The migration is a highly complex demographic phenomenon that responds to diverse and very difficult to determine causes. In Mexico as a result of the political, economic and social changes that have taken place during the last decades, they have created a new environment for the agricultural sector. These changes, coupled with a lack of public policies to support the countryside, have favored the widening of the poverty gap, provoking a profound social inequality and having among its many consequences the international emigration, mainly to the United States. The international emigrants from the agricultural sector are one of the less studied and understood subjects, there are no specific statistics or an in-depth analysis of the consequences of this phenomenon. In this context, the objective of this research is to present an analysis that allows to glimpse the general situation in which the international emigrants who were working in the agricultural sector in Mexico, just before their flight to the United States, and from a model of binary logistic regression, provide an approach to the factors that influence the probability that international migrants have been engaged in the agricultural sector.

The source of information for the study comes from the ENOE, which is produced by the National Institute of Statistics and Geography (INEGI), can be obtained from information on sociodemographic, occupational and labor data at nationally. The ENOE information is obtained through overlapping panels in which the same household is monitored for up to five quarters, and in each quarter about 20% of the households participating in the sample are broken. By integrating the information of two contiguous combs of the same household one can identify the absent ones in the home, and with the support of the questions about the country of birth and the destination of the absentee, a subgroup of people can be created who are the international emigrants.

Based on this methodology, it was possible to estimate gross international migration rates from the first quarter of 2005 to the first quarter of 2015. The Figure 1, shows the annual averages. Because the number of mexican-americans (those born in the United States of Mexican parents) is a considerable number, it was decided to omit them and take into account only those born in Mexico. According to ENOE numbers, both the number of migrants and the number of mexican immigrants has decreased in the last decade. Even before the economic crisis of 2008 there was a reduction in the volume of international mexican emigrants, however it is evident that it was from that year that the reduction was greater so that in 2005, for each ten thousand habitual residents in our country counted 109 exits for change of residence abroad, in 2009 the rate was reduced to more than half and by 2014 to less than one third (Figure 1).

Fuente: elaboración con base a la metodología de estimación de migrantes internacionales del INEGI.

Figure 1 International migrants and immigrants born in Mexico, 2005-2014 (rates).  

What is striking, and the subject of this research, is that during the period 2005-2014, the highest percentage of international emigrants who were employed in Mexico belonged to the agricultural sector (Figure 2).

Fuente: elaboración con base a la metodología de estimación de migrantes internacionales del INEGI.

Figure 2 Mexican emigrants employed by sector of economic activity, 2005-2014 (%).  

However, when the migrants arrive in the United States, the vast majority of them are employed in construction, or in the hotel and recreation sector, the agricultural sector only reaches 4.8% and 4.9% in 2013 and 2014 respectively (^{CPS, 2014}) (Figure 3).

Fuente: elaboración a partir de estimaciones de la current population survey (CPS), suplemento de marzo 2005-2014.

Figure 3 Mexican migrants employed in the United States, by economic activity, 2005-2014.  

Estimates from the ENOE indicate that the majority of international migrants who belonged to the agricultural sector over the period 2005-2014 are men (97.6% on average); in contrast, migrants immersed in another sector the percentage of men (84.1%), the majority of the agricultural migrants are younger than the non-agricultural ones, they have, in average, 29 and 30 years respectively. Likewise, the largest proportion of agricultural emigrants are concentrated mainly in small communities (less than 2 500 inhabitants), with an average of 76.7% of the population, while non-agricultural communities are concentrated in urban communities, 75.4% on average. In terms of marital status, on average 60% of agricultural emigrants are married or in free union, a higher percentage than non-agricultural workers (53%).

Among the characteristics of agricultural emigrants, there is an increase in the level of primary education (6 years of schooling) in 2005 to incomplete secondary education (7 years of schooling) in 2014, in the case of non-agricultural workers, the number of years studied from 9 to 10 years. The agricultural migrants had lower incomes than non-agricultural workers, on average more than 85% had incomes of no more than two minimum wages, of these, on average, 41.2% did not receive income, while non-agricultural migrants only 40.2% earned up to two minimum wages.

As for the labor profile of agricultural emigrants, the majority of them are unpaid subordinate workers (40% on average), followed by self-employed workers (29.2% on average), in contrast, the vast majority of non-agricultural migrants are unpaid subordinate workers (73.3%). It is also observed that the distribution of the employed population by type of employment (formal and informal) has changed relatively little between 2005 and 2014, above 95% of agricultural emigrants are employed in informal activities, while non-agricultural migrants have on average, a lower percentage of labor informality (72%).

In order to determine the factors that determine the probability that an international migrant has been immersed in the agricultural sector, a logistic model was used, because it allows to introduce a mixture of categorical and quantitative variables, it is possible to calculate the parameters of quantification of known risk in the literature as “odds ratio” from the regression coefficients (β) of the independent variables introduced in the model and the dependent variable (the one to be modeled, Y) is categorical, (binary in our case), which simplifies the representation of phenomena.

Logistic regression expresses the probability that the event in question will occur as a function of certain variables, which are presumed relevant or influential. If this fact we want to model or predict is represented by Y (the dependent variable), and the k (independent and control) explanatory variables are designated by X₁, X₂, X₃, ..... X_k the general equation (or logistic function) is:

Where: β_1, β₂ β_3…. β_k= = are the parameters of the model; exp= denotes the exponential function. This exponential function is a simplified expression corresponding to raising the number e to the power contained within the parentheses, where e is the Euler number or constant, or base of the neperian logarithms (whose approximate value to the thousandth is 2.718) (^{Jay, 2008}).

In the Table 1 gives a brief description of the variables that were used in the analysis. The effects of the explanatory variables on the probability of belonging to the agricultural sector are analyzed through a logistic model where the independent variable is dichotomic that takes the value equal to one if the migrant belonged to the agricultural sector and zero if it was in another sector. In this way, we try to estimate the probability that an international emigrant has been immersed in the agricultural sector. In order to obtain the data of the emigrants, in the first instance were merged, the bases of the first quarter of 2013 to the first quarter of 2015, after which a primary association or independence exercise was carried out between the explanatory variables and the dependent variable in order to explore the bi-variant associations through an analysis of contingency tables, this allowed to obtain, in a general way, an idea of the independent effects of the different variables on the probability that an emigrant had belonged to the agricultural sector before leaving.

Table 1 Description of the independent variables.  

*= la región tradicional comprende: Aguascalientes, Colima, Durango, Guanajuato, Jalisco, Michoacán, Nayarit, San Luis Potosí y Zacatecas; la norte: Baja California, Baja California Sur, Coahuila, Chihuahua, Nuevo León, Sinaloa, Sonora y Tamaulipas; la centro: Distrito Federal, Hidalgo, México, Morelos, Puebla, Querétaro y Tlaxcala; y la sur-sureste: Campeche, Chiapas, Guerrero, Oaxaca, Quintana Roo, Tabasco, Veracruz y Yucatán (CONAPO, 2012).

Subsequently, analysis was performed to detect possible confounding variables, if necessary, adjust or control, and thus prevent the primary relationship between the explanatory variables and dependent is spurious. In the same way an analysis is performed to detect the variables that could be effect modifiers or interaction in terms of increasing or decreasing the effect of the main relation.

Having performed the corresponding analyzes, the logistic regression was performed using the IBM SPSS 19 statistical program (^{Valderrey, 2010}). The method chosen is that of “forward” using the criterion of the likelihood ratio to contrast the new variables to be introduced or drawn from the model. Three measures “abstract model” are given to globally assess its validity: the first is the value of -2LL and the other two are coefficients of determination (R²), similar to that obtained in linear regression, expressing the ratio of the variation explained by the model. A perfect model would have a value of -2LL very small (ideally zero) and R² close to one (ideally one) (Table 2).

Table 2 Summary of the model. 

The data suggest that the model explained 51% of the variability of the studied phenomenon that is sufficient given the complexity of the phenomenon and the limitations of the source. Subsequently is shown the global adjustment test of the model known as the “Hosmer and Lemeshow test” (Table 3).

Table 3 Hosmer and Lemeshow test for migrant model.  

What is wanted in this test is that there is no significance (the opposite of what is usually usual) because the null hypothesis is that the proposed model conforms to what is observed. Therefore, a p-value greater than 0.05 implies that what we observe fits sufficiently into reality.

Finally the variables in the equation are presented, their regression coefficients with their corresponding standard errors, the value of the Wald statistic to evaluate the null (β_t=0), the statistical significance associated, and the value of the OR (exp (B)) with their confidence intervals (Table 4).

Table 4 Variables in the equation.  

With this data we can construct the logistic regression equation:

With this equation we can predict the probability of having the result (Y= 1) of "agricultural sector". Thus, an individual who is male (sex= 1), the locality in which he or she resides is less than 2 500 inhabitants (locality= 1), the number of years of study are less than nine (years of schooling= 1), their income is not more than two minimum wages (income= 1) and serves on labor informality (formality= 1) would, according to the best estimate model, a likelihood that the migrant has been busy before leaving:

This predicted probability (as is greater than 0.5) is classified as the migrant associated with the agricultural sector.

The significant variables in the best estimated model reflect the precariousness and poor quality of life sustained by people immersed in the agricultural sector before emigrating.