For many years the analysis and design of flexible pavements has been based primarily on empirical methods. However, with the development of new technologies and better understanding of material behavior, research centers around the world are developing mechanistic-empirical methods that allow a more efficient use of materials in road construction, which can be translated into a better investments in road infrastructure. The aim of this study is to develop permanent deformation models for two granular materials used in Costa Rican roads as a sub-base or base and a subgrade soil, in order to predict the response and deterioration of these layers within a flexible pavement structure. Based on a typical regression analysis, it was initially concluded that the variation in material behavior depends on factors such as the number of loading cycles moisture content, and; deviator and confining stresses. These factors were then defined as the predictor variables that constitute the model and describe the permanent deformation development for each material. The models' efficiency was improved by using a statistical technique based on panel time series since the autocorrelation coefficient for each material is considerable as it deteriorates, which shows that, in fact, the collected data is correlated in time. In addition, the variance associated with the heterogeneity for each test condition shows that most of the unobservable error can be quantified when using panel data models, implying that very little information is not captured by the developed model when the estimation is performed.