Digital Money: The Rise of the Entrepreneurial State

Citi Public Digital Money | The Rise of the Entrepreneurial State | March 2022 21 Citi Public Digital Money | The Rise of the Entrepreneurial State | March 2022 20 This section presents a detailed summary of Digital Money Index (DMI) construction methodology. The Index is constructed from 77 data feeds (referred to as indicators) originating frommultiple opensource databases. The DMI website provides more details on the set of indicators used to inform the index. These feeds are grouped to compose a single index score following several layers of normalisations and transformations described below. II. Methodology a. Pre-Processing and Deprecation First, to produce a more up to date index some of the previously included indicators were deprecated due to discontinuation or irregularity of survey/ database updates, correlation analysis of indicators, and poor reporting frequency. Furthermore, to make the index more current a host of newer indicators were added across many open sources. b. Processing There were lot of gaps in the dataset and two methods were explored for dealing with missing data. In the last iterations of the DMI the procedure was to calculate a K-cluster mean. A k-means unsupervised clustering algorithm was applied to a subset of indicators which contained no missing values. This model was re-run under several different initial conditions and optimization of the corresponding variance- ratio criterion led to a consistent identification of 10/11 clusters. The K means average was used to replace missing values. However, this year, missing values are replaced with zeros. This method pushed the overall index scores lower by “penalising” non-reporting countries rather than make any assumptions about country similarities. Appendix: 2022 Index methodology — what has changed? c. Normalisation & Dropout Once the data was collected, cleaned, and matched, the individual indicators were min-max normalised (0 to 1) and aggregated by categories (called pillars). The previous version of this index also made use of dimensions which inherently added unequal weights to each indicator. However, the current version of the index aggregates indicators at the pillar level without weighting them at the dimension level. Furthermore, given the role of governments in driving money readiness and adoption, this year’s index has altered the weightages of the 4 pillars that define the overall index. Government and market support (pillar 1) has been weighed at 40% while the 3 other pillars (Financial and technology infrastructure, availability of solutions and propensity to adopt) were weighed at 20% each, compared to all of them being weighted equally (25% each) until 2020. Finally, the pillars were in turn also min-max normalised and aggregated to produce a final mean which is the DMI index.