account risks that are based on the recent historical data of single assets or portfolios.
This works relatively well under stable market conditions in which recent historical
behaviour is an acceptable predictor of the expected market behaviour in the near
future. However, it is also recognized that under abnormal market conditions historical
data cannot be adequately used as an indication for expected market behaviour. Con-
sequently, during such unstable periods, traditionally calculated risk values cannot be
trusted. This deceptively suggests that market risk measures are reliable unless there is
an abnormal market. In practice, such risk measures have a very limited applicability
under unstable market conditions and wrongly suggest approximate risk indication.
In essence, the effect of a sudden change in market conditions on future risk predictions
is neglected.
17.5 NORM goals
NORM aims to improve the reliability of risk metrics by incorporating the impact of
sudden market changes. The final goal is threefold
- To determine the likelihood that a market will suddenly change. This result can be
used to adjust the confidence level of traditionally calculated risk metrics. - To provide qualitative impact factors to adjust risk metrics. This result can be used to
adjust the actual value of the risk metric in combination with an adjusted confidence
level. - To provide quantitative impact factors to adjust risk metrics. Quantitative impact
factors would enable more accurate risk metric calculation while maintaining original
confidence levels.
17.6 NORM uses semantic news analysis technology
The technology that NORM uses to achieve these goals is semantic analysis of financial
news messages. We define ‘‘sudden market changes’’ as being preceded by significant
events that appear in financial news and affect one or more companies. NORM focuses
on the equity market, which enables it to restrict the definition of ‘‘events’’ to things that
affect the value of one or more equities. These events are, for example, not only company
takeovers, mergers, changes in C-level management, etc., but also market-dynamic
changes such as oil price changes and regional instability around production facilities.
A state-of-the-art semantic analysis technology is used to automatically detect
events in the news and correlate them with the occurrence of sudden market changes.
In order to distinguish between relevant and irrelevant events, a dataset of 20 years of
financial news history will be analysed and compared with the intraday market values of
a set of selected equities. This can be used to test the three goals described above.
Subsequently, a proof-of-concept application will be constructed that implements the
results, can run on current news and market data, and will be tested by selected early
adopters.
NORM—Behavioural finance with news-optimized risk management 321