However, IE theory has yet to be extended to the selection of performance measures for BGs in the early phases of development compared to their more mature counterparts.
Thus it can be seen that selecting appropriate measures to evaluate BG performance is an especially challenging task, and one that requires the unique characteristics of BGs to be taken into account.
2.2 Measuring BGs' Performance Throughout their Development Processes
Although liabilities of newness do influence BG strategy and performance (Sleuwaegen and Onkelinx 2013), assigning them the generic label of young firms does not truly reflect the heterogeneity in the age of BGs. Rather, BG firms advance through numerous phases of development from initial idea generation through to growth and consolidation (Rialp-Criado et al.
Given that the challenges faced, strategy formation process undertaken, and decision making logic applied by BGs is unique to their phase of development (Rialp-Criado et al.
First, the pre-start-up and new venture creation phase can be quiet extensive (Hewerdine and Welch 2013) and have considerable implications for the BGs post-founding strategy and behaviour (Pettersen and Tobiassen 2012).
Second, immediately after founding, BGs enter their early international entry/development phase.
2004), suggesting that some BGs move into a phase of international growth and consolidation (Rialp-Criado et al.
2.3 The Influence of BGs' International Strategies on their Development Process
The internationalisation processes of BGs can also be expected to influence the pace at which BGs progress through the various phases of their development.
Furthermore, the performance of BGS with partitioning 3 turns out to be insensitive to random symmetric permutations of the input matrix and is much better than that of GS under the same symmetric permutation.
To compare the performance of BGS with partitionings 1 (y,n,n,l) and 3 (y,y,n,l) and to test our hypothesis further, two sets of fifty sparse matrices of order 500,000 with densities of 0.625 x [10.sup.-5] and 1.25 x [10.sup.-5] are randomly generated using Matlab.