–Fuller t	6.4243	0.0000
Pedroni	Modified Phillips–Peron t	14.4032	0.0000
	Phillips–Perron t	−1.7542	0.0397
	Augmented Dickey–Fuller t	−1.8538	0.0319
Westerlund	Variance ratio	2.3638	0.0090

Data availability

Data will be made available on request.