Exploring the relationship between entry requirements and throughput rates for honours students

Type: 
Research Article
Number of pages: 
6
Published: 
28 September 2017

Export citation

Views 117
PDF110 downloads
EPUB62 downloads
XML55 downloads

Abstract: 

In order for a student to enrol in an honours programme at the University of KwaZulu-Natal (UKZN), a weighted average mark for their final year of undergraduate study must exceed a particular threshold value. Students are then ranked according to this weighted average mark, with entry into the honours programme offered on a top-down basis, within the constraints of teaching resources and space. A proposal has been made at UKZN to remove existing barriers for entry into an honours programme, i.e. to allow entry to any student who has completed a 3-year undergraduate degree with a major in that discipline. The impact of such a decision was investigated. By lowering the requirement for entry into an honours programme, one is expected to predict how a new cohort of students will perform. Apart from obviously having a lower weighted average mark for their final year of undergraduate study, these new students may also differ in other unobservable ways which need to be accounted for. In a regression modelling context, one is asked to predict outside the range of a collected data set. A Heckman selection model was used to account for a possible self-selection bias that may arise because the subpopulation for which a prediction is required (namely those new students who will now be able to enter an honours programme), may be significantly different from the population of UKZN undergraduate students who are currently permitted entry to an honours programme.

Significance: 
  • A modelling technique that accounts for a possible sample selection bias was used to determine the impact of lowering the entry requirements into the honours programme at UKZN to allow entry to any student who has completed a 3-year undergraduate degree.

Keywords: 

Heckman model; sample selection bias; regression model
Views and downloads are with effect from 29 January 2016.