IFRSassistant - Shiny Contest Submission


Authors: Asitav Sen

Abstract: IFRSassistant is being built to provide convenience to Finance and Accounting consultants and SMEs in finance industry.
(Update: I received complains about the app crashing. Have updated allotted resources.)
IFRS reporting, especially estimated loss calculation is a fairly complicated process involving Monte Carlo simulation, forecasting, survival modelling (or other predictive algorithm) and financial mathematics. It is time consuming and has its share of hassles. This app attempts to avoid the hassles and to perform the multistep analysis in a few clicks. One can download the report as well in pdf.

Full Description: IFRS reporting is fairly complicated affair. IFRSassistant attempts to bring some relief. This app, at this moment is capable of simulating and estimating credit loss across a lifetime (up to 5 years).

IFRS 9 requires a complex calculation of simulated possible losses. It starts with estimating probabilities of an asset going bad at different points in the asset's life. There is a need to model this using macroeconomic data. Then, estimated exposure on default is calculated and finally the reduction (or not) on these possible exposure is also estimated, based on the estimated value of sales of the asset/hypothecated asset or collateral. The app does these and provides information at various stages of the calculations and simulations. The final information that the app provides is the overall estimated loss by year, up to 6 years (current + 5 years).

The app loads with sample data. Data is shown in a table. This is followed by showing some basic information about the portfolio. Then the user needs to select country of the portfolio and click a button to fetch macroeconomy data of the country from IMF. Currently quarterly GDP and Price information is used. After the data is fetched, the combined data is shown in another table. In the meantime, these macroeconomic data is forecasted as well. (Forecast plots are not shown in the app, but is provided in the pdf report).


The user then need to click a button to start model selection. A survival model is selected and a plot showing distribution of probabilities of default by year is generated.
In the next step the user needs select parameters like the discount rate for present value calculations and some estimations of the depreciation of collateral value.
Using the discount rate, the exposure at risk by year is calculated. And combining with other parameters, Monte Carlo simulation is performed. This simulation provides a wide range of possible losses per year.
Simulation result of each year is plotted and a final plot showing weighted sum of possible losses by year is also generated.

At the end, the user can provide her/his name (Used as author in the generated pdf) and download a report in pdf.

This is a small beginning of a bigger effort to make complete IFRS reporting a very easy affair, at least for the SMEs. For more details, please do not hesitate to visit the github page.

Note: As you may imagine, the app is memory and CPU intensive. Nonetheless, I have uploaded the app in Rstudio Cloud and ShinyApps.

Keywords: IFRS, Finance, Monte Carlo Simulation, Survival Analysis, Forecasting, Credit
Shiny app: https://asitav-sen.shinyapps.io/IFRSassistant/
Repo: GitHub - asitav-sen/IFRSassistant: Shiny app for IFRS provisioning and estimated loss report
RStudio Cloud: Posit Cloud


Full image:

1 Like