INTRODUCTION

Agriculture is a relevant sector in the economy of many countries as it serves as a cornerstone for ensuring their food supply and security, while also contributing to job creation and various other benefits. The Mexican agricultural sector is characterized by its heterogeneity due to the coexistence of both subsistence producers and agro-industrial companies, as indicated by ^{Yúnez Naude (2010)}. In this regard, based on the 2007 Agricultural Census (^{Inegi, 2012}), ^{Stabridis (2022)} shows that, despite subsistence production units (with up to two hectares in size) accounting for half of the total, they only possess 5.5% of the total agricultural land, whereas the large units (over 20 hectares) possess 65% of productive lands. These larger units predominantly utilize irrigation water systems, employ the majority of the labor force, and cultivate crops through one or more cycles annually. Over time, they have developed agricultural practices that facilitate export to international markets through producer-exporter associations or under production contracts with transnational companies.

Based on the aforementioned points, we can categorize agriculture in Mexico into two primary types: the first revolves around the production of staple crops such as corn and beans, among others, which generally hold lower commercial value; and the second focuses on horticultural products (e.g., tomatoes, peppers, etc.) and fruit-bearing trees (berries,³ cherries, nuts), which have a high commercial value. The labor demand patterns associated with these types of production ensure year-round, consistent export stream to regions like the United States, Europe, or Asia, resulting in substantial profits for these production units. This creates a heterogeneous demand for agricultural labor that depends on the type of production unit, whether commercial or subsistence. Regarding the latter, they typically rely solely on family labor; small commercial units tend to hire workers for short seasons; whereas large productive units that supply the domestic and export markets sustain a continuous demand for wage labor, which is fulfilled by agricultural laborers.

Farmworkers are (mostly) temporary workers who endure precarious⁴ working conditions. Many of them are forced to migrate from their places of origin—usually from the poorest states—to the major agricultural regions located in the western and northwestern parts of Mexico in pursuit of improved wages. Precariousness characterizes the occupational landscape of agricultural laborers (^{Lara Flores, 2011}). Notably, a significant portion of these laborers are indigenous individuals, of which between 10 and 15% are women, as evidenced in the data from the 2020 Population and Housing Census (Censo de Población y Vivienda 2020) and the National Survey of Household Income and Expenditure (Encuesta Nacional de Ingreso y Gasto de los Hogares [ENIGH]) (^{Inegi, 2020}, ²⁰²¹). The intersection of these variables reveals a wage gap in comparison to their counterparts, specifically non-indigenous individuals and men, respectively. While disparities in skill levels contribute to a portion of this wage gap, another significant part can be attributed to occupational segregation and discrimination.

The objective of this research is to analyze the impact of gender and ethnicity on the wage gap among farmworkers in the northwest region of Mexico, using the census sample data from the 2020 Population and Housing Census. Through the application of matching methods, we aim to uncover the direct influence of gender and ethnicity on the wage gap among agricultural laborers and determine whether this association can be attributed to discrimination. The advantage of employing this methodological approach is that it not only allows us to isolate the effects of gender and ethnicity on wages from other characteristics, but also facilitates the examination of the impact of additional variables such as age, education, and migration status. The study focuses on the northwest region of Mexico—encompassing Baja California, Baja California Sur, Sinaloa, and Sonora—due to its significance as a major agricultural area, characterized by high agricultural production value, substantial export levels, and a continuous demand for farmworkers throughout the year.

The results of the analysis reveal that women’s wages tend to be 7% lower than those earned by men, and this difference increases between the middle and upper segments of the distribution. A similar trend is observed when considering the ethnicity variable, as indigenous individuals earn 17% less than non-indigenous individuals, and this gap becomes more pronounced in the upper segments of the distribution. In other words, the wage gap widens as it moves towards higher percentiles. When the gender and ethnicity variables are intersected, it becomes apparent that the lowest wages are earned by indigenous female farmworkers. It is considered that a portion of this disparity can be attributed to discrimination.

The article is divided into six sections: the first addresses the conceptual framework regarding farmworkers, the second considers the literature review on wage gaps, the third outlines the methodology used, the fourth describes the data source, the fifth presents the results, and in the final section, the conclusions are presented.

AGRICULTURAL DAY LABOR IN THE NORTHWEST REGION

THE WAGE GAP

METHODOLOGY

As previously mentioned, the BOD serves as the initial approach to studying the wage gap. This model differentiates between the aspect defined by group composition differences and the component defined by wage structure, which is associated with discrimination. However, despite the method’s practicality, it is considered to have several drawbacks. For instance, it only provides accurate estimations when parameter specifications are linear, as noted by ^{Barsky et al. (2002)}. Similarly, ^{Weichselbaumer and Winter-Ebmer (2005)} argue that the unexplained component doesn’t solely encompass discrimination but may also involve other omitted variables—such as productivity differences or occupational segregation—. In this regard, ^{Ñopo (2008)} points out that the BOD takes into account differences in worker distributions without considering that not all are necessarily comparable, potentially introducing biases in the estimations. To address this concern, the author employs exact matching techniques to analyze the wage gap among individuals with similar observable characteristics. ^{Frölich (2007)} also utilizes PSM to estimate the Average Treatment Effect on the Treated (ATT) related to gender differences in the wage gap.

For this article, matching techniques are applied to separately estimate the effect of gender and ethnicity on the wage gap among day laborers. For the simultaneous modeling of gender and ethnicity in the wage gap, IPWRA models are used. Each of these methods is briefly explained below.¹³

INFORMATION SOURCES

The utilized database is the census sample derived from the extended questionnaire of the 2020 Population and Housing Census, which is representative of all municipalities and localities with more than 50 000 inhabitants—where representative samples were collected—. Similarly, data was collected through a census that included all households in municipalities characterized by a low Human Development Index, those with a high concentration of indigenous and/or Afro-descendant population, and those with fewer than 1 300 occupied households (^{Inegi, 2021}). The census sample consists of approximately one-tenth of Mexico’s households. From this sample, only farmworkers residing in the northwestern region were selected, considering the population aged 12 to 75 years in order to comprehensively capture the agricultural workforce. Therefore, the study subsample is constituted by the following variables:

The averages of these variables are shown in Table 4.¹⁶ The average wage is 33.3 MXN per hour, being higher for men than for women (33.7 and 31.7 MXN, respectively). Similarly, non-indigenous day laborers earn more compared to indigenous ones (35.1 and 29.6 MXN, respectively). Notably, the gender wage gap is more pronounced among non-indigenous farmworkers (35.6 and 33.6 MXN for men and women, respectively) than among indigenous laborers (29.6 and 29.4 MXN for men and women, respectively), where it is nearly negligible. These findings offer valuable insights into the wage disparities rooted in both gender and ethnicity within the northwest region.

Regarding the percentage of female farmworkers (23%), it is observed that this figure is higher among indigenous individuals (28%) compared to non-indigenous individuals (21%). One-third of the farmworkers (33%) self-identify as indigenous, with this proportion being higher among men (40%) than among women (31%). In terms of age, the sample’s average is 37.4 years, with women (35.1 years) being younger than men (38 years). Similarly, indigenous individuals are younger (35.9 years) compared to non-indigenous individuals (38.1 years).

The average years of schooling are at the middle school level (7 years), demonstrating a very similar pattern between men and women. Additionally, non-indigenous individuals have 0.7 years more education compared to indigenous ones (7.2 and 6.5 years, respectively). This difference is smaller than the one observed when comparing workers without distinguishing their occupation. Indigenous language speakers constitute 14% of the total, with a notably higher percentage among workers who identify as indigenous (40%) than among non-indigenous individuals (1%). Approximately 30% of farmworkers in the northwest are migrants, which represents the highest percentage among all regions.

In the case of migrant farmworkers, the percentage of women (40%) exceeds that of men (27%). Similarly, the percentage of indigenous migrants (44%) is higher compared to non-indigenous migrants (23%). In all cases (total, men, women, indigenous, and non-indigenous), the farmworkers have a spouse. Regarding extreme labor poverty, it is observed that 15% of day laborers experience extreme poverty, with the rate being higher among men (both indigenous and non-indigenous) than among women. Approximately 40% of farmworkers reside in rural areas, defined as areas with fewer than 2 500 inhabitants. Day laborers in the northwest who live in a municipality of the Zona Libre de la Frontera Norte (ZLFN) constitute 13% of the total, with non-indigenous women workers having a slightly higher presence in ZLFN municipalities (15%).

RESULTS

The results of gender-based matching are shown in Table 5, where both the results obtained by PSM (Propensity Score Matching) and Mahalanobis matching¹⁷ are presented. All the obtained results are significant at the 99% level, with the effect of being a woman on the average wage gap hovering around 4%. The results are expressed as a percentage of the gap and are in reference to the matched portion (matching).

According to ^{Meara et al. (2020)}, the treatment effect of being a woman on the wage gap is the difference between the matched groups. In other words, when comparing men and women with similar characteristics (age, education, etc.), it is observed that women receive 4% less in wages, which could be associated with discrimination and other unobservable characteristics such as productivity. In the occupation of agricultural laborers, it is difficult to see occupational segregation that does not favor women, as there is rather a segregation based on the type of production unit towards those that operate with more informal schemes, such as piece-rate payment.

Table 6 presents the results from ethnicity-based matching. The effect of ethnicity is significantly unfavorable for indigenous farmworkers, who receive a lower salary by 16% (ranging from 15.7% to 17.7% with Mahalanobis matching and 15.7% to 15.9% with PSM) compared to non-indigenous workers. This 16% divergence could be attributed to discrimination and, in some instances, is evident across various regions of the country where indigenous laborers are often assigned physically demanding tasks with limited opportunities for additional compensation. They might also prioritize working in informal production units that offer piece-rate payment without access to benefits. For the case of the United Kingdom, ^{Meara et al. (2020)} show that considering occupational segregation can yield slightly different results.

For the case of multi-treatment estimations, the results are presented in Table 7¹⁸ based on ^{Cattaneo (2010)} and ^{Cattaneo et al. (2013)}. When considering indigenous men as the reference category, the effects of the remaining categories on the wages of farmworkers are obtained.

The values of the logarithm of wages and the treatment effect are reported. In terms of the mean, it is observed that non-indigenous women earn 7.5% less than non-indigenous men, controlling for observable characteristics, so that 7.5% is the effect of being a non-indigenous woman. Indigenous men receive a salary 16.5% lower than that of non-indigenous men. The most disadvantaged category is that of indigenous women, as their salary is 24.1% lower than that of non-indigenous men, reflecting the vulnerability experienced by indigenous female laborers through their earnings.

Now it’s worth questioning whether these gender and ethnicity effects persist throughout the wage distribution. This can be understood by examining the results by quantiles: at the 10th percentile, it is observed that the gender effect diminishes, with ethnicity playing a more significant role (among non-indigenous individuals, both men and women, the gender difference is only 1%). However the differences between non-indigenous men and indigenous men and women amount 22.2%.

As we move towards higher percentiles, the gender difference becomes more significant, and the ethnicity difference increases, becoming very high at the 90th percentile, where non-indigenous women earn 15% less than their male counterparts. The most extreme difference is observed among indigenous female laborers, whose salaries are 36.5% lower than those of non-indigenous men. This increase in the wage gaps at the higher percentiles indicates the presence of a glass ceiling,¹⁹ as measured by ^{Arulampalam et al. (2007)}. Table 8 provides a comprehensive comparison of all categories.

When comparing indigenous men to non-indigenous women, it is observed that the former have a 9% lower salary than the latter on average. This difference is higher in the lower part of the distribution (19.8% at the 25th percentile). However, in the middle and upper parts of the distribution, the differences in wages between these groups are not significant. In the case of comparing women, the effect of ethnicity can be observed: non-indigenous women have a 16.7% higher salary than indigenous women on average, and when exploring this difference across the wage distribution, it widens at the distribution extremes and becomes insignificant at the median. At the 25th percentile, non-indigenous women have a 21.1% higher salary than their indigenous counterparts; similarly, this difference remains at 21.1% at the 90th percentile. This could be interpreted as the presence of a sticky floor²⁰ and glass ceiling.

The comparison between indigenous men and women allows us to observe the gender effect among indigenous individuals. On average, indigenous women tend to earn a salary 7.6% lower than their male counterparts, but when examining the difference across the wage distribution, it is observed that it only exists at the high end of the distribution, namely, a 16.8% difference at the 90th percentile.

The significance of these results lies in the interaction between gender and ethnicity effects on wages, which can vary depending on the part of the distribution being examined. Thus, it is demonstrated that the gender-ethnicity interaction (indigenous women) indicates a substantial difference ranging from 21 to 36%. This difference is only diminished at the median of the distribution. In this manner, both effects contribute to the widening of disparities.

To corroborate the robustness of the results obtained with the IPWRA model, we conducted a comparison with the estimates of gender and ethnicity gaps using the RIF decomposition methods (^{Firpo et al., 2018}), using the commands developed by ^{Rios-Avila (2020)}. The results presented in Tables 9 and 10 show that the gender and ethnicity gaps favor male and indigenous workers. It is observed that most of the gap remains unexplained by the characteristics of the groups, which does not necessarily imply that this is due to discrimination but, as noted by ^{Meara et al. (2020)}, reflects the high degree of heterogeneity among the groups, something that is evident from Table 4, where differences in age, education, and migration status are evident. Therefore, it is considered that the results obtained with the matching and IPWRA models are a more suitable approach for modeling the studied effects compared to the RIF models.

CONCLUSIONS

The objective of this article was to analyze the effect of gender and ethnicity on the wage gap among farmworkers in the northwest of Mexico, both on the mean and across the wage distribution, to determine if there is a component of discrimination. These effects were found using matching methods, considering gender or ethnicity as the treatment, and then, with the use of IPWRA models, to simultaneously evaluate both effects.

The results obtained through matching reveal a direct negative effect of both gender and ethnicity on wages, indicating that in an environment where farmworkers with similar observable characteristics are compared, women tend to have lower wages than men, and indigenous individuals have lower wages than non-indigenous individuals. The effect of ethnicity is greater than that of gender (16.5% vs. 4%). One advantage of matching estimates over the BOD approach is its ability to control for heterogeneity, thereby significantly reducing selection bias in observable characteristics.

To assess the simultaneous effect of gender and ethnicity across the wage distribution, IPWRA models were employed. The results in the mean indicate that non-indigenous men earn the highest wages, whereas indigenous workers, both men and women, have the lowest wages.

When these gaps were analyzed across the wage distribution, it was observed that in the lower part of the distribution, only the effect of ethnicity was significant, with the gender effect not being significant. However, as we moved towards higher percentiles, the effect of ethnicity decreased slightly while the gender effect increased. This is why non-indigenous women earn almost the same wages as their male counterparts at the 10th percentile, but much higher than indigenous individuals. However, at the 90th percentile, non-indigenous women earned wages 15% lower than non-indigenous male laborers, and the gap between indigenous workers and non-indigenous men increased, signifying the presence of a gender-related glass ceiling. In the case of indigenous men compared to non-indigenous men, the effect of ethnicity decreased as we moved along the distribution. When comparing the effect of ethnicity in women, it was observed that the gap was larger at the extremes, with no discernible wage differences at the median. This dual pattern suggests the simultaneous existence of both a sticky floor and a glass ceiling.

Lastly, when comparing the gender effect among indigenous laborers, it was observed that wage differences between indigenous men and women were not significant from the 10th percentile to the median of the distribution. However, in the upper part of the distribution, lower wages for women became apparent, indicating the presence of a glass ceiling. As mentioned earlier, there is limited research on the wage gap among farmworkers. For instance, in the United States, ^{Fisher et al. (2021)} found that women earned 6% less than men, partly due to discrimination. In Mexico, ^{Stabridis and Salgado Viveros (2022)} discovered that female laborers earned 15% less than their male counterparts, with a portion of this difference attributed to discrimination.

The negative effects observed in wages concerning gender and ethnicity cannot be solely attributed to factors like discrimination. There are also unobservable characteristics at play, such as the type of payment (e.g., integrated salary, mixed, piece-rate), potential occupational selection (certain tasks may be assigned to women or indigenous individuals, or they may choose them for higher pay), or gender-related challenges that women might face in accessing extra payments for longer working hours, potentially resulting in lower wages. However, it is possible to attribute these negative effects, at least to some extent, to gender and ethnicity discrimination.

It is worth highlighting that in future research, it would be valuable to include gender-specific labor supply at the local level as a variable that may affect wages, as well as incorporating information about producers that could have an effect on wages or influence labor supply. The current challenge stems from the lack of a comprehensive survey or census that captures the specific characteristics of agricultural laborers or the productive units they work for, similar to what is collected in the United States with the National Agricultural Workers Survey (NAWS) or similar agricultural surveys in Chile, Peru, and Colombia. The National Agricultural Survey (Encuesta Nacional Agropecuaria) conducted by Inegi offers limited information, and its accessibility is restricted.