Joining me right now is Doctor Reem Hassan Bir, associate professor of accounting at Graceland University and Cairo University, whose latest research asks exactly that.
Doctor Reem, welcome back to Capital Markets.
Thank you, Basil, and thanks always for having me.
Dr.
Rehm, your newest research argues that estimating expected credit losses depends on things like customer behavior, transaction characteristics, and the reliability of forecasts.
Egypt is about to run a live experiment on exactly that.
From your research, what actually makes a credit loss estimate reliable and what makes it dangerous?
Absolutely.
When we talk about credit losses, it's really a longer scale or a bigger scale to discuss.
There's a lot of variables or indicators that the bank or the financial institution should take into consideration before giving the or extending the credit or reject the credit.
One of them, the quality of the data, the model that the Bank or the financial institution used to estimate the losses, and the credit losses have different indicators.
One of them also the assumption quality, the realistic assumption.
Some banks, they extend the credit without realistic assumptions, and some banks they are very strict with the assumption and they work under the umbrella of the central bank of Egypt or with The country.
So when we need to say what is reliable and what is dangerous, it is the data about the customer, about the financial situation, and more specifically the credit cash flow and the cash flow of the operation.
And this is because it reflects how the customer will pay back the loan.
Right, let's talk more about the phrase more data because more data sounds like better information, but does more data necessarily mean uh better credit decisions, or can it simply produce a more sophisticated looking version of the same bad decision?
Absolutely, Basil.
It's not necessarily that we have a lot of data.
It means that our decisions will be more reliable.
There is a terminology, or we used to say in information system, data overload.
When you have a lot of information, a lot, a lot of information, you are like lost between this information.
So more information is not necessarily always.
That will lead us to the perfect decisions, but the quality of the data, the quality of information can lead the decision maker or the management or the people who make the decision to extend the credit can make the correct answer.
So again, we went back to the quality of the data, the model validation, the realistic assumption can all work together.
To make the decision of extending the credit is correct, right?
So once a lender uses behavioral and cash flow data to approve credit, should that same data also flow into how it calculates expected credit losses and provisions?
And if it doesn't, what does that mismatch look like in the financial statements?
More, more focus for any financial institution before extending the credit, the focus to the cash flow and the behavioral cash flow, how the customer or the client will behave on the cash flow.
We know that there is a big difference between accrual based and cash based.
We focus on the cash flow when we check the financial statements for.
Customer or the client, we check how behavior has cash flow, how it's easily generated, the cash flow, how it moves from the minus to positive, but don't focus on the profit because profit is always misleading.
Yes, they are making a profit because of the accrual base, but in fact they are not making cash.
And when you pay your credit, you use cash.
So this is where the auditor comes in.
An auditor can expect a loan agreement or verify collateral, but how does an auditor independently challenge credit model processing millions of data points?
Of course, the auditors play a significant role in credit and auditing all this process, but it's not about millions of transactions.
Back in then, long, long, long time ago, the auditors upgraded their job.
They are not doing a full audit.
They are doing auditing by sampling, so they are, and more specifically in extending credit, they focus on the model validation, how the model is.
Developed, they don't really check the millions of transactions and millions, millions, trillions of transactions for the customer or the client or the financial institution.
They focus on how the credit model developed, how the data are validated, how the assumptions are realistic, how the data came from.
They have high quality data, so they are validating the model more than auditing the daily millions transactions to do the job of auditing.
And now to the accountability question, a very important question.
Who ultimately owns the risk if the model turns out to be wrong?
Is it management, the model developers, the risk department, and does the auditor now need a fundamentally deeper technical capability than the profession has today?
The answer is ultimately the management.
The managers are responsible at the fairest place about their decisions, about their financial statements, because the managers who are responsible for preparing financial statements, even if I ask for IT experts to develop the model, the responsibility is only about how to develop the model, but making the decision.
Using the financial statements, making the credit extensions, it's only about management and as you mentioned, auditors also play a role here, but it's not the responsibility about the decisions.
They are a role or the responsibility about auditing the model as we discussed, but also, as you said, the auditors have to.
Increase their skills to learn more about IT and technology, data analytics, all these what we use nowadays in the technology, all these technology and AI tools we have as auditors to use them, upgrade our knowledge.
Don't just take about paper-based or computer-based auditing process or technology.
And since we're kind of living the financial inclusion era, as some experts would coin it, perhaps there is a financial inclusion promise right here.
Data could open credit to businesses that banks have always turned away, but could it also create a new kind of exclusion where an entrepreneur is rejected because an algorithm reads their digital behavior badly with no loan officer to appeal to.
Of course we absolutely can say it loudly that Egypt is in the era of financial inclusion.
According to the latest statistics update from the Central Bank of Egypt, we have like 79% of the Egyptian people have active accounts in the financial institutions, different banks in Egypt, like more than 55 million of Egyptian people, they are included in the financial system.
But at the same time, yes, there is.
Inclusion and exclusion, as you said, the pass, but according to by using this technology, this technology only rely on numbers, on the data that you provide for the model, so the decision will be depending on the data that you provide.
Yes, sometimes they reject to extend the credit for some customers because the data from or the numbers tell them no, don't extend it.
Like for example, as you said, some entrepreneur.
They have a negative cash flow because maybe their operation depends on the season.
Like for example, in Egypt we have now Sahel season, so maybe during the summertime they have making positive cash flow, but in the winter they don't make cash flows.
So this instability or volatility in the cash flows tell the tell the system or the credit system not to extend the credit.
But at the same time this is always what I say we Do not rely heavily on AI technology without human judgment.
So when we use the technology and the technology results said don't extend the credit, but at the same time you have a human being, the manager or the credit manager, to say we need to extend this because there is a behind the number that we know as a human we can understand to extend this credit for the entrepreneur, right?
So the human factor is indispensable actually in this formula.
Uh, Doctor Rimaambi there, uh, it's always a great pleasure to have you with us.
Thank you very much for joining us today.
Thanks for having me, Pastor.
Thank you.
