
Media Contact
Holly Frew
Office of Communications & Marketing
Robinson College of Business
[email protected]
There is no science to business forecasting, no common set of expectations or tools. Handicapping football utilizes a more sophisticated process.
Usually the best you can hope for is to not be outrageously wrong. It's a low bar, and even then business forecasters rarely meet it. Or have you forgotten that recession that just didn't happen? At least we warned about the collapse of the housing market in 2008. Oh, right. We weren’t.
This headline from a 2022 edition of the Harvard Business Review summed it up best. "Business Forecasts Are Reliably Wrong — Yet Still Valuable." "They are exercises of imagination, which studies have shown are rarely correct in their particulars," the article stated. "Despite being reliably incorrect, savvy leaders can find strategic value from forecasts."
But how much longer will "reliably wrong" be an acceptable benchmark, especially in these volatile times where economic survival requires knowing today what will happen tomorrow?

Baozhong Yang
Professor of finance at Georgia State’s Robinson College of Business Baozhong Yang says part of the problem has been the costliness of conducting managerial surveys on such a large scale -- and doing so with the frequency necessary to adjust to market volatility.
"There's so much useful information available," said Yang. "The challenge has been organizing it all in a useful form.”
Yang, along with Robinson professors Manish Jha and Jialin Qian and Michael Weber, from Chicago Booth and NBER, tackled this challenge in their recently released paper, "Harnessing Generative AI for Economic Insights." As the title suggests, tomorrow's technology is today's solution.
The researchers utilized AI to analyze transcripts from corporate conference calls and extract information on managers' expectations for various economic indicators, like GDP, production, employment and investment.
"Our idea was to convert conference calls into managerial qualitative data," Yang said. Using available technologies, they were able to collect a wide range of economic factors from inside the firms.
From this, they've created a new measure, the "AI Economy Score," which summarizes managerial forecasts for the economy. This score reflects future GDP and economic factors including industrial production and employment.

Manish Jha
"AI-generated expectations can forecast economic trends at different levels and over long periods," according to the researchers. "Additionally, the study finds that managers' expectations have predictive power because managers have firsthand information and act on their beliefs, impacting the economy. Overall, this research demonstrates that AI can help provide valuable insights into economic trends, aiding researchers, policymakers, and investors."
Yang said the score will be useful to several sectors. Low AI Economy scores portend a challenging economic environment with adverse market conditions that will make it tough to do business. High AI Economy scores reflect optimism, surging revenue growth and positive financial performance, say the researchers.
Yang said the topics discussed in these conference calls produce significant indicators. In gloomy economic conditions, you see firms acting cautiously. Optimism is on the wane.
But when the economy is on the upswing, discussions will generally center around discussions of sales growth, improving market conditions, organic revenue growth and overall positive business performance.
The transcripts were fed into a generative AI model programmed to give answers regarding managerial expectations of the future state of the economy and the individual firms.
Data from more than 120,000 conference calls, representing 5,513 unique companies, was collected. The numbers were used to generate indicators at the national and industry-sector levels.
"We have the predictive power," Yang said. "We just haven't known how to use it effectively."
Another innovation: Separating data by industry, recognizing that some sectors may defy prevailing economic trends.
Yang foresees the AI Economic Score becoming accessible to the general public.
The ultimate goal is a better-informed customer or investor. Already existing technologies in forecasting, market analysis, policy simulation, and risk assessment can significantly aid stakeholders.
Maybe even advance from reliably wrong to reliable.