Location: Conf. Hall 2, HQ2-01A-830
The iLab, ITD, SPR, and STA recently held the Big Data for Surveillance Challenge . The Challenge ran over four months and sought proposals for how to use big data—both new data sources or analytical techniques—to improve the Fund’s surveillance work. It generated 24 submissions, with 921 staff participants. Eight final teams attended a four-day innovation bootcamp , and pitched their ideas during the Annual Meetings. The three winning projects will now move into the iLab Accelerator program, with seed funding to implement their proofs of concept. The iLab is also in discussion with the remaining finalists about how best to further their projects.
· Maksym Markevych (LEG)
· Maksym Ivanyna (ICD)
· Ruben Atoyan (EUR)
· Steve Dawe (LEG)
· Samuel Romero Martinez (EUR)
Idea: We see significant potential in combining SWIFT’s big data and machine learning tools to analyze patterns of financial flows and detect possible financial crimes. For example, Moldova recently experienced a major money laundering scandal. An analysis of SWIFT data helped inform the mission team’s discussions with the authorities. Our project’s goal is to set up a red flag system that would automatically identify jurisdictions, specific payments, or financial flows that suggest possible systemic money laundering, banking frauds, tax evasion, corruption, or pressures on correspondent banking relationships.
Takeaways: The Challenge provided a unique platform for support, feedback, and accelerated development of ideas. It also included a boot camp for us to refine potential solutions and a pitch event where we distilled complex issues and proposed a potential solution in short speeches. We learned that our innovative ideas could add significant value to many streams of Fund’s work.
· Diane C. Kostroch (STA)
· Rasmane Ouedraogo (AFR)
· Rita Mesias (STA)
· Pheabe Morris (LEG)
· Luisa Nolan (UK Office for National Statistics (ONS))
· Pedro Rente Lourenco (Vodafone Group)
Idea: Our idea is to use Mobile Money Transfer (MMT) data to fill existing data gaps and derive new or more accurate estimates of remittances, financial inclusion, and other high-frequency indicators. This would allow us to more effectively monitor macroeconomic and financial developments in low-income countries (LICs). To transfer insights to other IMF member countries, we will compile a digital toolbox that documents lessons learned in how to access anonymized individual data and includes instructional materials. The goal is to help LICs enhance existing statistics by accessing MMT data from telecommunication companies and provide new insights for surveillance. Three East African countries (Kenya, Tanzania, and Uganda) have agreed to participate in the pilot project.
Takeaways: The Challenge was a rich learning experience for two reasons. First, the bootcamp sessions were extremely helpful to refine and design our project in a way that lets us present it to a general audience in an “elevator pitch.” Second, we learned a lot from the training presenters and other participants of the Challenge.
· Chengyu Huang (Research Officer, SPR)
· Andre Leitao Botelho (Data Scientist, ITD)
· Yang Liu (Data Scientist, ITD)
· Agustin Roitman (Senior Economist, SPR)
· Yunhui Zhao (Economist, SPR)
Idea: Changes in sentiment are important drivers of market stresses and financial crises. Yet measuring sentiment has always been difficult, and previous studies mainly relied on subjective and infrequent surveys. Recent developments in big data and other techniques make it possible to extract information from new data sources. Building on our previous work on news-based sentiment indicators, our project will apply cutting-edge techniques to a large news dataset. The goal is to enhance the IMF’s early warning systems with a new set of sentiment indices, focusing on predicting fluctuations in critical macro-financial indicators, such as exchange rates, stock prices, and sovereign bond spreads.
Takeaways: The Challenge was a very enriching experience, and we learned a lot about how to pitch a project highly relevant to the Fund. Among the lessons we learned are these. First, make it operational. A project that is practical, policy-oriented, and accompanied by a user-friendly tool is more likely to gain traction. Second, make it client-oriented. Direct and early interactions with end users could help eliminate blind spots and enhance the usefulness of the project. Third, start small, and stay agile. Ours is a new topic; we need to move step by step and keep an eye on what other people are doing.
“It has been great to see both high levels of staff engagement and ingenuity in the proposals. The ingenuity of these initiatives is the result of teams drawn from across departments, backgrounds, and hierarchy. You will see that rank has not predetermined leadership. This is a very welcome approach that showcases empowerment and rewards risk taking”. —Louis Marc Ducharme, Director, Statistics Department
Who Drains Bond Market Liquidity in an Emerging Market?
ITD, WHD, and MCM
Yang Liu, Hui Miao, Christian Saborowski
Can Big Data Help Us to Measure Remittances, Financial Inclusion, and Other High-Frequency Indicators?
STA and AFR
Diane Kostroch, Rasmane Ouedraogo, Rita Mesias
Lessons from Tax Reforms in Low-Income and Developing Countries (LIDCs)
FAD, AFR, APD, MCD, and ITD
Valerio Crispolti, Roberto Perrelli, David Amaglobeli, Papa Niang
Climate Risk Index
MCM, SPR, and RES
Alan Feng, Jorge Chan-Lau, Dalya Elmalt, Felix Suntheim
Transform Open-Source Satellite Images into Country Sustainable Development Indicators
STA, AFR, FAD, iLab, IEO, ITD, and SPR
Ayan Qu, Sandeep Sreekumar, Chenju Chakravarthy, Nicoletta Batini, Jiaxiong Yao, Ian Parry, Denisa Popescu, Alexis Meyer Cirkel
Does Sentiment Lead Market Stress? A Machine-Learning Forecasting Tool Using Sentiment Index
SPR, AFR, and ITD
Yunhui Zhao, Yang Liu, Agustin Roitman, Chengyu Huang
Using SWIFT Data and Machine Learning for Financial Integrity Surveillance
LEG and ICD
Maksym Ivanyna, Maksym Markevych, Steve Dawe
A Dynamic Projection and Recession Tracker
WHD, RES, and ITD
Emilio Fernandez Corugedo, Li Zhao, Kadir Tanyeri, Li Tang, Shuyi Liu, Allan Dizioli
This is the final event of the Big Data for Surveillance Challenge for IMF staff, where the finalist teams will pitch their ideas to a group of judges including Louis Marc Ducharme, Statistics Department Director, and YOU! The event attendees will have a chance to cast their vote to represent one “virtual judge.” The winners will receive funding of up to 50K towards their project and will be invited to participate in the IMF iLab Accelerator Program.
Join the event to hear the proposals on how data or techniques can support surveillance in the following areas:
Leading Indicators – Identify key turning points
Activity – Monitor key economic activities
Risks – Monitor key economic risks
Data Gaps – Address data gaps or lack of timely data, especially in low income countries and fragile states, or support the measurement of SDGs.
The challenge ran over the past four months and sought proposals from all IMF staff on these topics and the finalist projects were shortlisted based on the criteria of impact, extensibility and team strength. The IMF runs about 2-3 internal challenges for staff a year, and are launching the first external innovation Challenge on Anti-Corruption on October 18, 2019. Read more here.
Louis Marc Ducharme, STA Director, IMF
|Who Drains Bond Market Liquidity in an Emerging Market?||ITD, WHD, MCM||Yang Liu |
|Can big data help us to measure remittances, financial inclusion and other high frequency indicators?||STA, AFR||Diane Kostroch |
|Lessons from Tax Reforms in Low-Income and Developing Countries (LIDCs)||FAD, AFR, APD, MCD, and ITD||Valerio Crispolti |
|Climate Risk Index||MCM, SPR, RES||Alan Feng |
|Transform Open-source Satellite Images into Country Sustainable Development Indicators||STA, AFR, FAD, iLab, IEO, ITD, SPR||Ayan
Alexis Meyer Cirkel
|Does Sentiment Lead Market Stress? A Machine Learning Forecasting Tool Using Sentiment Index||SPR; AFR; ITD||Yunhui Zhao |
|Using SWIFT Data and Machine
Learning for Financial Integrity Surveillance
||LEG, ICD||Maksym Ivanyna |
|A Dynamic Projection and Recession Tracker||WHD, RES, ITD||Emilio Fernandez Corugedo |