This paper examines the suitability and usefulness of OECD definitions and cut-off levels for classes of R&D intensity as a classification scheme for developing countries by developing a method for determining R&D intensity classes from data mining techniques. This improves on the original methodology of Hatzichronoglou (1997) by using clustering methods to define the members and boundaries of a technology class, instead of arbitrary choices of cut-off levels for technology classes. The approach may be implemented on a global level, but is used to examine new categories of R&D intensity, using data on South Africa. It identifies the 'electronics-communications' sector as the only high R&D intensity industry, and it determines that only two R&D intensity classes were evident between 2010 and 2015 in South Africa. The results suggest that there is currently no high-tech industry in South Africa, and that a strategy that seeks simply to promote the export growth of such sectors is naive.
Human Sciences Research Council (Sat,) studied this question.