The world’s largest pension fund is learning how artificial intelligence (AI) will help it find, watch and assess its fund managers.
Japan’s ¥158 trillion ($1.4 trillion) Government Pension Investment Fund (GPIF) Global has a high outsourcing ratio in contrast to smaller funds, and with outsourcing has come less than stellar results over the past decade, it said in a March report. Everything except foreign bonds hasn’t produced alpha in that timeframe. The plan’s portfolio mix is almost an even split between stocks and bonds with a sliver of short-term assets.
“On the other hand, payments to asset management companies and asset managers amounted to 9.9 billion yen for domestic bonds, 13.7 billion yen for domestic stocks, 12.8 billion yen for foreign bonds, and 34.5 billion yen for foreign stocks (a cumulative total for three years). They are not negligible and make no sense,” the document added.
The fund assessed the situation, and began a collaboration with Sony Computer Science Laboratories, which led to the development of a prototype system that would use AI to collect trading data to determine the investment styles of its money managers.
The program, dubbed the “Style Detector,” analyzed eight different investment strategies (high dividend, minimum volatility, momentum, value, growth, quality, fixed weight, and technical) to help determine how its managers operated. It then programmed the system to monitor the styles against various market and corporate performance as well as varying fund conditions. The fund could then use this data to determine which strategies, and in turn which managers, are more aligned with its goals.
In the future, the fund will utilize the Style Detector to determine what works best for its performance, and managers will have to up their game if they want to play in the GPIF sandbox.
“The fund managers will be aware of the fact that GPIF has the ability to analyze investment styles in real time. The style and drift of individual fund managers will also be detected, and that may encourage a certain discipline among the fund managers. In addition, GPIF will be capable of analyzing the behavior of multiple fund managers from a comprehensive perspective—a bird’s-eye view of fund manager investment styles,” said the report.
Should a manager’s performance become questionable in relation to its fees, GPIF will know why and may choose to either cut them or give them a second chance. “When asset management companies recognize that GPIF has the ability to independently analyze their investment styles and intends to continue development of even more advanced technology, they will recognize that they cannot justify their results with only qualitative explanations,” the report said.
Should it choose to keep an asset manager, the manager may need help from some AI of its own to stay in the race. “As a result, asset management companies will take action to improve the efficiency of their investment process by introducing more sophisticated technologies, including AI, to explain their behavior and be accountable for their investment practices,” said the paper.
GPIF believes the process will accelerate machine learning in the asset management world, while also ending the “dependence on individual persons from management strategies, and promote optimization.” It also says the developments will further promote the science and technology of asset management.
“First, such a system benefits GPIF by contributing to the construction of a more sophisticated manager selection system. Second, it benefits the asset management industry as a whole if GPIF’s insights can be shared with fund managers to improve their performance and risk management,” the fund said.