Houston voices

Focusing on data can enhance business forecasting, Houston researcher finds

A new, data-intensive technique can create a better profile of a firm and its profit forecast. Photo via Pexels

Earnings summaries are the corporate version of a Magic 8 Ball, something used to forecast future performance and profit. But Rice Business professor Brian Rountree has found that magic has its limits, and that by delving into a few additional areas of interest, investors can get a more accurate prediction of a company's future earnings than current techniques allow.

Plenty of studies analyze how to use performance summaries to calculate a firm's potential and future profits. Building on the abundant literature around this approach, Rountree, working with colleagues Andrew B. Jackson of the UNSW Australia Business School and Marlene Plumlee of the University of Utah, devised a new, additional technique for forecasting profits. By dissecting an assortment of operating details, the researchers discovered, it's possible to create a more precise forecast of a company's financial future.

Rather than replacing prior work on the subject, Rountree's team delved deeper into the significance of details within existing data. Their focus: whether including a firm's market, its overall industry and any unique activity specific to the firm makes for a more reliable profit forecast. Their conclusion: Firms can indeed improve their predictions if they separate returns on net operating assets (RNOA) into separate components and use those figures in their projections.

Normally, firms use market and industry related data to create future profit predictions. For example, a major oil company might use data on market conditions and the overall state of the oil industry to build its profits prediction. The resulting financial literature might be peppered with statements such as, "Like the rest of big oil…" or "The overall market for oil remains soft."

While this type of data is typically used to make projections, Rountree and his colleagues used the market and industry information more formally by creating the equivalent of stock return betas — a statistical measure of risk — for corporate earnings. In addition, they allowed for adding firm-specific information to market and industry information to help forecast earnings.

To conduct their study, Rountree's team used Compustat quarterly data to calculate firm, industry and market RNOAs from 1976 to 2014. Next, they broke these figures down and separated the results into different categories.

Their resulting formula differs from the conventional approach because it doesn't rely on one average set of market and industry-related data for each firm. Instead, it assumes varying factors for each company. The devil is in these details: Calculating specific market, industry and firm-idiosyncratic components improves the chances of forecasting profits correctly.

Correctly breaking down and separating profitability details to plug into the new formula is no small task. Separating company data into just three components requires up to 20 quarters of figures about prior profitability.

Once the information is processed, a researcher must then be vigilant for "noise" — incidental, irrelevant data that can lead to errors. Finally, Rountree warns, the breakdown process may not work as well for forecasting bankruptcy as it does for profits.

Used correctly, however, the technique is a practical new tool. By breaking down profitability into market, industry and firm-specific idiosyncrasies, researchers can improve forecasts strikingly compared to conventional calculations of total RNOAs.

The most accurate profit forecasts in other words, demand more than just a figurative shake of an industry Magic 8 Ball. To find the most reliable information about future earnings, a company instead has to flawlessly juggle years' worth of specific details about their particular firm. But the reward of planning based on a correct forecast can pay for itself.

------

This story originally ran on Rice Business Wisdom. It's based on research by Brian Rountree, an associate professor of accounting at Jones Graduate School of Business at Rice University.

Trending News

Building Houston

 
 

Houston-based Melax Tech has developed multiple Natural language processing tools that are used by more than 650 health care and life science organizations. Photo via MelaxTech.com

Melax Tech Partners, a leader in natural language technology processing, announced a new partnership with the University of California at Irvine that will help researchers derive insights from the UCI Health Data Science Platform’s electronic health records system and improve patient care.

Melax will implement its signature text annotation tool LANN to pull information from clinical notes, and its CLAMP product to develop natural language processing customizations through the use of AI, according to a statement from the company.

“There has been a strong desire among UCI researchers to have the capability to analyze free-text clinical narrative data using cutting-edge NLP technologies," Kai Zheng, chief research information officer at UCI Health Affairs, says in a statement. "We are delighted to have this opportunity to work with Melax Tech to deploy their AI-driven annotation and analytics tools to help our researchers advance their research agenda by leveraging the vast amount of free-text data that our health system has accumulated in the past two decades.”

Natural language processing, or NLP, allows organizations and healthcare groups to sift through and analyze massive amounts of data at a rapid rate through the use of machine learning and AI. Houston-based Melax Tech, founded in 2017, has developed multiple NLP tools that are used by more than 650 health care and life science organizations, according to its website.

In addition to the recent partnership with UC Irvine, Melax has also recently partnered with Vanderbilt University Medical Center and the University of Western Pennsylvania on similar clinical projects.

Melax has also used its platforms to pull vital information from datasets relating to COVID-19, in both medical and social settings.

In March 2022, it was awarded a Phase 1 NIH Award, valued at $300,000, to develop informatics tools based on COVID-19 datasets with the San Diego Supercomputer Center at UC San Diego. The tool aims to help researchers better understand vast amounts of virus-related data and connect findings with other similar results.

In August, Melax also received another $300,000 grant from the National Institute of Allergy and Infectious Diseases to develop NLP-based algorithms that will "model, extract and synthesize vaccine misinformation from multiple popular social media sources," according to a statement. Melax will also develop a visualization that presents its findings on the misinformation into a compressible format.

"This is a very real topic affecting culture at present," Andre Pontin, CEO at Melax Tech, says in a statement. "And shows that we as a collective business and group of experts continue to be on the cutting-edge of science in the NLP and AI domain."

Trending News