The Client
In Feb 2015, We have had an opportunity to work with a top wealth management firm based out of the US.
Project Background
The client had a large amount of high-frequency financial data which can be widely used to study behaviour of alternative assets in relation to market and other macro-economic issues, including price discovery, market trends, strategic behaviour of market participants, and modeling of real-time market dynamics. The Client was looking for a strategic partner who can analyze the data on multiple dimensions enabling them to facilitate real time decision making.
Analytical Challenges
High-frequency financial data possess interesting and unique challenges to econometric modeling and statistical analysis.
- The number of observations in high-frequency data sets was overwhelming
- Data often recorded with errors and need to be cleaned and corrected prior to direct analysis
- Discrete price movements, nonsynchronous trading, and bid-ask bounce was distorting inferences based on standard statistical models
- Need to build a statistical measures of the forecast performance
Project Objective
- Provide a practical guide to high-frequency financial data analysis by preparing data for analysis and creating relevant variables, and to performing basic descriptive and graphical analysis
- To build correlation among the different market indicator, to enable real time decision making
Our Solution
Functions were created in R programming to perform data manipulation and descriptive analysis over specified trading periods, and to construct variables frequently used in the analysis of high frequency time series. We preformed the following activities for the descriptive analysis of high frequency financial time series:
- Creation of data sets and functions that contain standard and complete information from the database.
- Loading of data sets for the further analysis by creating functions and set the path.
- Data examination and cleaning by filtering the data using the functions.
- Construction of market variables (Price Change, Duration, Spread between bid and ask quotes, Exchange rate, GDP growth, etc.) that are of direct interest to high-frequency financial data analysis.
- Implementing the functions for the observation of trade direction and volatility.
- Uncovering the Statistical Properties of Market Variables.
Results and Conclusions
- Financial market activities were observed via periodic calendar patterns.
- The extracted calendar patterns used to adjust market variables before fitting econometric models to high-frequency financial data.
- Historical stress testing was done with 85% accuracy on price discovery.
- Pro-active defense strategy for client portfolio specifically focusing on long term capital gain.
Technology:
- MS SQL Server
- R (for analysis and reporting data creation including R based programming)
- MS-Excel and MS Word (for interactive reporting)
