Atributos de gobiernos corporativos universitarios y predicción de los niveles de divulgación de información: una aplicación de árboles de regresión

Abstract

In the area of university governance and management, Abello (2018), Abello et al. (2018) and Abello et al. (2019) have shown that on average the level of information disclosure by Latin American universities is low and very heterogeneous; they have also shown that the attributes of university corporate governments or maximum collegiate bodies (MCC) are determining factors in the disclosure of information. Although the data from more than two hundred universities that were used in this work represent a high number of observations, they are only a proportion of all Latin American universities; likewise, the collection of information to determine the levels of information disclosure could become complex, the need to predict and estimate what would be the expected level of disclosure from other universities emerges. In view of the above, this work aims to predict the levels of information dissemination of Latin American universities considering the attributes of their corporate university governments or top collegial bodies (MCC). To achieve the above objective, it is introduced with a technique used in other disciplines and for other purposes called regression trees. While there are other techniques, such as linear regressions, classification and regression trees are part of machine learning techniques to build predictive models from available data (Loh 2011). Thus, for the purpose of this work, regression trees may be an appropriate methodology to predict the levels of information dissemination in Latin American universities. The technique will generate nodes or branches according to the attributes of the CQI, which will allow typing the universities to later make a prediction of the level of information dissemination considering these attributes. The name of the technique derives from the common practice of describing the process of data partitioning by means of a decision tree and one of the advantages of regression trees is that they are easy to understand and interpret (Hernandez 2020). The results of this work indicate that the universities whose CQI are more diverse and more independent have the highest rates of information dissemination, while on the contrary, the universities whose CQI are less diverse, the number of members is 49 or less, and they are also less independent, are the institutions that have the lowest rates of information dissemination.