The Monetary Authority of Singapore plans to change the ways in which it collects data from financial institutions. It means to reduce duplication as time goes on and automate the ways in which financial institutions send it data, starting on 31 March.

For one thing, the MAS will follow a new doctrine of 'data reusability' (its intention to use data that it collects once for many purposes) so that financial institutions need not send it the same data twice.

Also, the regulator is going to allow financial institutions to decline any requests it might make for structured data that they have already provided in their regular regulatory submissions to it (i.e. in their regulatory returns). This rule, which is only to apply to data that the regulator has asked for or received after 31 March, will eventually apply to other data submitted to the MAS through surveys and one-off requests. The aim is to eliminate all duplication in data requests by the end of 2019.

Machine readability

All new regulatory returns from financial institutions to the regulator from 1 April onwards will have to be in machine-readable formats, i.e. in forms that allow computers to read and process data automatically. The machine-readability format will be extended to new surveys and ad-hoc data requests next year.

The regulator is going to seek feedback from financial institutions about the provision of data in the machine-readable templates before next year. It will take this feedback into account before the second phase of implementation.

From 31 March on, submission templates in machine-readable format will apply to new regulatory collections (mandatory data collections issued under MAS-administered law) that contain only structured, alpha-numerical data (unstructured data, such as emails or annual reports, need not be submitted in machine-readable format); or that contain only data that exists in electronic form; or that are issued under an MAS-administered law on or after 31 March.

Machine-readable templates will include both partially machine-readable formats, such as MS Excel-based files, and more fully machine-readable formats, such as CSV files. Regulated entities may only submit data using these templates. This will not apply to data collected under existing regulatory collections.

How grainy should collected data be?

The dark side of this policy is the fact that firms are going to have to give the regulator far more data than before. Where it can, it wants to collect detailed data on underlying transactions instead of aggregated statistics. For example, instead of asking banks to submit separate statistics on loans extended to each industry and in each country, it might ask them for each loan labelled with the industry and country with which that loan is associated. It claims that its aim here is to reduce the reporting burden that financial institutions have to shoulder, but it is hard to see how this cannot lead to greater reporting volumes. It goes on to say that it is "working with the industry" (i.e. talking to a small number of selected banks) to work out how granular its requests ought to be.