More data help to better assess price risks

27 October 2015

Professor of Financial Econometrics Peter Boswijk conducts research into the risks of securities portfolios. He is delighted with the availability of increasingly sophisticated data. ‘Our study contributes to making financial risks are ever more manageable.’

In the financial world Big Data is indispensable. Investors base their decisions to a large extent on variables such as corporate profits, trading volumes and historical price trends and the way they correlate. This data that may help to understand the functioning of the market and perhaps make predictions about its future development. Quite often, investment decisions are left entirely to computers that rapidly react to several  such variables, in what is referred to as ‘algorithmic trading’.

‘Big Data is a challenge for econometrics’, says Boswijk. ‘Financial data can be used to find explanations and make predictions, but data from other areas may also give new insights. Such as, for example, the analysis of messages from social media or newspapers and the relationship between these data and share prices. This requires a broad cooperation between different disciplines.’


Volatility of financial securities

Boswijk is currently focused on a small component within the field of financial data. His research focus is on the volatility of market prices of securities, such as stocks or bonds. Volatility is a measure of movements in the prices of these securities. If anything can be said about the volatility of market prices, this gives a clear insight into the risk incorporated in such shares or bonds, says Boswijk.

According to Boswijk volatility plays a role in selecting securities within certain portfolios or in determining prices of options on stocks. Volatility is also being used in risk management. ‘The latter is important to me. This involves measuring the probability of a future decline or increase in value of an entire portfolio of securities. This kind of risk assessment is important for example for banks that have to constantly keep an eye on whether they have enough capital to even out the high-risk portfolios they hold. Bank supervisor are also interested in this kind of information.’

‘In order to say something about volatility in the future, it makes sense to look at past volatility’, says Boswijk. ‘Historical returns, of, for example stocks, have little predictive power for future price trends, however volatility can be predicted by looking in the past.’

Better results

Until recently, researchers tried to predict volatility by establishing historical time series based on relatively crude data. The volatility of a stock on a particular day would be established based upon returns over the past few days, calculated on the basis of closing prices. The availability of more sophisticated data offers new possibilities. ‘Regarding equity returns, I also incorporate data that is generated during the day. This data already existed, but since recently they are also being recorded in historical data sets.’

According to Boswijk, the use of more sophisticated data has two positive effects. ‘First, this enables us to see more closely how high volatility really was. This way, we can improve the comparison between our model predictions and their actual outcome. This, in turn, enables us to make better choices in creating the right model.’ It is also important to improve the quality of the predictions themselves. ‘Thanks to this process, our estimates of future volatility are considerably better than they used to be.’

Boswijk nevertheless encounters all kinds of statistical problems that make it difficult to make good predictions. ‘The large amounts of data we use should be adjusted for certain noise. This requires new and special techniques. In addition, one has to take into account the fact that the data we use is not always continuous. Securities sometimes show large price fluctuations during the day. This causes a certain discontinuity in the data that has to be separated from the normal volatility one wants to measure.’

When measuring the volatility of a portfolio as a whole, one should also measure the interdependence of the volatility between individual securities, known as covariance. ‘One problem is that securities are not traded throughout the world at the same time. If one fails to take that into account, one could sometimes come to the conclusion that there is no connection between certain effects at all, when actually there is one. This can be solved with statistical techniques.’


‘I hope this research will make risk more manageable’, says Boswijk. ‘Banks work with mandatory capital requirements, which were formerly based on less accurate and – more importantly - sluggish risk measures. By being able to better estimate future risks, buffer requirements can be adapted to changing market conditions more rapidly.’

A better prediction of future volatility could theoretically mean that a lower degree of uncertainty needs to be built into capital requirements. However, supervisors tend to require increasing buffers these days. According to Boswijk, this is caused by the fact that the risk of portfolios is also determined by other properties, in addition to the volatility, which are currently less predictable.

More information? Email

By Bendert Zevenbergen

Published by  Economics and Business