2.1. The aggregation procedure

The first stage of the analysis combines different elements of deprivation at the individual level to identify those that are long-term-employment deprived. In this stage, we use a dimensional cut-off k which identifies people that face deprivation in many dimensions, while τ refers to the period of time by which a person is considered long-term employment deprived. The second stage refers to the aggregation of individuals’ information into an index that characterizes the level of long-term employment deprivation in society. There are, however, many ways to complete this aggregation procedure according to certain desirable properties. The simplest measure available is the long-term employment deprivation headcount H ^LT , which is the fraction of the population that is employment deprived during a period of time τ . The index can be expressed as:

that is, the number of long-term employment deprived q ^LT , over the total population (n). However, the measure is insensitive to the number of deprivations (^{Alkire and Foster, 2011}) and the number of periods the long-term unemployment deprived experience (^{Alkire et al., 2014}). That is, given the dimensional cut-offs k and τ, if an individual is identified as long-term employment deprived and becomes deprived in the additional dimension of period of time, the long-term employment deprivation headcount does not change.

In order to overcome the limitations of the long-term employment deprivation headcount H ^LT , we follow ^{Alkire and Foster (2011)} and ^{Foster (2009)} to obtain a normalized population average of long-term employment deprived. In other words, we add up the means of all dimensions m and periods of time t, to obtain an aggregate index. The proposed dimension and time-adjusted multidimensional deprivation index ED ^LT , is defined as:

where d _ij (k) is the deprivation of individual i in dimension j,W _j is the weight assigned to dimension j, such that ∑j=1mwj=m.The index proposed in (6) can be further disaggregated into partial indices that convey meaningful information on different features of long-term multidimensional deprivation. Therefore, the index can be re-expressed as:

where A ^LT is the average intensity of poverty among the long-term deprived, and D ^LT the average duration of deprivation among the long-term employment deprived. The ED ^LT index in formula (7) is the sum of normalized gaps of employment deprivations that predispose individuals’ employment prospects in the labor market, divided by the possible highest value of this sum nm. The total measure can be decomposed into three components: a) the headcount ratio of the population that is employment deprived during a period of time τ ; b) the intensity of deprivation among the long-term employment deprived, in other words, the share of weighted deprivation that long-term deprived people experience in the period in which they are multidimensional employment deprived; c) the average duration among the long-term employment deprived which measures how people experience deficiencies in the labor market. Persistent conditions of insufficiency might severely predispose opportunities in the labor market. The duration of employment deprivation may increase the likelihood of low payment and informal work and hinder opportunities in the labor market in which experience may be relevant. The decomposition allows us to understand the incidence, intensity and duration of those who are long-term employment deprived to better assess the situation in the labor market.

3.1. Depiction of the Mexican labor market

Before the 2008 financial crisis, the Mexican labor market was characterized by low unemployment rates, together with high informality and low wages in an economic context of slow growth. After the crisis, though, the levels of unemployment and informality increased (^{Rodríguez, López and Prudencio, 2013}). The government’s response to tackle this shock has been weak (^{Freije, López and Rodríguez, 2011}) and, as in the rest of Latin America, there is an increasing working impoverished population in Mexico (^{Beccaria et al., 2011}; ^{Rodríguez, López and Prudencio, 2013}); meaning that for many people having a job is no longer enough to meet their basic needs. Moreover, the level of long-term unemployed in Mexico, although small, has remained the same since the hike of the crisis. In Figure 1 we compare the United States and Mexican unemployment patterns. We present both the short- and long-term levels of unemployment. In the case of the United States both the short- and long-term unemployment levels increased sharply in 2008, but they have decreased almost to pre-crisis levels. That is not the case for Mexico. Both the short- and long-term unemployment levels have stayed relatively the same since the financial crisis.

Sorce: Own calculations using INEGI and Boreau of Labor Stadistics

Figure 1 Unemployment trends in the United State and Mexico 

The status of employment -and whether this employment is for-mal or not- varies considerably by gender, level of education, region or industry. We study this further by observing the employment pat-terns presented by the household heads. For Mexico, the status of the household heads in the labor market and their level of education are important determinants of the household’s well-being. At the same time, this group presents a lower attrition rate than other household sub-groups. We thus focused on this group for both the description of the labor market and the construction of the proposed index.

Constructing five period panels for 2007 and 2014, we present in Table 1 the employment distribution of the household heads by different categories. The results presented correspond to panel averages. We observe that most of the male household heads are fully employed, 45.33 percent of them working in the informal market and 39.04 percent in the formal market, while in 2014 these percentages were 42.08 and 38.92, respectively. When considering the underemployed, in 2007, 7.19 percent of men fall into this category, of which 5.97 percent points correspond to those working in the informal market. For 2014, 8.36 percent of men were underemployed, of which 6.94 percentage points were in the informal market. All of these changes between 2007 and 2014 are small, in the range of 1 to 2 percentage points. In the case of women, the biggest proportion of workers is employed in the informal sector, followed by those employed in the formal sector. Once more, most of the underemployed belong to the informal sector. Finally, in contrast to men, in 2007 34.6 percent of female household heads were out of the labor market, and 33.83 percent in 2014.

These results vary significantly when dividing the groups by education level. Individuals with six years or less of education are for the most part fully employed in the informal sector: 54.68 percent in 2007 and 52.98 percent in 2014. Only a small proportion is fully employed in the formal sector, 17.89 and 15.92 percent for the periods analyzed. When considering the household heads with more than 12 years of education, most of them work in the formal sector and are fully employed. Interestingly, the unemployment levels do not vary significantly with the educational level. Nevertheless, the same does not apply with people out of the labor force, which decreases according to the educational achievement of the household head.

When dividing the groups by rural and urban regions we observe that the results from the rural sector are very close to individuals with six years or less of education. Similarly, the patterns of the urban population are parallel to those with higher education.

Finally, when dividing workers by industry we observe a clear lag in the workers employed in the primary sector given that almost all of them work in the informal sector. Workers in the construction industry also present relatively high rates of informality and, along with the primary sectors, these industries present the highest rates of underemployment. Both in regard to informality and underemployment, these groups are followed by workers in commerce, restaurants, and transport. In general, changes between 2007 and 2014 are small when dividing the work force by gender or region (rural/urban), although they are relatively higher when creating subgroups by the level of education or by industry.

The income trends of these groups, together with the differentiation by medical insurance, reveal a continuing decrease of income in real terms in all sub-groups. Results shown are from panel averages that were constructed for each quarter. For example, data for 2014 first quarter would correspond to a five-quarter panel of individuals whose first observation was in the first quarter of 2014, the panel of the second quarter of 2014 includes those individuals whose first observation was in the second quarter of 2014, and so on. On the horizontal axis we label only the years for clarity. On the vertical axis we report the real income of the household head at 2010 prices. Results are reported in thousands of Mexican pesos.

Finally, Table 2 shows workers’ mobility in the labor market, and how their mobility has changed between 2007 and 2014. In general, mobility decreased during this period, that is, in 2014 a higher proportion of individuals remained in the same status in the labor market. Although fortunate for the employed, the unemployed are striving to find a job. In 2007, before the financial crisis, 9.96 percent of the unemployed remained in that position after five quarters, whereas 74.10 percent of them found a job, 26.69 percent in the formal market and 47.41 percent in the informal market. In contrast, at the beginning of the 2008 crisis the proportion of individuals who were still unemployed amounted to 9.56 percent, but the percentage of those who had found a job decreased to 68.92, the formal market having registered the greatest decrease. Currently, in the 2014 post-crisis period, an even smaller proportion of the unemployed found a job, namely 66.03 percent, mainly due to a hiring decline in the informal market -44.66 percent versus 49.80 percent in 2008-, and secondly due to an increase in people who remained unemployed -12.05 percent vs. 9.56 percent. Moreover, most workers who were already out of the labor force remained in that category and nearly all of those who did manage to enter the labor market found a job in the informal sector.

Figure 2 Income patterns 

In general, unemployment among household heads is low, but as ^{Beccaria et al. (2011)} have pointed out, employment may not be sufficient for a household to escape from poverty. The real income of the household heads is decreasing in all the subgroups considered, as well as their mobility in the labor market. In this context, it is important to study the dimensions that may contribute to the resilience of individuals to excel in the labor market and to increase their mobility and well-being. In the following section we describe the dimensions considered in the proposed long-term employment deprivations index.

4.1. Set of labor dimensions

The long-term employment deprivation comprises a set of attributes that characterize different aspects related to individual qualities that offer good opportunities of well-paid jobs in the labor market. We now discuss the choice of attributes, given both their impact for the analysis and the fact that they are subject to specific value judgments. Therefore, it should be discussed as explicitly and as firmly as possible in employment public policy literature. The traditional approach to labor market indicators tends to have a macroeconomic drift and sometimes the indicators are not as relevant in the developing world as they are in developed economies, and hence they do not provide an accurate picture of labor markets in these economies (^{Lugo, 2007}). On the other hand, the ^{World Bank (2013)} report on jobs stresses the idea of a “decent job” in which rights and laws serve as a basis to avoid non-decent jobs. This refinement in definition implies that full employment is not the final goal, since a good job is rather that which has the highest value for society.

Recent studies on decent jobs focus on the legal aspects of the activity, which often misrepresent labor market vulnerabilities, even if in most cases they ensure protection (^{Lugo, 2007}). The set of attributes we are interested in are related to the prospect of finding a good job that guarantees an individual’s well-being. A combination of the ^{Lugo (2007)} and ^{Rodriguez, Cardozo and Parra (2014)}, against the backdrop of ^{Alkire et al. (2014)} multidimensional poverty selection process was used for outlining the attributes and domain in the analysis. Therefore, our definition is broader that the idea of a “decent” job, as it considers indicators that reflect labor market vulnerabilities and the values and norms of the specific society.

The views of experts inspired our first set of domains and indicators, to which we added household and individual features considered in the literature on decent jobs and labor market indices and the assessment of existing data and data availability. Hence, our list of indicators includes four dimensions and eight indicators, as shown in Table 3.

The percentages presented are for the sample of household heads for the 24 to 65 year-old cohort. Educational achievement is defined as the number of members in a household having completed the schooling years as prescribed by law (primary and secondary level). As we can see, the education deprivation among household heads was 8.75 percent in 2007 and 6.14 percent in 2014. Although deprivation in education decreased, the number of households being working poor increased 12 percentage points in this sample. We define an individual as a working poor if the household’s per capita income that comes from labor is not enough to purchase a basic food basket. Finally, we note that the highest percentages are observed in the category of security and legal protection, which displays the lack of law enforcement concerning the legal requirements to which Mexican workers are entitled.

Table 4 presents various labor market indicators and how they relate to the chosen dimensions. We use five labor market indicators, where U stands for unemployment, UP for unemployment and part-time workers, GP for general pressure in the labor market, defined as the unemployed and the employed who are seeking a job, UND stands for underemployment, and, finally, CCO for workers in critical conditions.¹

In the first three rows we report the percentage of individuals that fell under the labor market indicator reference at least t times. For example, for unemployment (U) in 2007, 5.65% of the individuals were unemployed in at least one quarter. This percentage drops when considering individuals who were unemployed durng two quarters, and it is minimal when t = 5. We observe that the 2014 percentages are higher in every category and also that, although the rates of unemployment are low, 29% of workers labored at least one period in conditions considered critical. In line with ^{Joutard and Sagaon (2006)}, we observe that unemployment in Mexico shows very short durations, which is also the case for the rest of the labor indicators. In this sense, the heads of household seem to be mobile, but many of them fell at one point in time into conditions of labor vulnerability.

For the second part of the table, we take as our sample individuals who fell at least once under the labor market indicator referenced, and then we observed how many of them were at the same time deprived in each of the dimensions specified. For example, all those who were unemployed at least once, 8.80 percent were deprived in the dimension of educational achievement in 2007, and 5.51 percent in 2014. We further note that for the dimension of educational achievement the percentages observed in 2014 are lower than in 2007 despite the fact that the labor market conditions seem to have worsened in 2014. In the rest of the variables, the figure is diverse.

Table 3 shows that 21.58 and 33.47 percent were labor poor in 2007 and 2014 respectively. An individual is considered working poor if the per capita labor income of the household is not sufficient to buy a basic food basket. In Table 4 we observe that of all the house-holds where the head was unemployed, 70 percent were labor poor in 2007, and 80.39 percent in 2014. Those percentages increased when considering other categories, with the only exception being the underemployed. These results are consistent with ^{Inchauste (2012)} who argues that status in the labor market is the most important determinant of poverty within the household. Finally, the highest incidences of deprivation correspond to the dimensions of informality and the lack of contract. Note that the percentages presented are panel averages.

4.2. Robustness checks

To check the robustness of the multidimensional index we measured the head count (H ^LT ) for different cut-offs of k. As expected, there is a reduction in H ^LT for every increment in the cut-off k. This holds for both the long-term and short-term measures. Here we define the short-term deprived by setting the time dimension cut-off as 1 ≤ τ ≤ 2. The results are presented in the Appendix 1.

When taking k = 3, the levels of the short- and long-term deprived are relatively the same, and a gap of around 20 percent appears when taking k = 2. Also note that long-term deprivation is relatively higher in 2014 than in 2007, and it is also the case for short term deprivation with the exception at k = 1.

Now, when varying both cut-offs k and τ we see a high variability of ED ^LT ; the results are shown in Table 5. First, when considering the union approach under k = 1 and τ = 1, the incidence of the deprived individuals is as high as 92.13 percent and the adjusted headcount 19.86 percent for 2007. In 2014, we have an increase in incidence to 96.37 and 21.94 percent for the adjusted headcount. At the other extreme, when taking the intersection approach for the time dimension (τ = 5) along with k = 3, we observe that the headcounts are 11.56 and 12.57 percent for 2007 and 2014, and when taking into account the censored headcount we have 5.66 and 6.19 percent.

When observing both the intensity A ^LT and the duration D ^LT indexes, we note that the duration is relatively high throughout the table, whereas A ^LT varies more. From now on we will use the cut-off of k = 3 and τ = 3.

Now, considering the second part of Table 5, using the chosen cut-offs we observe that the index does not follow the unemployment rate closely. Nevertheless, it does approximate considerably to the rest of the labor market indicators, which are more related to either the level of satisfaction of the individual in the labor market, or the definition of the precariousness of work in general. In other words, ED ^LT as a measure seems to resemble the capabilities of the individual to work at non-precarious jobs. In Figure 3, we estimate the headcount H ^LT for each quarter from 2007 to 2014 using the cut-offs of k = 3 and τ = 3, and we compare it to the levels of unemployment and underemployment, and workers in critical conditions. As can be attested, there are also similar patterns in the level of H ^LT and the level of workers in critical conditions in the period under examination.

Figure 3 Comparison between labor indexes 

Now, to further analyze the behavior of each dimension and its relative importance in ED ^LT , Table 6 exhibits the censored head-counts of each dimension and their respective relative influence for various cut-offs of τ .

In addition, we also present the change observed from 2007 to 2014. In general, the dimensions of working poor, informality, lack of contract, lack of benefits, and experience have worsened, whereas the dimensions of educational achievement, illiteracy, and health disability have improved. Nevertheless, as observed in the changes presented, the differences between the two panels are small, with the exception of the working poor and the informality levels when considering τ = 1. When observing the relative importance of each dimension, the most important variables in both periods were lack of contract and informality, followed by illiteracy and working poor. Interestingly, the relative importance of the variables related to productivity, which are education achievement and illiteracy, decreased the most, and the variable that fared the worst was working poor.

Finally, making use of the sub-group and decomposition properties of the ^{Alkire et al. (2014)} measure, we estimated the results for each state. Using the cut-offs of k = 3 and τ = 3, we generated deprivation maps by state for the variables H ^LT A ^LT and E D ^LT in Figure 4.

Figure 4 Map of ED^LT (2007, 2014) 

For the measure of ED ^LT , the 2007 map generates a clear division between the northern, central and southern regions. The lighter colors represent higher levels of deprivation. Therefore, we observe that workers in the north are less deprived than those in the central and southern regions. The state that performed the worst was Chiapas. In 2014, these divisions were less strong, mainly due to a deterioration in the labor characteristics of the workers in the states of Chihuahua and Coahuila.

The maps for the headcount H ^LT A ^LT and D ^LT can be found in Appendix 2. For the former we observe, on the one hand, that in 2007 the more prosperous northern region, along with Mexico City, presented the lowest levels in the head count (H ^LT ). On the other hand, the southern border had the worst performance. In 2014 the figure changes slightly, and some states in the north -as we pointed out before- presented a deterioration in their labor market conditions, reflected in higher levels of H ^LT .

The picture for the intensity in deprivation is varied. In 2007 the state of Tabasco performed the worst in this category, and for 2014, the state of Puebla. Finally, in regard to duration, the south presents the longest spells of deprivation both in 2007 and 2014.