1 ‘Social Sharing’ By Means of Distributed Computing: Some Results From A Study of Hans-Jürgen Engelbrecht Massey University August
2 1.Introduction Information and Communication Technologies (ICT) are General Purpose Technologies. One of many associated innovations: Distributed computing, grid computing. Enables non-commercial sharing of physical, rivalrous goods via the Internet: Such ‘social sharing’ is a form of economic production (Benkler, 2004).
3 ‘Shareable goods’ Sharing of computing power and bandwidth. Two features of ‘shareable goods’ (Benkler, 2004): They are lumpy (PCs come in discrete units). They are of ‘mid-grained’ granularity (PCs are widely privately owned and systematically have slack capacity).
4 ‘Shareable goods’ ctd. What determines the extent of ‘social sharing’? Technological conditions, but also cultural practices and tastes (Benkler, 2004) and social and legal conditions (David, 2004).
5 Prime example of a voluntary non- commercial Internet-based distributed computing project: Launched in May Download screen saver. Analysis of Arecibo radio telescope data. the most powerful special purpose supercomputer in the world.
6 ctd. Worldwide phenomenon (except for Mauritius, Palestine and Vatican City). Incentives build into client interface, e.g. user and results data. By Dec. 2004, there had been: More than 5 million contributors. Providing over 2 million years of CPU time (more than 1000 years of CPU time during the last day alone).
7 ctd. SETI country data available for: Dec. 10 th, 2002; Dec. 11 th, 2003; Dec. 13 th, Dependent variables used in the regression model: SETI participants per capita. SETI results per capita (measures actual outcomes and is arguably a better Internet- intensity variable than ‘hours of use’).
8 3.Explanatory variables What determines cross-country participation and its intensity? Aim: To include as many countries as possible. Therefore, modelling is severely restricted and I use only a few key explanatory variables in the regressions: ITU’s ‘Digital Access Index’ (DAI). GDP per capita (gdp). The ‘Human Development Index’ (HDI). Country group dummy variables.
9 The Digital Access Index (DAI) ITU: The DAI tries to measure “the overall ability of individuals in a country to access and use ICTs…”. It provides the first truly global ICT ranking. The DAI is a composite index made up of 8 underlying indicators to capture: infrastructure (fixed telephone & mobile telephone subscribers), affordability (Internet access price), ‘knowledge’ (adult literacy, school enrolment), quality (broadband subscribers, international Internet bandwidth), actual usage of ICTs (Internet users).
10 Components of the Digital Access Index (DAI), 2002:
11 The DAI ctd. Hypothesis: The DAI is a positive and statistically significant determinant of participation and its intensity. This would mean: On average, participation and its intensity across countries matches inter-country differences in ICT accessibility.
12 Other explanatory variables GDP per capita (in PPP adjusted US $): Traditional proxy for ‘standard of living’. Key explanatory variable in numerous ICT and Internet diffusion studies. It is expected to be a positive and statistically significant determinant of participation and its intensity.
13 Other explanatory variables ctd. The HDI: A composite index which has emerged as the preferred measure of ‘development’. It measures important dimensions of human development neglected by gdp, such as: living a long and health life and being educated. It is best included alongside DAI and gdp as an additional explanatory variable.
14 Other explanatory variables ctd. Country group dummy variables: ITU’s “developed & advanced countries” versus ‘the rest’. Alternatively: 6 regional dummy variables (similar to Caselli and Coleman II, 2001). See “Appendix: Country List”.
15 4.Regression analysis Matching data for 172 countries. Dependent variables alternatively in 2004 levels and changes. Most regressions estimated in double-log form. OLS with White’s heteroscedasticity correction. Box-Cox regressions.
16 Regression results
17 Regression results ctd. Increasing DAI and gdp by 1% increases dependent variables by a similar %tage (elasticity of ‘change in results per capita’ with respect to DAI somewhat lower). DAI, gdp, and the general divide between rich&poor countries can explain most of the cross-country variation in participation and its intensity (see R 2 s). HDI dropped from preferred regressions (DAI and HDI highly correlated).
18 5.The global digital divide By Dec. 2004, developed & advanced countries (about 15% of the sample population) accounted for over 90% of submitted results. But: Indications of a slowly narrowing global digital divide! Growth rates for ‘users’ and ‘results’ higher in ”the rest”.
19 Developed & advanced countries versus ‘the rest’:
20 Developed & advanced countries versus ‘the rest’ ctd.:
21 Developed & advanced countries versus ‘the rest’ ctd.:
22 6. Concluding comments Further research needed: For a less heterogeneous group of countries. This would allow more sophisticated modelling. More sophisticated models are needed to enable more specific policy conclusions. Will non-commercial ‘social sharing’ via the Internet become a dominant mode of economic production? There is huge potential for it, but commercial distributed computing might greatly affect its realization.