Salary Costs of Employment Vulnerability in Benin

Mincer argued that the positive relationship between an individual's education and subsequent income can be understood to reflect the effects of increased education on productivity [6]. This relationship is not straightforward or simple. Schooling and education are not synonymous, the educational content of time spent in school varies from superb to miserable. It is therefore not surprising that there are correlations between the level of education, measured in years spent in school, and the income of individuals, although positive, are relatively weak. However, when one takes the average of the wages on the groups of people different in their levels of education, the clear and strong differences emerge. This definitively established human capital as a determinant of remuneration. Human capital can be understood in two ways, either through education or through experience. There is indeed a close relationship between education and experience in that it is possible to make a trade-off between spending time in school and spending time working to gain experience. This link therefore underlines the obvious fact that from a theoretical point of view, experience is also a determinant of both employment and salary. The idea that experience also determines pay was taken into account in salary equation. It should be noted that this analysis of has remained in the logic of neoclassical theory by treating the formation of wages within a certain framework. This is only true when the labour market operates under the assumptions of competition. But the segmentation theory developed has shown that several employment segments can coexist in the same market, obeying different operating modes. Thus, the secondary segments are associated with vulnerable and poorly protected jobs. Labour supply decisions are based on the trade-off between labour income and the cost of that same labour. Thus, at the optimum, the wage differences correspond to the marginal rate of substitution between consumption and working conditions. On the other hand, the firm's decision to request is based on the trade- off between the need to pay compensation to workers in relation to the arduous or dangerous nature of the tasks to be performed and the need to improve working conditions. Developed the theory of hedonic prices aims to account for the wage differentials between individuals of the same qualification [7]. Hedonic prices are the implicit prices revealed to economic agents through the observed prices of differentiated products with specific characteristics associated with them. The basic idea of this theory is that jobs that are associated with poor working conditions compensate for this disadvantage by increasing the wage rate. The heterogeneity of wages can then be illustrated based on jobs with varying degrees of hardship. According to his preferences, an economic agent is ready to accept a riskier job if it is compensated by a high remuneration [8]. The arduous nature of the work has a very broad connotation because it encompasses aspects such as work accidents, occupational diseases, and a degraded work environment. The trade-off is therefore done at this level, between the risks associated with the job and the level of remuneration. If the worker receives non-monetary benefits from his employer (holidays, pensions, etc.), his salary may turn out to be low. Thus, in this type of job, the economic agent seeks to maximize his utility by minimizing the arduousness of his work. In this new demand system, goods are of three kinds: leisure, non-monetary characteristics of employment, and marketable goods that can be consumed. Companies, for their part, are making a trade-off between the costs of equipment that would provide safer working conditions on the one hand and the bonuses and benefits they must grant to workers who accept these risky jobs on the other hand. Under certain assumptions (homogeneity of individuals, heterogeneity of work environments, perfect information in relation to wages and working conditions, perfect mobility on the labour market), wages differ between workers but all have the same utility. Thus, to encourage workers to agree to work in poor conditions, companies must offer high wages. The relaxation of the assumption of homogeneity of worker preferences certainly introduces uncertainty into this reasoning.

Unacceptable forms of work are described as including conditions which deny fundamental principles and rights at work, endanger the life, health, liberty, human dignity and safety of workers or keep households in conditions of poverty [9]. The issue of relative vulnerability, which is the comparison of the conditions faced by different groups of workers competing in the same labour market, has generated wide scientific interest. Differential vulnerability in the labour market may be based on education, age, family responsibility, occupation, industry, well-being, and the absence or presence of labour market protections [10,11]. The notion of vulnerability thus refers to several aspects in the economic literature. Specifically, job vulnerability is linked both to intrinsic aspects of the job and to factors relating to the workers themselves. It is obvious that a one-dimensional measure of vulnerability cannot adequately capture the situation of workers. The most satisfactory approach to measuring vulnerability is therefore multidimensional. Vulnerable employment can be identified by reference to certain characteristics of employment linked to the risk of lack of decent employment [12]. Given the plurality of approaches to measures of employment vulnerability in the empirical literature, we present in Table 1 a summary of the dimensions generally encountered. Analysis of the table shows that the vulnerability measures generally take into account the following elements: working conditions, working environment, remuneration and protection of working rights (Table 1).

This article has two purposes. Analyse the effect of vulnerability on wages and test the hypothesis of compensating differences.

Construction of an employment vulnerability index for Benin (IVEB)

The notion of job vulnerability is difficult to measure from just one aspect. An overview of the literature shows that it is a multidimensional notion even if the dimensions it incorporates are not universal. The approach adopted here follows this logic. The reasons for the choice of dimensions are obvious. It is first and foremost a question of statistical availability and then aspects of interest which are the subject of the study. The variables we have chosen for the vulnerability measure are quite similar to those used. These dimensions are in line with a definition of employment vulnerability which takes into account the aspects that can constitute a source of employment instability.

The dimensions selected for the Benin employment vulnerability index (IVEB)

The measurement of vulnerability does not follow a unique approach in the literature. Studies thus use different dimensions even if certain dimensions can often appear in several studies. Two reasons fundamentally justify this plurality of approach in the literature. The first is that important vulnerabilities in one environment may not have the same significance in other environments. The second reason is linked to data limitations which make it not possible to follow the same approaches to measuring vulnerability from one country to another. Despite this plurality of approach, a criterion must nevertheless guide the choice of dimensions / variables to be taken into account in measuring job vulnerability. Indeed, this criterion relates to the fact that the variables selected must be clearly defined, well measured and verifiable if necessary. In view of these different aspects, we have retained three dimensions for 14 variables in the measurement of vulnerability. Elements retained in these different dimensions are found in the variables retained. Thus, the dimensions retained for the IVEB are as follows:

First dimension: working conditions

The first dimension taken into account relates to working conditions. It takes into account the various risks to which employees are directly exposed at their workplace and within the framework of their professional obligations. This dimension includes eight variables addressing different types of professional inconvenience to workers. According to, these job inconveniences have a negative impact on the well-being of employees and must therefore be associated with a salary increase. They also generate health and safety risks.

Third dimension: security and employment law

The last dimension relates to job security. Indeed, an important characteristic of vulnerable jobs is the low level of job protection. This aspect is even more important when we take into consideration the configuration of the labour market in Benin, which is predominantly informal. This represents a source of vulnerability for jobs. To give an example, more than 75% of jobs in Benin are not covered by any form of employment contract. In order to take into account this particularity of the labour markets of developing countries, we have retained two variables. These variables relate to paid holidays and the protection of pension rights. The table 3 presents a summary of vulnerability indicators (Table 3).

IVEB aggregation

The aggregation technique chosen in this study is consistent with that used. Indeed, once the dimensions are identified, it is a matter of aggregating them to make an index that will be considered as the measure of vulnerability. In this sense, two questions arise. The first is that of weighting. Should all sub-aspects of vulnerability be given equal importance? This debate does not have a universal response. This is moreover in line with the debate on the weighting of multidimensional indices in general. In this article, we have chosen to adopt an equal weighting. Indeed, an unequal weighting of the different components of the composite indicator can bias the results because individual preferences and by extension the responses depend on the individual context. The second question is which method to use. In our case, this is the summons. Each variable entering into the composition of the index is binary. Thus, the coding of binary variables is such that a low value of the index means low vulnerability.

Starting from the variables (D) selected, the index is as follows:

With: Y the natural logarithm of the wage; ? the constant term; X a set of variables; E the average of the years of schooling validated by stratum and ?i, the error term.

Correction of the endogeneity of the IVEB and of the schooling: estimator of IV2SLS (Instrumental Variables Two least Squares)

In this second estimate, we are only interested in the endogeneity of the two variables that are the IVEB and education. As in the previous case, we maintain the schooling instrument and we use the approach proposed to avoid an endogeneity bias [15]. This approach known as IV2SLS is a two-step process. First, we do a regression of the endogenous variable, which here is the IVEB, on all the instruments and the exogenous variables of the model. In the second step, the salary equation is estimated by regression replacing the IVEB with its predicted values obtained from the estimate in the first step. The equation we estimate looks like this: [16-30].

Yi = a + b1IVEBP + b2IVEBP² + ?Xi + cE + vi

With: Y the natural logarithm of the wage; has the constant term; IVEBP the predicted value of IVEB; X a set of variables; E the average of the years of schooling validated by stratum and vi, the error term.

Table 3: Summary of vulnerability indicators.

Simultaneous correction of endogeneity and selection biases

In this last estimate, we correct for both biases simultaneously. To do this, we use the IV2SLS estimator described above with the addition of the inverse of the Mills ratio to correct for selection bias. Ultimately, the IVEB is instrumented by its predicted values, schooling by the averages of years of schooling validated by stratum and finally the selection bay is corrected by the inverse of the Mills ratio. This inverse is obtained from the estimation of a Probit model of labour market participation. The estimated equation is as follows:

Yi = m + c1IVEBP + c2IVEBP2 + ?Xi + c3E + c4IRM + Wi

With: Y the natural logarithm of the wage; m the constant term; IVEBP the predicted value of IVEB; X a set of variables; E the average of the years of schooling validated by stratum; MRI the inverse of the Mills ratio and wi, the error term.

Robust results

We adopted a sequential estimation approach in order to compare the results and thus test the robustness of the effects.

Hypothesis of the existence of compensating differences

The objective is to test the hypothesis of compensating differences. Indeed, according to the theory of compensatory differences, vulnerable work therefore associated with poor working conditions will be rewarded by an overvaluation of its remuneration. The wage differential paid compared to its equilibrium level then appears as a compensation that the firm pays to the employee. To test the hypothesis of the existence of compensating differences, we adopted a method which is similar to that used previously.

The data

The main objective of the Integrated Modular Survey on Household Living Conditions (EMICoV) is to set up the bases for a permanent mechanism for monitoring and evaluating household living conditions in general and the poverty reduction program in particular. The specific objectives of this survey are, among others, the study of monetary poverty, the study of poverty in terms of living conditions, the study of subjective poverty, the determination of  the level of unemployment in  Benin and its determinants, measuring the extent of land conflicts in Benin, etc. Benin has 12 departments and each department is considered as an area of study where all the key indicators of the survey will be provided. A sample allocation specific to each department was applied. The allocation of each department was then distributed proportionally among the municipalities and according to the urban and rural environment. This corresponds to a stratification at the level of municipalities and by urban and rural environment. The EMICoV-2015 survey effectively reached 21,402 households in 920 enumeration areas. Table 4 presents description of variables (Table 4).

Table 4: Description of the variables.

The results show that the level of vulnerability affects wage formation on the labour market in Benin. Specifically, the intensity of vulnerability has a non-linear effect on wages. Despite the difference in sign of this nonlinear effect through the different estimates, this result reflects a contradiction with the theoretical predictions of the theory of compensatory differences. Indeed, the observed reversal of effect indicates that different levels of vulnerability may correspond to the same level of salary. In other words, it is possible that two agents, one with a low level of vulnerability and the other with a high level of vulnerability, have the same salaries. This would indicate that the level of wages is not fundamentally linked to the state of vulnerability of economic agents. Authors such as in a study on the effect of vulnerability on wages in the main cities of WAEMU also found a non-linear effect. The explanation for such a result certainly lies in the informal nature of most activities which increasingly marginalize vulnerable workers. With the quantity of employment comes great vulnerability. In this logic, vulnerable employment appears as a double penalty for the worker. In fact, the worker in a vulnerable job suffers initially, the disutilities of difficult working conditions and a deprivation of salary compensation in a second time (Table 5).

Is there a wage compensation mechanism for vulnerable workers?

To answer this question satisfactorily, it is necessary to relate wages and intensities of vulnerability. In accordance with the approach described in the methodology, Chart 1 shows the evolution of the level of wages according to the intensity of vulnerability. The wages used for Chart 1 are taken from the predicted values of the wages after estimating Equation 3 in Table 5. Analysis of the chart complements the analysis of the estimation results. Indeed, the graph shows a non-uniform variation in the trend between wages and vulnerability. This variation shows, considering the higher levels of vulnerability, a drop in the level of wages.

Table 5: Effect of vulnerability on wages.