The importance of good data in selecting stocks
Bart Van der Grient
Bart van der Grient explains how Robeco ensures good data quality, which is vital for selecting stocks and conducting empirical research. “Our approach makes the stock selection and portfolio construction process more transparent,” he says.
Bart van der Grient is the equity researcher who is responsible for data quality at Robeco Quantitative Research. He uses and develops tools to generate new stock rankings and to optimize the portfolios. Another important element of his work is the creation of historic databases for empirical research.
Any asset manager can buy data on stocks. What makes Robeco different?
"For us it is very important to stay in control. We sometimes use external tools, but for the majority of our work we develop our own. This in-house approach makes the stock selection and portfolio construction process more transparent. If the stock selection model indicates a particular trade in the portfolio, we know why. For example, it could be because of a stock’s valuation, or because it helps to decrease risk. We can also look more closely at the underlying variables if the trade is unexpected. There are external platforms that can do many things at the same time, but we prefer to start with the raw data and build in periodical checks using several applications."
'For us it is very important to stay in control'
"We have created many of these checks. For example, we analyze stocks that have moved significantly in our ranking. These changes can be legitimate, for example, as a result of good corporate earnings. But if a change results from an incorrectly handled stock split, we adjust for this event."
"A second thing that makes Robeco different is our experience, especially in emerging markets. We do not just have experience in our Quantitative Equity and Quantitative Research departments, but also in the fundamental emerging markets equities team. Close cooperation increases the quality of data, because in emerging markets equities there are many exceptions. For example, sometimes not all of a company’s net earnings go to its shareholders, because the government also receives a share of these earnings. We can check such cases with the fundamental team, because they have more in depth knowledge of individual companies. These discussions also help to choose the best data providers. We have preferred data suppliers, but if for a particular stock or group of stocks another data provider is better, we use this one instead. "
"We also cooperate closely with the fundamental team in choosing the best share class to trade in. You want to trade in the class where trading costs are lowest. This instrument does not necessarily have the best quality data linked to it. Therefore we developed tools to combine data from different sources and listings. For example, we can trade in the liquid ADRs (American Depositary Receipt) of a particular firm, while we select stocks based on the data of a domestic share class."
How do you guarantee good historic data quality for research?
"We use a bottom-up approach to build our database and always check the accuracy of the data. For example, by reconstructing the MSCI World Index with individual stocks and comparing it with the data of the index provider, we can validate return data. Such checks are done for all variables in our database."
"We especially want to make sure that we have data on all stocks, including those that no longer exist – in order to prevent survivorship bias. Even academic papers do not always take survivorship bias into account. It is important to include the data of firms that no longer exist, for example, because the conclusions about the performance of high risk stocks may otherwise be overly optimistic. For every successful high tech company such as Apple, many others have fallen by the wayside."
"One of the most challenging projects we worked on was gathering good data on frontier markets. Take for example, Pakistan, or countries in the Middle East and Africa. We have invested considerable time in building a high-quality database, which hasn’t been easy; there are practical problems such as different trading days. For example, in countries where there is Sunday trading – does it count as the beginning of one week or the end of another? We have frequent contact with data providers on these matters and sometimes they change their dataset as a result of our findings."
"In addition to the fact that we are fully prepared to include new countries in our universe, this database has allowed us to test our models on such frontier markets, which has provided additional evidence for the strength of our strategies."
‘In emerging markets equities there are many exceptions’
What is the level of cooperation with the portfolio managers of the Quantitative Equity team?
"Cooperation is very good. Whenever trades are proposed there is contact. There is a division of responsibilities – we are responsible for the data, run the models and conduct the research aimed at further improving the models."
"The portfolio managers of Quantitative Equity implement the model outcomes in the portfolios after checking the stocks rankings to apply the four-eye principle and monitor positions as well as risk exposures. We can comment on a particular trade and they can come to us with questions; so there is a feedback loop. It helps that many of the portfolio managers also have a background in quantitative research. Together we make sure to work with the highest quality data."