Anglia Ruskin University
Predicting the Impact of Social Capital for Microfinance through Crowdfunding: A Neural Network approach.
Emanuele Giovannetti and William Davies
The notion of innovation ecosystem has been recently utilised to shape policy advice for developing countries, (ITU 2016a). The building blocks of innovation ecosystems reside in the reciprocal relations among the ecosystem agents and in the social capital (Durlauf & Fafchamps, 2005) required to sustain them. ICT platforms are key for the formation of these relations and the underlying social capital across geographies. A key example of an ICT platform is that of Crowdfunding (Kromidha and Robson, 2016, Davies and Giovannetti 2016). This paper has the main objective to understand “How Social Capital, mediated through ICT-enabled crowdfunding platforms, can address the problems posed by the lack of traditional credit markets for micro-entrepreneurial activities.” In order to understand the role of ICT-mediated Social Capital we address the necessity of harvesting large datasets, including qualitative and quantitative information on thousands of crowdfunded microfinance projects. Our data was web-harvested from the Kiva platform, an international, not for profit, crowdfunding platform founded in 2005, whose role is to connect lenders and borrowers, financing early innovation and micro-entrepreneurship. Over the past ten years Kiva has enabled more than 1.5 million people to fund over 2 million borrowers in over 80 countries. The result has been nearly $1 billion dollars lent to borrowers and repaid at a rate greater than 97%. This paper introduces and analyses a large data set of KIVA crowd-funded projects, generated by the authors through a specific web harvesting programme, and studies the impact of social capital on lending, based on the centrality properties of a latent network formed by the set of bilateral relationships established among projects. These linkages are derived through the presence, or absence, of joint lenders, creating bilateral linkages between projects, via direct information sharing. The results were analysed utilising both Ordinary least squares (OLS) regression and a backwards propagating neural network. Backwards propagating neural network utilise the connections between inputs, a set of hidden units and outputs to minimize the error of nonlinear functions (Buscema, 1998). They have been shown to provide better performance than alternative classification methods, in the context of crowdfunding (Li et al 2018). Our results empirically support this finding with the backwards propagating neural network providing greater fit for the model in comparison to the OLS regression. Furthermore, our results show how the network eigenvector centrality (Newman, 2012), a key complex network relational variable used as a proxy for social capital, affects these crowd-funded projects’ raised amount of lending. The paper’s results show, as expected, that a project’s social capital, has a significant and positive predictive impact on the amount of finance raised and, hence, on the project’s viability. Finally, we have obtained time series of aggregate raised funds for the entire platform since its inception, and we compare, results of traditional time series forecasting with backwards propagation forecasting algorithms, both at aggregate level and across different project categories.
William Davies, MSc (ARU) collaborates with the Centre for Pluralist Economics at Anglia Ruskin University, has obtained a B.Sc. Economics from the University of Bristol. Continuing with acquiring a Master’s degree with merit from Anglia Ruskin University in International Business Economics. Worked as an economic consultant for three years and is currently working on a PhD at Anglia Ruskin University considering the topic of crowdfunding. Alongside developing his PhD, William has worked as a research assistant at the university, across multiple different projects. William, alongside his supervisor Emanuele Giovannetti, has recently published a paper to the journal of Technological Forecasting and Social Change, entitled “Signalling experience & reciprocity to temper asymmetric information in crowdfunding evidence from 10,000 projects”.