Poverty and labor informality in Colombia

There is much discussion about the measurement of informality. Different approaches adopt distinct criteria to identify which people work in the informal sector or have an informal job. Some of the most common perspectives used to study informality are the legal records or tax payments of businesses, the number of workers hired by an establishment, different job features (occupational position, nature of labor relations, levels of productivity), and compliance with labor laws and regulations.

In 2003, the International Conference of Labour Statisticians (ICLS) proposed a specific way to measure informality that involved two different approaches: measuring employment in the informal sector and measuring informal employment (ICLS, 2003; Hussmanns, 2005). This is the perspective adopted in this article based on the microdata of the National Continuous Survey and the Integrated Household Survey.

The National Continuous Survey was conducted from 2002 to 2005 and the Integrated Household Survey from 2008 to 2013. The two surveys have different formats, methodologies, and populations of reference. Therefore, throughout the article, the examined period is divided into two (2002–2005 and 2008–2013) based on the survey used.

On the one hand, wage-earning workers in a small establishment (five or fewer employees), domestic workers, unpaid workers, non-professional own-account workers, and non-professional employers of a small establishment comprise employment in the informal sector (EIS). Professional self-employed individuals who have more than 15 years of education are excluded because they tend to engage in the formal economy if they want to work.

On the other hand, all wage earners and unpaid workers whose employer does not obey the labor laws comprise informal employment (IE). These workers do not have access to labor rights. In this article, as in the majority of related studies, the proxy variable for identifying informal employment is the receipt of payment (or not) from the pension social system.

Specifically, the following question is considered: “Are you currently paying into the pension system?” If the answer is “no,” that wage earner worker has informal employment.

This perspective was developed by ICLS (2003) and explained by Hussmanns (2005). The figures for Colombia in 2013 are shown in Table 1, which shows that 57.5% of urban Colombian workers were working in the informal sector and 59.6% had informal employment. Self-employed accounted for more than 60% of informality, and 30.8% of workers worked in the formal sector with formal jobs. As found in other studies, the extent of informal employment was higher than the extent of labor in the informal sector in most Latin American countries (Maurizio, 2015). This means that if only one perspective of measurement is taken into account, the level of informality could be over- or under-estimated. This article adopts both perspectives (EIS and IE) to carefully calculate the different effects of changes in the two types of employment on the reduction in monetary poverty.

There exists a debate about how poverty can be accurately measured. A simple way of identifying poverty is to check whether a household does not have enough income to stay above a certain threshold. This methodology considers “unidimensional” indicators, using only one variable to classify people as poor or not poor. Other recent approaches take into account several variables to identify poverty. However, given the aim of this article, monetary poverty is chosen as the measurement of poverty. In this paper, a household is considered poor if its household income per capita (HIPC) is lower than the poverty threshold estimated by the National Administrative Department of Statistics of Colombia

Importantly, no data are available for 2006 and 2007 because in those years, a change in the survey was made. A commission of several institutions merged data from different years.

This study estimates the most common poverty indicators, that is, the Foster-Greer-Thorbecke (FGT) (1984) indicators. These indicators are calculated using the poverty threshold, the level of HIPC, and the different levels of poverty aversion:

where N is the total population, x_i is the HIPC for i, z_i is the poverty threshold corresponding to i,

There are several poverty thresholds (z_i) depending on the region and geographical references.

Poverty decreased in Colombia during the first decade of the twenty-first century. The poverty headcount ratio was reduced from 46.3% in 2003 to 27.2% in 2013. This trend is similar to that of Latin American countries, whose poverty levels were reduced by approximately 40% in the same period (ECLAC, 2014a). Some countries were more successful than others in the fight against monetary poverty: Argentina, Peru, and Uruguay reduced their poverty by more than 60% . In Colombia, the level of informality decreased slightly, but as shown in Figure 2, it increased in some years of the period analyzed.

Between 2005 and 2008, there is a different trend in the level of informality that may be the result of the change in the survey in those years.

Figure 2

Therefore, poverty and informality did not change to the same extent, and that could be the result of different links between low income and the labor market. Figure 3 shows changes in poverty and informality. Poverty decreased more than 40% when every indicator is considered, and the decrease in poverty severity is even higher than the decrease in the percentage of the poor population. However, throughout the period analyzed, the informal urban sector represented a labor problem that did not change noticeably. The reduction in informal sector employment was 5%, whereas the reduction in informal employment was 11%. This suggests a certain lack of connection between the improvements in the quality of living and the persistence of labor market problems, such as informality.

Figure 3

Figures 4 and 5 show the earning differentials between formal and informal workers (EIS/IE). Kernel-based nonparametric estimation methods are used to determine the monthly and hourly earnings of formal and informal workers. There is a close relationship between the two informality approaches: the mean income is similar. The same situation occurs with formal workers (EFS/FE); their mean labor income is similar in both definitions of formality but higher for workers with formal employment. In addition, there is a bunching around the minimum wage for formal workers. Both Figures 4 and 5 illustrate that there is a critical gap between formal and informal workers, and the latter have higher dispersion. This dispersion is seen in monthly and hourly earnings, but it is greater for monthly earnings because of the dispersion of the number of hours in the jobs of informal workers.

Figure 4

Figure 5

A primary conclusion from the figures is that most workers with the lowest income have informal employment or are working in the informal sector. However, this is not evidence of segmentation, nor does it mean that the difference is explained by informality. It could be the consequence of low skills or personal characteristics related to low productivity. If this were true, the gap between formal and informal workers could partially be explained by segmentation, as will be shown in the next section of the article.

The gap between formal and informal workers depends on, among other factors, the kind of informality and occupational position. Table 2 shows that workers with the lowest average income are informal self-employed and informal wage earners who are working in the informal sector; both groups have mean incomes equivalent to 54% of the mean income of total workers. Despite the low average incomes of informal workers, the figures demonstrate that there are large differences within informality. Even in the formal sector, there are many employees with low incomes, and the main references are those who are working in that sector but lack coverage by labor law, whose incomes are 27% lower than the average. Overall, it is possible to conclude that even though there are heterogeneities by occupational position, informal sector employment and informal employment are sources of low earnings.

This section presents the estimation of individual earnings gaps, which are measured based on Jenkins’ methodology (1994). This methodology consists of a counterfactual analysis that estimates what the earnings of informal workers would be if they were working in the formal sector or had formal employment. This methodological approach goes further than traditional estimations of Mincer’s functions because after the econometric predictions for formal and informal workers are performed, a detailed set of individual gaps is calculated for every individual. This means that workers with different personal characteristics would have different gaps (Beccaria and Groisman, 2008; Jenkins, 1994).

The first step is to estimate earnings functions (monthly and hourly), such as those presented in equation (2). They are estimated separately for formal (F) and informal (I) workers (J = I,F). The same estimations are carried out for the two types of informality (EIS and IE) to compare the gaps when different definitions of informality are considered.

The xi,J'$x_{i,J}^{'}$vectors comprise the personal characteristics of workers that influence their level of earnings, such as age (experience), education (non-lineal relation), gender and relationship to the head of household, and sector of the job and geographical control variables. Each equation estimates a vector of coefficients. If there is any difference between the two vectors, it could be quantitative evidence of labor segmentation because one group (formal–informal) does not have the same treatment as the other.

Those estimations assume a conditional mean of zero (E[e_i,J |x_J ]= 0), but this is not plausible due to the sample selection bias. Because the most important result is the coefficient vector, unbiased parameters need to be warranted. To correct the potential sample selection bias in equation (2), Heckman’s methodology (1979) is included in the estimations. In this application, the sample selection bias correction is different from the traditional methodology for two reasons: first, because, when Mincer’s function is estimated for formal (informal) workers, there are two groups that are excluded from the estimations: people who are working in the informal (formal) sector and people who are not working; second, because, differently from the case in which the estimations are performed by gender, there is a probability that informal/formal workers could change sector and potential for self-selection in the decision to work in the informal/formal sector.

Those estimations assume a conditional mean of zero (E[e_i,J |x_J ]= 0), but this is not plausible due to the sample selection bias. Because the most important result is the coefficient vector, unbiased parameters need to be warranted. To correct the potential sample selection bias in equation (2), Heckman’s methodology (1979) is included in the estimations. In this application, the sample selection bias correction is different from the traditional methodology for two reasons: first, because, when Mincer’s function is estimated for formal (informal) workers, there are two groups that are excluded from the estimations: people who are working in the informal (formal) sector and people who are not working; second, because, differently from the case in which the estimations are performed by gender, there is a probability that informal/formal workers could change sector and potential for self-selection in the decision to work in the informal/formal sector.

Consequently, there are three labor market statuses: working in the informal sector (EIS and IE) (y_1i = 1), working in the formal sector (y_2i = 1), and out of the labor force (y_3i = 1). The first step of the Heckman methodology is a multinomial logit selection model (3); the s_i vector includes variables that affect the likelihood of being in each of the mentioned labor market statuses (y_ji), which are the same personal characteristics used in equation (2), plus a dummy variable indicating whether the household has children and the marital status. This has been proposed by several studies as an alternative when the selection bias is not bivariate (Bourguignon et al., 2007; Dahl, 2002; Lee, 1983) and it was applied by Pradhan and van Soest (1995) to estimate the formal and informal labor earnings of Bolivia.

Based on the results of equation (3), Lee (1983) proposed to calculate the Mills ratio with the cumulative distribution function and the standard normal density. The second step is shown in equation (4), which contains λ^i,J,${{\hat{\lambda }}_{i,J}},$the inverse Mills ratio, the coefficient (ψ_J) of which establishes the direction and significance of the selection bias.

Then, there are two incomes available for every informal worker: first, the predicted earnings and, second, the simulated earnings, which are equal to the earnings of an informal worker if he/she were paid as a formal worker. Equation (5) illustrates the individual gap for each informal (I) worker i:

The estimated labor income for each informal worker (ŵ_i,I) is the result of estimating the income function (4), whose sample selection bias is corrected by the Heckman two-step meth(r^i,I)$\left( {{{\hat{r}}}_{i,I}} \right)$odology. The counterfactual income for each informal worker is calculated considering personal and occupational characteristics and also considering the earning structure of formal workers (β_F). It is important to examine how the inverse Mills ratio is calculated. Because some of the individual gaps for informal workers could be the result of differences in the selection process, they are supposed to have formal workers’ sample selection when the counterfactual income for informal workers is analyzed:

Equation (6) shows how the counterfactual income for each informal worker i is calculated. The personal gap is the result of comparing the estimated and the counterfactual income. Consequently, there is a vector with the distribution of the gap. It is important to have complete information for every worker, instead of an average gap, to perform an analysis of different populations and occupational groups. Based on that estimation, in the next section, quantitative evidence of the relation between poverty and informality is shown through microsimulations of incomes for informal workers and changing rates of informality.

Table 3 shows the average individual earnings gaps associated with informality considering both approaches to informality, IE and EIS, and hourly and monthly earnings for the period from 2002 to 2013. The gap is higher when monthly earnings and the IE approach are considered. On average, the monthly individual gap for informal employment is 43.6%, whereas the gap for the informal sector is 42.3%. This means that the same informal worker would noticeably increase his/her labor earnings if he/she were employed in the formal sector or had informal employment, or at least if the structure of earnings were the same as that of better jobs. The figures illustrate a trend of a reduction in the gap between 2002 and 2005 and a steady leveling of the average gap from 2008 to 2013.

These gaps are, partially, empirical evidence of segmentation by informality type in the Colombian labor market. The gaps are not the result of personal productivity but are the consequence of sectoral differences and the type of employment. Many other studies have found the same result using different methodological approaches and examining several countries (Appleton et al., 2005; Beccaria and Groisman, 2008; Charmes, 2009; Maurizio, 2013). Therefore, informality, which includes several types of workers, such as those working in small establishments, those with nonregulated labor relations, or those with family occupations, is a source of low incomes that cannot provide resources to overcome monetary poverty.

In this approach, the methodology is static, which means that the analysis considers only the last year in the study period (2013). This estimation tries to quantify the potential effect of formalization. Nevertheless, it is not a real scenario if informality does not exist, but this simulation should be interpreted as an analytical exercise with the intention of analyzing the quantitative importance of the relationship between poverty and labor informality (Beccaria and Groisman, 2008; Maurizio, 2015).

The estimation takes into account the individual gaps associated with informality, as shown in the previous section of the paper. Based on those results, the exercise is to calculate the simulation of household income per capita (HIPC) if informal workers were paid as if they were formal workers. The estimated (HIPCEsth,)$\left( HIPC_{Est}^{h}, \right)$and the simulated (HIPCSimh)$\left( HIPC_{Sim}^{h} \right)$HIPC for household h are given by

where w^ih$\hat{w}_{i}^{h}$is the estimated labor income for formal worker i from household h;w^jh$h;\hat{w}_{j}^{h}$is the estimated labor income for informal worker j from household h;NLkh$h;NL_{k}^{h}$is the nonlabor income for member k from household h; F and I are the number of formal and informal workers in each household h; M_h is the number of members in household h; and r^jh$\hat{r}_{j}^{h}$is the simulated income for informal worker j from household h.

The formalization effect (exclusion of individual earnings gaps by informality) on poverty indicators is equivalent to the difference between the indicator calculated with HIPCSimh$HIPC_{Sim}^{h}$and HIPCEsth,$HIPC_{Est}^{h},$keeping the same threshold of poverty z_i:

The earnings gaps between formal and informal workers are related to the low incomes of poor households. Table 5 shows that if workers with informal employment were paid as if they were formal workers, keeping the same personal and productivity characteristics, the extent of poverty would decrease by 44.5% when monthly labor income is considered. A higher formalization effect is found when gap and severity poverty indicators are analyzed; those indicators would be reduced by 51%.

The higher effect on the gap and severity of poverty is the evidence that workers who belong to poor households have informal employment, and they are people whose households have the greatest vulnerability and the lowest income. Therefore, improving labor market conditions (formalization process) would be a key factor in decreasing the level of poverty in the country. The estimated effects are consistent with the average gaps associated with informality, which are shown in Table 3; this means that the higher the gaps are, the larger the formalization effect is.

Figure 6 shows the formalization effect in a situation where continuous change in informality is simulated. The initial situation is characterized by the actual informality rate (59.6%) and poverty indicator in 2013 (28.4 headcount ratio, 10.5 poverty gap, and 5.5 severity); it is shown at the top of the figure. When informality is eliminated, the x-axis is zero and the effects on poverty indicators are changed, as shown in Table 5. In the middle, there is also a continuous change in poverty indicators when informality changes. As expected, the greater the reduction in informality, the higher the effect on the reduction in poverty. The order of formalization depends on the likelihood of a person being an informal worker: the beginning of the effect is calculated based on people who have a higher possibility of being formal workers. Once the order of formalization is organized (based on a probit model), the formalization effect is the elimination of the individual earnings gap; then, the new income is estimated for every worker, and the HIPC is simulated for each household. Based on that income, the new poverty indicators are calculated.

Figure 6

Initially, the reduction in informality is more effective in reducing the headcount ratio than in reducing the gap and severity. This is because workers who have a lower likelihood of being informal workers also have a smaller individual earnings gap associated with informality.

This methodology implies that the beginning of the process is more costly because workers who have a lower likelihood of being informal have higher labor incomes even though they have a smaller gap.

Notably, this approach is a partial equilibrium estimation (ceteris paribus), and there are several aspects that could affect the figures shown in this exercise but are not included in this methodology.

A helpful comment from an anonymous referee suggested the importance of explaining in greater depth this limitation of the empirical exercise that is developed in the article, which similar studies have had to face and recognize.

Boeri and Garibaldi (2007) consider that stronger control over informality does not necessarily reduce it; on the contrary, such a policy could boost the unemployment rate. Based on their general equilibrium model, the authors conclude that the potential effect of formalization is restricted by the macroeconomic environment and changes in productivity (Boeri and Garibaldi, 2007). In addition, informality would not be reduced automatically even if there were economic growth; rather, it should be promoted with labor and redistribution policy.

The results analyzed in this section show the potential impact of reductions in informality and in the earnings gaps associated with informality on poverty indicators. However, poverty and low incomes depend on many other factors, such as quality of employment, education, economic structure, and the health system. This indicates that even if there were no informal workers, poverty would be persistently high, with a headcount ratio of 15.8% and a poverty gap of 5.1.

This section presents the actual effect of changes in informality (rate and gaps) on poverty through microsimulations. The starting point is the change in the rates of informality and poverty (Figure 2) and in the earnings gaps by informality type (Table 3). The estimates are performed considering two periods, 2002–2005 (National Continuous Survey) and 2008–2013 (Integrated Household Survey). Poverty indicators are estimated taking into account three types of household incomes

Nonlabor incomes are not changed.

In this equation (10), there is a temporal component (t) because the aim is to compare the rate and the gap associated with informality between two moments in time (t and t−1). The main difference between the definition of HIPC in equations (7) and (10) is that the estimation of labor earnings (w^j,πt−1h)$\left( \hat{w}_{j,{{\pi }_{t-1}}}^{h} \right)$for each informal worker j of household h is performed with the penalty of informality in the initial year (π_t-1).

To estimate the effect of changes in the informality rate on the extent of poverty through the increase in labor incomes, the HIPC is calculated as in equation (10) for the year t but simulating the degree of informality from the initial year (t−1). Informal workers are classified by estimating a probit model that estimates the likelihood of a person being an informal worker. Based on the results of the estimated probability, workers are organized depending on whether they were informal in 2005 and 2013 (final years). Because Colombia’s informality rate declined between the periods analyzed, the classification is from formal to informal workers until the informality rate of the initial year of comparison is achieved (t−1).

If the informality rate had increased, the simulation would have changed the classification of workers from informal to formal.

The total effect of changes in informality on poverty is the sum of two variations: the gap associated with informality (11) and the informality rate (12). This approach aims to measure how informality affects the extent of poverty; however, this is just a partial perspective, and several essential changes in the labor market, such as labor participation, unemployment rate, average change in labor incomes, returns of education and experience, or nonlabor incomes, have not been analyzed. As a consequence, a large change in poverty is not explained by informality but is the result of factors that are not considered in these estimates.

Table 6 shows the results of microsimulations that measure the actual effect of changes in informality on poverty indicators in Colombia from 2002 to 2013, considering two different periods (2002–2005, 2008–2013).

The long period is divided into shorter years because two household surveys are considered.

Between 2002 and 2005, when poverty reduced annually much less than in the next period (2008–2013), the role of informality in the reduction was more important; however, the stronger role was partially a consequence of the small decrease in poverty indicators. In that period, the informality rate decreased between 3% and 4%, whereas the informality earnings gap in monthly income reduced by 3%; these changes affected the poverty indicator to some extent. Specifically, the effect was 22.8% for the headcount ratio and was less notable for the other indicators, with 19.8% for the poverty gap and 17.9% for poverty severity.

Although the gap and severity of poverty were the poverty indicators with higher reductions, the influence of informality was smaller for these indicators. Changes in informality more strongly affected the labor incomes of workers whose household income was not far from the threshold of poverty; that is why informality was more important for the reduction in the headcount ratio than for the reduction in poverty severity.

The persistence of labor informality in a context of poverty reduction and acceptable economic growth indicates that improvements in the quality of life of people whose incomes are low have been the consequence of different factors, but the problem remains that their labor conditions have not improved. In addition, the trend shows the difficulties of achieving better labor conditions through formalization. This does not mean that the earnings of informal workers have not increased, but they could have increased as a result of several factors, such as changes in personal characteristics and remuneration. It is important to emphasize that even though thousands of informal workers have overcome monetary poverty, they still have a large earnings gap in comparison with formal workers, and for the majority of them, their labor conditions have not changed.