The impact of climate variability on agricultural employment in Mexico from 1980–2017

activity outside agriculture, 2) Output: individuals who left agricultural activity, 3) Input: those who started working in agriculture, 4) Remained in the same agricultural activity, and 5) Those who changed activities within agriculture (e.g. from agriculture to raising cattle; or from maize to alfalfa; or from food crops to cash crops / plantations). Regarding the classification of occupation categories, it is important to note that two classi- fiers developed by INEGI were utilized. Specifically, from 2005 to the second quarter of 2012, the Mexican Classification of Occupations (CMO) [60] classifications was employed. How- ever, starting from the third quarter of 2012, there was a transition to the National Occupa- tional Classification System (SINCO) [59], which provided a higher level of description for various activities. This shift allowed for more detailed distinctions, such as the inclusion of "Support Workers in Agricultural Activities" as a recognized category. For the ENE (1998 to 2004) [54], the data are presented both annual and quarterly, so we chose to use the interactive tabulations provided by INEGI for the National Employment Sur- vey. We selected the employed population tabulation and then applied the filters “Main Occu- pation” and “Total Workers” to identify the number of respondents employed in an agricultural activity. 2.3.2 Statistical analyses. 2.3.2.1 Agrarian abandonment. Drawing from the shortened dataset of EDER 2017, which exclusively includes individuals aged 12 and above, along with

the reclassification of unemployed individuals, general workers, and those engaged in the agri- cultural sector, we conducted an analysis to ascertain the average age of departure from agri- cultural occupations. Our aim was to rule out the possibility that abandonment is related to age. To accomplish this, we documented instances of individuals transitioning out of farming over their lifetimes and visualized this data using a time series graph. Using the same abbreviated dataset from EDER 2017, we calculated the number of unem- ployed individuals for each year. This was done to analyze the movements of the Mexican pop- ulation who left agricultural occupations, particularly to determine whether leaving these occupations resulted in unemployment or a transition to another occupation, either within or outside of agriculture. We counted the number of times individuals exited agriculture over their lifetimes, as well as the number of unemployed individuals during the period under analysis. To explore how climate variability impacts agricultural employment in Mexico, the rela- tionship between mean annual precipitation (mm) and both the planted area (ha) and the pro- duction value of rainfed agriculture (in thousands of pesos) was examined. Precipitation data were sourced from the Climate Research Unit [49], while data on planted area and production values were obtained from the Statistical Yearbook of Agricultural Production (SIAP) [61]. The data analysis involved adjusting the production value to account for inflation, using the

index provided by the National Institute of Statistics and Geography (INEGI) [62]. Subse- quently, the annual changes in income and planted area were calculated, expressed as percent- age variations. Finally, a correlation between mean annual precipitation and changes in the planted area was conducted. 2.3.2.2 Correlation between climate variability and agrarian occupation. Two scales of com- parison were carried out. First, for EDER 2017, we compared annual data on total agrarian occupation, rainfed occupation and animal husbandry occupation against the mean annual precipitation. We used Pearson’s correlation coefficient to compare these variables and calcu- lated the statistical significance using this method. For ENOE, we employed mean quarterly data on occupation transitions and precipitation, which we correlated using the Pearson coeffi- cient, and also computed the statistical significance with this method. Only significant correla- tions are presented (Fig 1). PLOS ONE Agrarian abandonment and extreme events PLOS ONE | https://doi.org/10.1371/journal.pone.0313891 February 10, 2025 7 / 22