Accounting and Machine Learning: Q&A with Anna Costello

What innovations or unique elements did you introduce during the pandemic?

As we all did, I got really comfortable using Zoom. Early on, I taught my class fully remotely, so I took advantage of the technology. I introduced things like real-time surveys and breakout rooms to facilitate small group discussions. I actually fully flipped the classroom. All of my lecture material went online, and I used our time together on Zoom to engage with my students in active discussion. The trick was to leverage the technology to keep everybody engaged, because I fully understand the effects of Zoom fatigue.

I noticed that it was a little bit harder to engage with my students on a personal level. When we have a typical classroom experience, I’m able to get to know the students because there’s all of that nice, informal discussion that happens before class when students are coming in, during breaks, or even after class. I had really gotten to know my students when we were interacting with each other in person. To try to get to know my students better while classes were virtual, I introduced Zoom coffee hours with smaller groups. That enabled me to get to know them on a more personal level outside of the classroom.

How can I train myself to be more sensitive to numbers?

I assume this question is about familiarizing yourself with financial-statement information. And so, the first thing I recommend that you do is seek out formal training by taking some of the great accounting courses offered here at Booth. There are some fantastic professors that will give you a really nice basis for understanding how to read financial statements and financial reporting more broadly.

The second thing I recommend is to learn by doing. In other words, pick up the financial statements of some of the companies that you’re interested in and browse through them. There’s actually a lot of interesting information in the narrative portion of financial statements, and they’re quite easy to read, so you’ll start to understand the numbers.

By reading multiple financial statements, you’ll also pick up some of the important jargon that’s used in the financial-reporting world. You’ll start to get some context for being able to understand actual financial statements like the income statement, the balance sheet, and the statement of cash flows.

How could machine learning and natural language processing contribute to empirical accounting research?

This is a great question. Machine learning and natural language processing are starting to play a larger role in accounting research. For one, they’re impacting the types of questions that we ask. For example, I have a paper that investigates the role of machine learning in the fintech lending space. Others have also looked at similar questions, such as the role of algorithms in trading strategies and in making other important corporate decisions.

Second, machine learning and natural language processing are now often used as tools to conduct research. Corporate financial reports are long. They’re dense. They often contain nonfinancial and nonstructured data. So machine learning and natural language processing are great tools that researchers can use to better understand the details and important insights from these reports. For example, machine learning can be used to classify reports based on what we would view as the sentiment of the manager writing the report.

It’s important that we use these tools, because we can gain even broader insights into what managers are thinking and what they’re projecting for the future of their companies.