Refinitiv on How to Harness the Full Value of Research Using Data
Steven Carroll of Refinitiv, an LSEG Business
Aug 26, 2019
The next digital ‘arms race’ will be around data. Just how do you process the ever-proliferating amount of information and turn commoditised content into actionable insights? As an investment adviser or portfolio manager, how can you identify which sources to trust? Steve Carroll, ASEAN Managing Director at Refinitiv, presented an absorbing Workshop to explain how building superior analytics can help better understand the investment landscape and ultimately deliver wealth management clients better performance with lower risk profiles. The Refinitiv solution, StarMine, provides analysts, directors of research, advisers and wealth advisory relationship managers with a unique set of tools to measure and manage analyst performance and generate new ideas, or perhaps cautionary alerts, for investors worldwide.
Carroll began by explaining that his Workshop was all about Refinitiv’s ‘StarMine’ solution to data mining and analytics. He made his detailed and fascinating presentation with reference to a screen-based slide show that gave delegates a detailed insight to StarMine, its capabilities and its applications.
Why choose StarMine?
“StarMine allows customers who use it to identify opportunities, save time, and zero in on the most viable investment ideas,” Carroll reported. “Using StarMine in the investment process is like adding an entire research department of PhD-level experts to your team. Our suite of quantitative analytics and models covers critical areas including value, momentum, ownership, risk, and quality.”
He explained that customers can make better, faster investment decisions using StarMine’s quantitatively derived outputs to simplify the stock selection process. “For example,” he explained, “StarMine’s SmartEstimates places the most weight on recent forecasts by top-rated analysts, helping you predict future earnings and analyst revisions.”
In addition, StarMine uses a range of quantitative models to evaluate any company, including valuation, momentum, and earnings quality. “Customers using StarMine can therefore introduce new angles to investment strategy and test investment hypotheses with StarMine’s analytics and models,” he explained, “as well as validate and benchmark their own quantitative methods.”
StarMine’s many shafts
StarMine comprises a variety of key components. Predicted Surprises offers directionally correct predictions of earnings surprises using weighted forecasts from top-rated analysts, the Predicted Surprise has a 70% accuracy rate for determining the direction of surprise. StarMine also produces Smart Estimates - “SmartEstimates use only the most recent estimates, with both a static and dynamic measure of relevance, then with those remaining place more weight on the Analysts track record. These estimates provide a better measure of all estimates types (Revenue, EPS, etc.) and can be used to provide more accurate growth rates, valuation ratios, etc.
StarMine also includes the Intrinsic Valuation Model, which offers customers improved accuracy and stock ranking ability with more robust and reliable equity valuations. “This,” Carroll explained, “means that StarMine can help advisers and researchers adjust for over-optimistic growth forecasts caused by bias in analyst estimates, and thereby improve their forecast accuracy and stock ranking ability. They can better identify cheap stocks poised for rebound, as well as overpriced ones likely to revert. They can also better predict the persistence of earnings, drawing on StarMine’s quantitative multi-factor approach.”
The world of credit
StarMine does not limit itself to equity and earnings. The StarMine Structural Credit Risk Model also captures almost 85% of default events in a 12-month horizon and bottom quintile of scored companies.
And on top of that the Text Mining Credit Risk Model is a unique quantitative signal that systematically analyses a large body of previously untapped qualitative data to help customer better predict credit risk.
“Taking a multi-pronged approach to predicting credit risk and default probability is highly valuable,” Carroll reported. “StarMine draws on complementary sources of data and analytical methods so you can quantitatively assess and predict credit risk. Our unique Text Mining Credit Risk Model identifies language predictive of credit risk by applying sophisticated algorithms.”
Finally, SmartEconomics provides enhanced forecasts of macroeconomic data and FX rates using the historical accuracy of contributors to Reuters polls and applying weightings.
Mining historic and real-time data
Stepping back from the detail, Carroll told the audience that what StarMine sets out to do is use the historical data at our fingertips to understand which analysts will give them more value.
“In the world of equities and stock selection,” he explained, “you can very quickly divide your investment universe into those companies likely to miss estimates versus those companies likely to hit or exceed. For the advisory community, it is a great tool for both idea generation and for risk management, in other words avoiding those companies likely to disappoint the market with their poor earnings.”
And he explained that the outcomes are not only related to earnings, as StarMine covers cash flows, revenues, book value. “In short, StarMine covers every single key metric that is in use when you are trying to assess the performance of a company. If the analyst covers any metric, then we have a smart estimate available.”
He highlighted how StarMine also includes a quantitative model, called the analyst revisions model. “This is often used by professional investors, by quantitative investment funds,” Carroll reported, “because it is a very strong forecasting of future change in revisions incorporating a lot of different data.”
Simple scoring metrics
He explained how, for example, Singapore’s CapitaLand scores extremely well, with a score of 92. “That means it is in the top 8% of companies in developed Asia, because it is experiencing positive revisions on both this year and next year,” Carroll noted, referring to his video presentation. “The 1-100 scoring metric is a really easy way for the investment community to determine what is going on with CapitaLand, or any other company. In simpler terms it is a measure of change in analyst sentiment, so whatever recommendation an analyst has, that is not necessarily a good predictor of price change.”
He also highlighted more of the big picture capabilities of StarMine. “Looking at the wider economic outlook,” he reported, “you can see the environment such as we have today, with significant uncertainty where people are nervous around the economic outlook, where they see a yield curve that is negative, where there is a significant chance of a recession in the next six to 12 months in the United States. When you have an environment with untold uncertainty like that, then the StarMine Earnings Quality model in StarMine performs extremely well. In fact, in the last 12 months, earnings quality has been our best performing model and particularly strong in emerging markets. It is helping investors to separate those companies with a low risk profile with attractive fundamentals, from those companies that maybe are reporting a little more aggressively.”
Carroll explained that in a brief talk he would not have time to go into that much detail on all of the different models that Refinitiv has built under the StarMine umbrella.
“One remarkable output I want to also highlight is our ability to look into investment funds,” he said. “In brief, we can take 2,400 asset management companies globally, look at all their different funds, mine down into their underlying investment characteristics and their holdings and then give them ratings based on roughly 25 factors that we use. This is incredibly useful in selecting out the best or most appropriate funds.”
Watching out for weakening financials
And as to the world of credit, he explained that StarMine is mastering the use of AI and machine learning to facilitate a probability of default for every company in the 23,000 strong equities universe StarMine covers globally.
“Using text mining technologies,” Carroll reported, “we can mine out any possibility of future financial distress. If you think about how scrubbed and cleaned a company’s annual report is, with lawyers having gone over every word, you’ll probably understand that there’s very little chance that you’re going to learn a lot about real risk in a report like that. There is a lot of other content out there that can add up to insights into their future credit stability, or lack of, that might produce an elevated risk profile for this company. And we do this every day for 23,000 equities.”
He explained that additional tools provide re-calculations on every company, every day, allowing customers to see almost in real-time, any key changes to creditworthiness. “Each evening, we produce a probability of default score, and we then back out of that to figure out what is the implied credit rating. If there is a significant difference between the credit rating of the rating agencies and the Refinitiv score, that can be flagged, and our research is showing that there is a very significant chance that we predict the next directional change of the ratings agencies. It is another great early-alert tool.”
And with that, Carroll closed his Workshop, remarking how comprehensive the StarMine suite of indicators are and how invaluable they can be for customers of all types, including the wealth management industry.
Managing Director, ASEAN at Refinitiv, an LSEG Business
More from Steven Carroll, Refinitiv, an LSEG Business
Latest Articles