Investment researcher StarMine seeks to stay ahead of the quantitative analysis it pioneered.
StarMine definitively jumped on the big-data trend by recently augmenting its forecasts of corporate health and credit-default risk with a tool that scans virtually everything written about companies in search of hints of future troubles. The tool aims to distinguish the Thomson Reuters division from competitors that are also seeking to automate the collection and analysis of new types of data, a strategy for which StarMine was an early pioneer.
Founder Joe Gatto started up the firm in 1998, during the dot.com bubble, when equity analysts became celebrities on newly flourishing cable business channels. His firm questioned whether any of the analysts’ forecasts were consistently good or at least better than others, and the approach he devised captured investors’ attention by successfully quantifying just that.
StarMine then branded the approach as “smart estimates,” which weigh analysts’ estimates by their historical accuracy and give extra weight to more recent forecasts, rather than applying a simple equal-weighted average. The result was a more accurate forecast of companies’ future earnings as well as revisions.
“When our smart estimate is different from the consensus, there’s more likelihood to be revisions in the direction of our estimate, and there’s greater likelihood to be an earnings surprise in the direction of our smart estimate when earnings are announced,” said George Bonne, director of quantitative research for Thomson Reuters StarMine.
Thomson Reuters bought the private company in 2008 and the division has retained its identity within the financial publishing behemoth.
“The brand has stayed intact, which probably means there’s a research team that continues to produce something unique and of incremental value,” said Paul Rowady, senior analyst at Tabb Group.
That value would appear to stem at least in part from expanding StarMine’s smart estimates model to assess the financial health and default risk of companies globally, now covering more than 40,000. The model groups various accounting ratios and industry-specific metrics into profitability, liquidity, leverage, coverage and growth components. It combines them into a final default probability that also takes into account the geographic region.
Services such as StarMine’s have become critical for corporate credit managers as well as procurement officers managing long, just-in-time supply chains, where the loss of one important link can dramatically impact production. Mr. Bonne said an important driver behind StarMine’s decision to pursue more accurate credit-assessment models was the use of clearly inadequate models leading up to the financial crisis, in which major counterparties imploded.
StarMine said the main advantages of its model include combining companies’ reported numbers and forward-looking analyst estimates; using metrics most appropriate for specific industries; and giving more weight to the most important accounting ratios for a specific sector. It claims its smart estimates model is more powerful than common alternatives such as the Altman Z-score and the Ohlson O-score, relying on more straightforward use of financial ratios.
“The model accurately predicts 80 percent of default events at the 20th percentile of model scores compared to 60 percent for the Altman and Ohlson models,” StarMine said.
Like many of its competitors providing more quantitative corporate-default or bankruptcy-forecasting models, StarMine builds upon the nearly 50-year-old Merton model, which models a company’s equity as a call option on its assets, in which the probability of default equates to the probability of the option expiring worthless. StarMine claims its improvements on the Merton model make it “considerably more accurate” in predicting defaults compared to the basic Merton model.
Those improvements include a leverage component, in which a higher ratio of liabilities to assets generally points to greater likelihood of default, as well as a volatility part in which a company with more volatility in the values of its assets is more likely to slip into default and bankruptcy.
Mr. Bonne called the third part of StarMine’s model, text mining, its most innovative feature and something no competitors touch on. Its technology looks for indications that companies are moving toward or away from financial distress, scanning any relevant text including news articles, conference call transcripts, corporate filings, and brokerage reports.
“That’s a completely different avenue for assessing financial health, and something complementary to the Merton and accounting-ratio analysis,” Mr. Bonne said.
It also highlights StarMine’s expertise in the realm of so-called big data, a concept that has exploded in recent years. Large companies have jumped on the bandwagon, creating analytics departments and executive positions to oversee them.
“The whole concept of extracting valuable information out of large quantities of data has really come into the limelight of everybody’s consciousness, and that’s what our expertise has been for 15 years,” Mr. Bonne said.
The big-daddy provider of corporate credit default information has long been Dunn & Bradstreet (D&B), which is known for focusing on financial ratios analysis. The company did not respond to inquiries about other, more automated approaches it may be offering or considering to evaluate corporate default or bankruptcy risk. However, it, too, is clearly seeking to take advantage of the trove of data it has collected. In June it announced the D&B Data Exchange, which offers clients access to the data housed by D&B and other companies joining the exchange, to help them identify growth opportunities.
Other players in the credit risk arena include Moody’s KMV, a long-time provider of quantitative risk-management tools, and CreditRiskMonitor (CRMZ). Both employ versions of the Merton model and, like StarMine, augment it with other types of analysis.
Mr. Rowady said Starmine’s use of data was more unique when it first emerged on the scene in the late 1990s, taking advantage of new data standards that made automated parsing of Securities and Exchange filings possible. He described the quantitative analysis behind smart estimates and other data analysis tools as essentially automating what human analysts have long done more manually and much less rapidly.
“By automating the whole process, you can compare a company’s metrics not only against its own historical data but data for its closest competitors, and so its predictive power becomes greater,” Mr. Rowady said.
Camillo Gomez, senior vice president of quantitative analysis at CRMZ, noted that the Merton model, because it’s using information about stocks, can generate a lot of fluctuation in default-frequency estimates. However, it fills in the gaps between the availability of more stable financial information, which usually is filed only once a quarter.
“Credit ratings tend to be sticky, but they are predictive of corporate defaults as well as bankruptcies,” Mr. Gomez said, adding all three components as well as the Altman Z-score are integrated into CRM’s FRISK score, which indicates the probability of bankruptcy of a company over a 12 –month horizon.
CRMZ specifically aims its services at corporate credit and procurement executives, the latter of which often use private-company suppliers. Consequently, CRMZ tracks trade payments for more than 1.5 million private and public companies. Especially for private companies, Camillo said, longer periods to make those payments can indicate bankruptcy may be on the horizon.
StarMine applies its own secret sauce, the smart estimates, whenever possible to financial ratios such as Ebit, Ebitda, cash flow, earnings, book value, net debt and earnings per share. The inputs come from sell-side analysts who typically create full models of companies’ financials out at least two years into the future. In the process they estimate numerous items on the balance sheet and income statement, and StarMine provides is smart-estimates model to those forecasts, weighing them on their historical accuracy and estimated age.
“So we’re not just relying on reported, sort of backward information, but we’re taking advantage of our smart estimates to make our assessments based on accounting ratio analysis more forward looking, timely and ultimately more accurate,” Mr. Bonne said.
StarMine is now moving beyond credit analysis and applying its smart-estimates method on a more macro level, given the focus on macro drivers in recent years, such as the weakening Chinese renminbi or the depressed European economies. Over the last month it has launched smart estimates for economists’ macroeconomic indicators as well as foreign exchange rates, and the team is currently looking at other macro trends such as asset allocation, probably out next year.