IFRS 9 ECL Walkthrough | EXCEL PROTOTYPE

So good morning guys, I am taking this session to walk through the IFRS 9 implementation framework and help banks understand the ECL calculation method. To demonstrate how ECL calculation is done, I am showing you an Excel-based prototype that you might be seeing on the screen. If, as a bank, you want such a tool for your bank, then you can definitely come in contact with us.

Now, what example have I taken here? I have taken a loan example. So, the bank had given a loan, a mortgage loan, to a customer one year ago for one lakh rupees, right? And we all know that mortgage loans are collateral, meaning you also take collateral against them. So, you suppose you had taken collateral for 60,000. This 75,000, I will tell you. You suppose one year ago you had taken collateral for 60,000 and you had given a loan of one lakh at an interest rate of 12 percent, right? At 12 percent, you had given a loan to a customer one year ago for one lakh against a collateral of 60,000, and the loan duration was 120 months. Say 120 months, or 10 years, right? Or say 40 quarters. 120 months, 10 years, 40 quarters.

Now, when you would have given this loan to this customer, you would have obviously evaluated this customer's creditworthiness. To evaluate the customer's creditworthiness, you calculate the customer's credit score, or you assign a rating grade to a customer, right? So, every bank has a master rating scale on the basis of which you assign a rating grade to the customer, like you can see here, Grade 1 to Grade 9. One is your best grade, and nine is your worst grade. So, at the time of origination, you had assigned this customer the grade CRR 2, Credit Risk Rating Grade 2, right? Okay.

Now, one year has passed. In one year, this customer has made all their payments on time, right? In one year, this customer has made all payments on time, and the loan is active today. This loan is active today. Can this loan cause the customer to default, leading to future losses for the bank? The answer is yes. Can this customer cause future losses for the bank? The answer is yes. For those losses, the bank has to make provisions, which is Provision for Doubtful Debt. Provision for Doubtful Debt. If this customer defaults tomorrow, right? No doubt, in the last 12 months, all EMIs have been paid on time, but if this customer defaults tomorrow, then in that case, the bank can incur losses, and for those losses, you should set aside provisions under Provision for Doubtful Debt.

Now, let me tell you a little about provisions. Earlier, what was the requirement in banks? Earlier, banks had an incurred framework. What does the incurred loss framework mean? It means that if a customer became bad, then we would set aside provisions for them. So, that was a reactive approach. It was such a bad approach that we were saying, someone became bad, and then set aside provisions for them. That was nonsensical, right?

IFRS 9 came, or rather, IFRS 9 came, and it talked about the Expected Credit Loss framework. Incurred Loss versus Expected Credit Loss. What was said in the ECL framework? Estimate how much losses may occur due to future defaults, and set aside provisions for them today, right? So, it's a proactive approach. What is its benefit? If I don't set aside provisions and enjoy all the profits coming in, or distribute them as dividends, then tomorrow, if there are losses, will I have sufficient reserves to bear the losses? No, if I don't maintain provisions. Therefore, it was said that estimate the future potential losses, even though this customer is healthy today, still estimate future losses, and based on those losses, set aside provisions today so that you don't face problems tomorrow. So, it's a proactive framework.

Now, how do we model these losses? How do we calculate these losses? Will this customer default tomorrow? I don't know. If this customer defaults, what are their chances of default? Do I know that? The answer is no. If this customer defaults tomorrow, then in that case, you will sell their collateral. Some amount will be recovered from the collateral, the remaining amount will not be recovered. The remaining amount will be your loss. So, how much amount will be lost? Is that known? The answer is no. And the biggest thing, when will this customer default? Do you know that? The answer is no. So, what are the chances of this customer's default? That is unknown. If they default, how much will be recovered, how much will be lost? That is also unknown, right? And third, when will they default? That is also unknown. So, in that case, how do we estimate losses?

So, what approach do we use? Think about it. The total duration was 40 quarters. One year has passed, so there are still 36 quarters remaining, right? So, I will assume that this customer will default either in the upcoming first quarter, or in the upcoming second quarter, or in the third quarter, or in the fifth quarter, or in the 10th quarter, or in the 20th quarter, and so on. So, I have 36 quarters. I will assume that this customer can default in any one quarter, either in the first quarter, or the second quarter, or the third quarter, or the fourth quarter, and so on, up to the 36th quarter. And whenever they default, right? If they default in the first quarter, then how much will be my losses? I need to estimate that. If they default in the second quarter, then how much will be the losses? I need to estimate that. If they default in the third quarter, then how much will be the losses? I need to estimate that.

Now, how are losses estimated? Basically, for that, you need to model three components. Which three components? Probability of Default, Loss Given Default, Exposure at Default. Probability of Default, Loss Given Default, Exposure at Default. And if you model these three components, then your expected loss for that quarter is simply Probability of Default multiplied by Loss Given Default multiplied by Exposure at Default. You need to model these three components separately for each quarter. Out of these, one component is easily known to you, which is Exposure at Default, because you have the loan amortization schedule. So, if I ask you, if this customer defaults in the upcoming quarter, right? Then how much amount is outstanding? Exposure at Default is, at the time of default, how much amount is outstanding, right? So, will you know that? Yes, by looking at the amortization schedule, you can tell me how much amount will be outstanding.

Here, on the left side, you can see the amortization schedule, month-wise, because our EMI is monthly. So, let me take the plot over here. As you can see, we have the amortization schedule. I have the amortization schedule here for 120 months, that is, 10 years. Outstanding balance at the beginning, your EMI, how much it is, that will remain the same. How much is the principal portion? How much is the interest portion? Outstanding balance at the end, how much it is.

Now, if this customer defaults in the upcoming first quarter, right? Then how much amount will be outstanding? Can I say the balance at the beginning of the 13th month, because 12 months have already passed, I am talking about default in the upcoming first quarter? So, the balance at the beginning of the 13th month plus three months of interest will be my Exposure at Default. If they default in the upcoming quarter, then what will be my Exposure at Default? As per the amortization schedule, I can see what the balance is at the beginning of the 13th month, right? Or say, what is the balance at the end of the 12th month, and I will add three months of interest to it. Why three months of interest? Because in banks, generally, the definition of default is 90 days past due. 90 days past due. So, if this customer stops paying from today, then they will not pay for three months, only then will they be called a defaulter, right? In the first quarter. So, if they don't pay for three months, then the outstanding amount you had will have three months of interest added to it, right?

Similarly, if I need to calculate the Exposure at Default for the second quarter, meaning if they default in the second quarter, then how much will be the outstanding amount? Then can I say the balance at the beginning of the 16th month plus three months of interest? Meaning, if they default in the second quarter, then they have paid on time in the first quarter, so your balance will reduce, right? The balance will reduce to 93,000. Earlier your balance was 9446. Now it is 93,000. To that, you will add three months of interest, so you will get the Exposure at Default. So, like this, can I calculate the Exposure at Default for every quarter? Can I calculate the Exposure at Default for every quarter? The answer is yes.

So, here on the right side, what have I done? As you can see, in column J, we have quarters, and in column L, I have calculated the Exposure at Default. How did I calculate it? The balance at the beginning of that quarter plus three months of interest. So, one component is modeled for me for every quarter, how much amount is outstanding. And obviously, this amount is decreasing, because if they default in the first quarter, then in that case, your amount due is more. If they default in the second quarter, then your balance has decreased because you have also made some payments, right? You have paid three EMIs, so your balance has decreased. So, you have the Exposure at Default. What are the remaining two components? Probability of Default and Loss Given Default.

Now, I evaluated the creditworthiness of this customer at the time of origination, and I assigned this customer a rating grade, CRR 2. Today, tell me, should I update this customer's rating grade after one year, or should I keep the same rating grade? I should update the rating grade based on new information. What new information would I have received about this customer? At the time of origination, I would have received application data for this customer. But when one year has passed, in one year, how much has this customer paid on time, how much have they delayed payments? All this data is also available. That is called behavioral information. How has this customer behaved in the last one year in terms of payments and in terms of delays? By looking at that behavioral information, I can update this customer's rating grade.

So, you can see a column called Current Grade. The current grade is, suppose, CRR 3, right? This customer has made some delays, or whatever they have done, because of which their grade has become CRR 3. Now, within the bank, for every grade, you get an average PD, right? For every grade, you will get an average PD within the bank. Banks calculate it like this under the Master Rating Scale. There is a master rating scale system in which we see that for every rating grade, on an average, over a longer run, what is the Probability of Default? How is Probability of Default calculated? For example, you see that in 2014, you had 100 loans, and at the end of 2014, two loans defaulted. So, what was your PD for 2014? It was 2%. Similarly, you calculated for 2015, 2016, 2017, 2018, 19, 20, 21, 22, and you are taking the average of all those PDs. So, you get a long-run average, that is, what is the average Probability of Default for loans in that rating grade? How many loans go into your defaulting grade? So, you get an average PD for every rating grade. That is called Through-the-Cycle PD. That is called Through-the-Cycle PD.

Can I use this Through-the-Cycle PD in IFRS 9, in ECL? That is question number one. Second, I will have the collateral value of this customer today. One year ago, this customer gave collateral of 60,000. The bank valued it today. Today its valuation is 75,000, right? And if the customer defaults today, then how much will be your outstanding amount? 97,321. If they default in the upcoming first quarter, then your outstanding amount is 97,321, and how much collateral is there in front? 75,000. This means if they default, assuming the collateral is sold at full value, then the loss will be 22,000. A loss of 22,000 on 97,321. Can I calculate Loss Given Default based on this? That if there is a loss today, then 75,000, I assume, will be fully recovered. The remaining 22,000 will not be recovered. So, my Loss Given Default will be 22,000 divided by 97,000. Can I use the collateral value of today and use the same Loss Given Default at different times? That is question number two.

So, what have I done? I have the updated rating grade of this borrower today. Against that, I have an average PD. Can I use that average PD for every time? And second, using the collateral value of today, I can estimate the Loss Given Default. And can I use that Loss Given Default for every time, for every quarter? The answer to both questions is no. IFRS 9 requires you to estimate forward-looking losses and set aside provisions based on them. What does forward-looking losses mean? It means banks need to estimate what macro-economic scenarios can occur in the future, and in different macro-economic scenarios, you need to see how much your losses will be at different points in time, and based on that, the provision you get, you need to keep it under the ECL framework, under P&L, right?

So, this means now I also need macro-economic scenarios, that is, how the economy will be at different times. Can you easily estimate that? That is also difficult. Think about it, you need to generate macro-economic scenarios for the next 39 quarters. That is very difficult. So, banks have a macro-economic analysis and forecast team, or as a consultant, we can help you to generate scenarios. So, here what have I done? I have generated three macro-economic scenarios: Base Scenario, Upward Scenario, Downturn Scenario, right? You can call it Base Case, Best Case, Worst Case, whatever you like, Pessimistic, Optimistic. You need to take all macro-economic scenarios, meaning you need to take a Base Scenario, that if the economy remains stable, it will be like this. If the economy booms, it will be like this. If the economy goes into recession, it will be like this, right?

In each scenario, we compute a macro-economic index, which is given by G values. Macro-economic index. Here you can see the G values, okay? That is your macro-economic index at different points in time. If this Z value is positive, it means your macro-economic index is good, right? In the upward scenario, you can see your G value is positive, meaning your macro-economic scenario is good. In the downturn scenario, you can see Z is taking negative values, meaning your economy is going to be bad. In the downturn scenario.

Now, this macro-economic index that we have taken, through G, we are taking a composition of many macro-economic variables. I have taken two macro-economic variables: one is the unemployment rate, and the second is the housing price index. You can take more macro-economic variables. So, basically, what are we doing? We are taking historical data of the unemployment rate and the housing price index. Here you can see the dark line, it has the historical data of the unemployment rate and the historical data of HPI, right? Based on that historical data, I am generating forecasts of how the unemployment rate will be in the coming future, right? Generating forecasts of how the unemployment rate will be in the coming future. You can see this is your downturn scenario, where the unemployment rate will be high, that's why I have kept it in red. This is your upward scenario, where the unemployment rate will fall, and there is a base scenario, a stable scenario. Similarly, in HPI, if your housing price index grows well, that is the upward scenario in green. This is your downturn scenario. This is your base scenario.

If you get forecasts for all these macro-economic variables, then you can forecast the macro-economic index. I have forecast that macro-economic index for each time in the future for all three scenarios: Base Scenario, Upward Scenario, Downturn Scenario. Now, based on this, what will I do? I told you that for every rating grade, I get an average PD, that is called Through-the-Cycle PD. I will adjust that PD and make it Point-in-Time, forward-looking. The technical term is Point-in-Time Forward-Looking PD. Meaning, I am looking at forward-looking macro-economic scenarios, and accordingly, I am adjusting my PD, my average PD. So, if tomorrow the macro-economic conditions are good, then my Point-in-Time PD will be lower than the Through-the-Cycle PD. If tomorrow my macro-economic conditions are good, then my Point-in-Time PD will be lower than the Through-the-Cycle PD. And if tomorrow my macro-economic conditions are bad, then my Point-in-Time PD will be higher than the Through-the-Cycle PD. And if tomorrow my macro-economic conditions remain around the base scenario, then my Point-in-Time PD will be close to the Through-the-Cycle PD only, right?

So, what are we doing? We are again assigning a rating grade to the borrower. Against that rating grade, I have an average PD, Through-the-Cycle PD. But I cannot use this average PD because IFRS 9 requires that you estimate Point-in-Time forward-looking losses. So, I am generating different economic scenarios. In those different scenarios, I am generating a value of the macro-economic index, which is given by G values. If the G value is positive, then the index is good. If the G value is negative, then the index is bad, right? In the base scenario, my G value is around neutral, which is zero. If tomorrow the macro-economic index is bad, G values are negative, then your PDs will be adjusted upwards, meaning your Point-in-Time PD will be higher than the Through-the-Cycle PD. If tomorrow your macro-economic index is positive, meaning macro-economic conditions are going to be good, then your Point-in-Time PDs will be lower than the Through-the-Cycle PD. So, we need to make that adjustment. I have made that adjustment here. How did I do it? Its methodology. I have used it, which gives me what? It gives me Point-in-Time PDs, right? Point-in-Time PDs.

So, here, if you see, you have the Point-in-Time PD for the first year, and this is your Through-the-Cycle PD. If my economic scenario is more of a downturn, more of a downturn, then you can see what your Point-in-Time PDs are: higher. Here it is 10 percent, here 12 percent, here 7 percent, here 8.2 percent. If tomorrow your macro-economic conditions are much better, then your Point-in-Time PDs are lower. This is 10 percent, this is 8 percent, this is 7 percent, this is 5 percent. So, basically, we are adjusting our Through-the-Cycle.

We are adjusting the PD based on the macroeconomic conditions. We are not going into the method of how we are doing it, as there are many methods. I am just demonstrating the process.

Secondly, I have the value of collateral today. Right? Apart from that, I have the growth rate of how the housing price index is going to grow. Can I assume that the way the entire market will grow, my collateral will also grow in the same way? Meaning, the collateral that has been received against this loan, will that collateral also grow in the same way?

So, what do I need to do now? Basically, I will grow the collateral at different times at different growth rates. For example, in the base scenario, how the index will grow at different times, that growth rate is already estimated by me. How are these estimates made? You can use some time series models to get this growth rate. I have estimated that growth rate under the base scenario. Using that growth rate, can I value my collateral at different times? Obviously, my growth rate is positive, so my collateral will also grow. And naturally, in the base scenario, your collateral grows because property prices always increase. So, I will calculate the value of the collateral, right? With a certain growth rate, how my collateral will grow over time.

So here, I have simply taken the initial collateral value into 1 plus the growth rate to calculate the collateral value at the end of each period. Secondly, I also have the exposure at default. We have calculated this: if there is a default in the first quarter, what will be your exposure at default? If there is a default in the second quarter, what will be the EAD? If there is a default in the third quarter, what will be the EAD?

Let me calculate a ratio using the value of this collateral and this exposure, which is the loan-to-value ratio. And then, based on this loan-to-value ratio, I can compute the loss given default. Think of the simplest scenario: if my loan-to-value ratio is 100%, meaning there is ₹1 of loan outstanding and ₹100 of collateral, then if I don't assume any costs, what will be the bank's loss given default? Zero. Because ₹100 is the outstanding amount, right? And ₹100 is the value of your collateral. So, loss given default is zero. This is correct if I assume there are no costs. In reality, banks also have many types of costs, recovery costs, collection costs.

Apart from that, if the customer defaults today, can you immediately sell their collateral? No. You will go to court, the court will pass a judgment, then you will sell the collateral. So, there is a time lag involved. It takes two to three years to sell the collateral. So, there is also an opportunity cost for the bank, a loss of time value of money. So, actually, if your loan-to-value ratio is 100%, then your loss given default is not zero. Banks will estimate at what loan-to-value ratio their loss given default is zero. For example, 70%. So, if there is ₹100 of collateral and ₹70 of loan, then suppose the bank incurs no loss. Right? So, my loss given default is zero if my LTV is equal to 70%. Because if I sell ₹100 of collateral, recover the collection cost, sell it over time, it won't sell immediately. Considering all those things, I think net net, I will get ₹70 in hand, and the loan is also ₹70. So, my loss given default will be zero.

So, by looking at the loan-to-value ratio, I can estimate the loss given default. How? I can say my loss given default is simply my loan-to-value ratio minus 70%. How did this 70% come about? The bank would have internally estimated that if the loan-to-value ratio is 70%, then in that case, against ₹100 of outstanding, there is ₹70 of collateral. Right? ₹70 loan, ₹100 collateral. Then the bank incurs no loss because all costs are also recovered, and the recovery obtained over time with the time lag also recovers the time value of money. So, ₹30 is for compensating for all those costs and time value of money. So, I am simply saying, whatever your loan-to-value ratio is, minus 70% is your loss given default.

So, tomorrow, if your loan-to-value ratio is 80%, meaning ₹100 collateral and ₹80 loan outstanding, then your loss will be 10%. Tomorrow, if your loan-to-value ratio is 100%, then in that case, your loss will be 30%. Right? So, in this way, a formula has been created: LGD = X - 70%, where X is the loan-to-value ratio.

Now, how to estimate the loan-to-value ratio? You have the outstanding amount, how much loan amount is outstanding if there is a default, that you have. You have computed the exposure at default. Do you have the value of collateral? Yes, you also have the value of collateral. You have calculated the collateral value at the end of each period by assuming a growth rate. But will your collateral sell at this price? No. Whenever the bank goes to sell the collateral in the market, because real estate is illiquid, it will sell at a slightly lower price. There will be some haircut applied to it. I have assumed that whatever the value of my collateral is, it will sell at 20% lower. Right? Whatever the value of my collateral is, it will sell at 20% lower than that. So, 20% is my haircut. Now, whether it is a 20% haircut or 30%, that banks do internally, but there will be some haircuts because you cannot sell the collateral immediately. And if you try to sell it immediately, it will sell at bad prices because real estate is illiquid. Right?

So, here I have calculated the selling price against each collateral value. What will be my actual selling price? And my selling price will be 20% lower, that is, I will be able to recover only 80% of the collateral value. Right? So, I have the collateral recoveries, how much collateral I will be able to recover. And I have the exposure at default. Can I calculate the LTV? Loan-to-value ratio? Yes, I can calculate the loan-to-value ratio as my outstanding loan amount divided by collateral recoveries. And then, by looking at the LTV, I can calculate the LGD. I said that the bank has internally estimated that if the LTV is 70%, then there is no loss. Because if the LTV is 70%, then by selling ₹100 of collateral, I recover enough that my entire ₹70 is recovered. So, loss given default is zero. In that case, I am using this simple formula. I have to compute LTV - 70% to get my LGD.

But tomorrow, if any customer defaults, will all of them go into recoveries? No. Some customers will also cure. Some customers will cure, meaning they will pay back again, they will improve again. So, if my cure rate is 25%, then only in 75% of the cases will this loss given default matter. Because in 25% of the cases, I will recover the full amount, as the account has cured, they will pay back your money. So, for this reason, I have also multiplied by 75. So, ultimately, what is my LGD formula? 75% (because 25% of accounts will be cured) multiplied by the maximum of zero, comma, loan-to-value ratio - 70%.

For the loan-to-value ratio, what do you need? You need the outstanding amount, how much is outstanding if there is a default. Second, you need how much you can recover from the collateral. So, you have taken that here. Based on that, you have the LGD for each time in the future. You can see that your LGD is higher initially, then your LGD is falling. Why is LGD falling? One is that your collateral is growing, property prices will increase. Second, your outstanding amount is also decreasing, as per the amortization schedule. If you default very late, then you would have paid many EMIs. So, as per your amortization schedule, the exposure at default is also decreasing, and your collateral recovery is also increasing. So, in that case, my LTV ratio is going down, and since LTV is going down, my LGD is coming towards zero.

So, in this way, in each scenario, can I calculate the loss given default for each time in the future? Yes. So, what I have done is, I have taken the macroeconomic index and estimated the PD at each time. Here, I have done it quarterly. Well, why are we doing quarterly modeling? Because within ECL, quarterly is the best time frame because the macroeconomic data you have is available quarterly. So, we prefer quarterly modeling rather than monthly modeling. Right?

Here, what I have done is, as per the macroeconomic index, I am estimating the PD for each quarter. For each quarter in the future. Right? As per the macroeconomic index of that time. As per the macroeconomic index of that time point. Meaning, in the first quarter, what will be the macroeconomic index, based on that, the PD. In the second quarter, what will be the macroeconomic index, based on that, the PD. Right? Under the base scenario. And in the base scenario, your macroeconomic index is generally neutral, around zero. Then we have the upward scenario, where your macroeconomic index is positive. There, the PDs that will come will be lower. Your LGDs will also be comparatively lower. And then we have the downturn scenario, where your macroeconomic index will be negative. Based on that, the PDs you calculate, the adjusted PD point in time, will be higher. And the LGDs you calculate will also be higher.

So, what you need to do is, you need to adjust the PDs at different times in different scenarios by taking the macroeconomic index, and you need to model the LGD. Right? Once I do that, then I come to the front sheet. In the front sheet, what I have done is, I have brought all those values here. As you can see, I have LGD base scenario for all 36 quarters. I have LGD upward scenario. I have LGD downward scenario. Similarly, I have PD base scenario, PD upward scenario, and PD downturn scenario. So, as you can see, in the downturn scenario, my PDs are higher. In the upward scenario, my PDs are lower. Right? Similarly, in the downturn scenario, my LGDs are higher. In the upward scenario, my LGDs are lower. Here, my LGD in the upward scenario is coming at 24%. Here, it is coming at 38%. And I have the exposure at default for each quarter.

So, I have the exposure at default for each quarter. I have the LGD in all three scenarios. I have the PD in all three scenarios. Can I calculate the ECL? Expected Credit Loss. Expected Credit Loss is simply exposure at default into probability of default into loss given default. EAD into LGD into PD for the base scenario. Then EAD into LGD into PD for the upward scenario. Then EAD into LGD into PD for the downturn scenario. Right?

Now, the losses that will come, they are standing at different times in the future. I need to make provisions today. But these losses are standing in the future. So, you need to take the present value of these losses. Right? You need to estimate these losses by discounting them into today's terms as well. For that, what do we do? We take the present value of the losses. To take the present value of the losses, I need a discounting factor. To take the discounting factor, I need an interest rate. Should I discount at a 12% interest rate? No. Under IFRS 9, there is a requirement that you compute the effective interest rate, that is, what interest rate a bank effectively earns. You charge 12% from the customer, give a loan of ₹1 lakh, but you would have also charged some fees, processing fees. So, effectively, you don't charge 12% from the customer. You charge some higher amount. Suppose that effective interest rate came to 12.2%. Or how can you calculate it? In any loan, there are cash flows involved. By computing the IRR of those cash flows, by computing the internal rate of return, you can compute this EIR. That came to 12.2%. So, using that EIR, I have calculated the discounting factors here. Using the effective interest rate, I have calculated the discounting factors. Using those discounting factors, I will take the present value of my losses.

So, I have the expected credit loss for each point in the future in all three scenarios: ECL base, ECL up, ECL down. Right? If I add these up, then I will have the total losses in the base scenario. How much is coming? 4616. Meaning, today, what was the outstanding amount of this customer? Today, the outstanding amount of this customer was 94,000, 94,486. Against that, how much provision do I have to keep? 4461. In the upward scenario, how much provision will I have to keep? 4.66. And in the downward scenario, how much provision will I have to keep? 13182. Is it logical? The answer is yes.

But these are the expected credit losses in your different scenarios. If you need to get a single number, then in that case, I need to provide some weightage to each of the three scenarios. What is the weight of the base scenario? What is the weight of the upward scenario? What is the weight of the downward scenario? Again, the macroeconomic analysis and forecasting team in the bank, or as a consultant, we can help you to identify these weights. If you have the weights for different scenarios, then by simply multiplying this ECL with this weightage, you can find out the weighted ECL.

So, here I have calculated that weighted ECL. And if I add the weighted ECL, then my ECL will be 469. 469. Meaning, after considering these weights for this customer, who has an outstanding amount of around ₹94,000 today, I need to keep a provision of 469. This much provision I will have to keep. Around 5% or 4.8%, whatever it comes to, you can calculate it. Here, equal to this divided by this outstanding amount, 4.93%. This much provision you will have to keep for this customer.

So, what we have done is, we have taken a customer who originated a year ago. Today, they have a certain outstanding amount. In the future, the customer can default, due to which the bank can incur losses. You need to estimate those losses. For that, you need to set aside a provision. Now, what are the chances of this customer defaulting? You don't know. If this customer defaults, how much amount will be lost? How much will be recovered? You don't know. When will this customer default? You don't know. So, we have assumed that this customer can default in the coming first quarter, or in the second quarter, or in the third quarter, or in the fourth quarter. They can default anytime in these 36 quarters. If they default, what will be the outstanding amount? The balance at the beginning of that quarter plus three months of interest. So, I will have the exposure at default for each quarter. That is, if they default in the first quarter, what is the outstanding amount? If they default in the second quarter, what is the outstanding amount?

Now, what do I need? I also need the probability of default for that customer and the loss given default. So, I had updated the rating of this customer. Based on that updated rating, I have the average PD for that customer. But I cannot use the average through-the-cycle PD because IFRS 9 requires that you compute your point-in-time PD and point-in-time losses by considering forward-looking macroeconomic scenarios. So, in that case, I have taken different macroeconomic scenarios: base scenario, upward scenario, downturn scenario. In each scenario, I am considering the macroeconomic index, which is given by the G value. If the G value is neutral, it means my macroeconomic scenario is base. If the G values are positive, it means the macroeconomic index is good, my macroeconomic scenario is upward, upturn. Right? And if my G values are negative, it means the economy is going to worsen. Right? My macroeconomic index is bad.

In different scenarios, whatever the value of your macroeconomic index will be, for each time in the future. Right? For each time in the future, based on that, you will adjust your average PD to compute something called as point-in-time PD. So, here I have computed that point-in-time PD for different times.

Well, these scenarios have a nature. If you look carefully, what is the nature? First, your unemployment rate will increase. Right? For example, in the downturn scenario, first your unemployment rate will increase, then it will come to normal. Similarly, in the upward scenario, first your unemployment rate will fall, it will come to normal. So, in the longer run, it will come back to around the base. That nature remains in your scenario. Right? I have designed the scenarios in that way. As you can see in the plot.

So, based on whatever the value of my macroeconomic index is, I have adjusted my average PD to compute point-in-time PD. If the macroeconomic index is bad, then my point-in-time PDs will be higher. If the macroeconomic index is good, then my point-in-time PDs will be lower. And if the macroeconomic index is neutral, then my point-in-time PDs will be close to the average PD.

Similarly, I also need LGD at each time. For that, I have the collateral value today, ₹75,000. But this collateral will grow with time. Right? It will grow or degrow, that will depend on the scenario. But you also need to consider the growth rate of the collateral. So, in different scenarios, you need to take the growth rate of the collateral. I am taking that as equal to the housing price index growth rate. So, the way the housing price index will grow, in the same way, I am assuming my collateral will also grow. Right? And I am valuing the collateral. But your collateral will never sell at this price. Your collateral will always sell at lower prices. So, you have to apply some haircuts. So, I have applied a 20% haircut because of the lack of liquidity, your collateral will recover less money compared to the value of the collateral. Right?

Okay, you have the recovered amount from the collateral. You also have a schedule of the outstanding amount at each time, which we have already calculated. Using both, I can calculate the loan-to-value ratio. Now, banks would have internally estimated that at a loan-to-value ratio of 70%, the bank incurs no loss. Because if the loan-to-value ratio is 70%, then in the 30% margin you have, all costs and all time value of money are recovered. Right?

So, what we are doing is, we will compare this loan-to-value ratio with 70%. The higher the loan-to-value ratio, the higher your loss given default. Right? But obviously, if a customer defaults, not all customers go into your recoveries. Some customers also cure, they improve. So, in that case, if 25% is my cure rate, then for 75% of customers, this loss will occur. So, in this way, the function is designed. We are also...

Many models can be used. The thing is that you have to take the collateral value at different points in time. You have to somehow incorporate the forward-looking scenario into LGD. So, I am taking the collateral value at different points in time. Based on that, I am calculating this LGD. And obviously, as time passes, your LGD will go towards zero because the collateral value is also increasing, and your exposures are also decreasing, right? So, I have PDs at different times, and I have LGDs at different times in different scenarios. If I multiply PD by LGD by D, I will get the expected credit loss for all the three scenarios, right? If I add that expected credit loss, you will have the provisions. How much provision you should keep in different scenarios, right? But if you want to get a single number, which you have to, you have to provide some weightage to these scenarios. So, if I have used certain weights now, my lifetime ECL is coming to 4657. How much is it? 4665. If I give more weight to the downturn scenario, obviously you will have to give weight to the rational scenario. If I give more weight to the downturn scenario, then my lifetime ECL is coming to 8644. If I reduce the probability of the downturn scenario, then my lifetime ECL is also reducing accordingly. Now, you might have found something strange here. What is it? That we said, estimate all the losses that can occur from a customer in the future, over the entire residual lifetime of the loan, and based on that, keep provisions. Isn't that approach too conservative? I am saying, estimate any loss that can happen in a lifetime, and based on how much loss can happen over the entire lifetime, set aside provisions today. Isn't that too conservative? The answer is yes. I will only take its one-year income, but a loss can happen anytime in the future, and if I estimate how much that loss can be and set it aside today, my profit and loss account will be very negatively impacted, right? So, IFRS 9 said that yes, this approach is too conservative. So, one resolution for that is that we will categorize all our accounts into different stages. So, you have to categorize every account into either Stage One or Stage Two or Stage Three. All the accounts in the bank have to be categorized into Stage One, Stage Two, or Stage Three. Your Stage One accounts are your healthy accounts. For those healthy accounts, you do not need to adopt a conservative approach. You do not need to keep provisions for lifetime ECL. You should calculate only one-year ECL, meaning, add the ECL of only the first four quarters and keep provisions only that much. For example, if I add the ECL of the first four quarters, my weighted ECL is only 232. So, if an account is a healthy account, you do not need to be that conservative. If it is a Stage One account, you will keep provisions against one-year ECL only. But if an account is your Stage Two account, meaning an account that has become unhealthy, that has deteriorated, then you have to be conservative for that account, and you will have to provide for lifetime ECL within your P&L. And if an account has defaulted, a defaulted account means a Stage Three account. It has already defaulted, so you have to estimate ECL considering a probability of default of 100%, because it has already defaulted, then the PD will be exactly 100%. There is no question of different scenarios here. What we will do here is just showing you what you need to do. First, you need to assess which stage the customer is in. If they are in Stage One, then you have to use their one-year ECL. If your customer has moved to Stage Two, then in that case, you have to keep provisions against lifetime ECL for them because the customer is unhealthy now. So, you have to be conservative, and you have to provide for lifetime ECL. Okay? How much is it coming to? 11,000. And if your customer has defaulted, then they will directly go to Stage Three. Then you have to estimate losses by taking PD directly as 100%. If you take PD 100%, the losses will be very large. So, how much provision do you need to keep? 6575, which is a very big number, right? So, you have to categorize every account into Stage One, Stage Two, or Stage Three. Now, how is this categorization done? All those accounts that are newly originated come into Stage One. Meaning, accounts that have recently originated are put into Stage One. And second, those accounts where credit risk is very low. Credit risk is very low. For example, when this account originated a year ago, it was newly originated then, so it must have gone into Stage One then. Today, after one year, I will check the one-year PD of this account. What is the one-year PD? So, if this account's rating grade is CRR 3, then the one-year PD is 0.522%. Suppose the threshold in the bank is what is the threshold? There is a threshold of 3% that if your PD is below 3%, then the credit risk is very low. Put it directly into Stage One. So, is its PD below 3%? No. So, we will not put it into Stage One. For now, I am saying, accounts that are newly originated come into Stage One. Is this account newly originated? No. When it was newly originated a year ago, you would have put it in Stage One. But today, this account has to be assessed: is there very low credit risk in this account? If there is low credit risk, then put it into Stage One without thinking. For low, there will be an internal benchmark. For example, 3%. Now, this borrower's one-year PD is coming to 0.522%. My low benchmark is 3%. So, as per this criterion, should it be put into Stage One or not? So, it is not put into Stage One now. For example, what is the criterion for Stage Two? Stage Two is called Significant Increase in Credit Risk. Those accounts where there is a significant increase in credit risk. Those accounts where there is a significant increase in credit risk should be put into Stage Two. Now, how do we assess a significant increase in credit risk? For that, the regulator says, check what your PD was at the time of origination and what your PD is today. If your PD has relatively increased, for example, by more than double. Now, whether to take double or triple, you estimate that internally. But if your PD has increased a lot compared to the time of origination, if I consider a maturity of nine years, what would the PD be? What is your PD today? If that PD has increased by more than double, then we will consider it a significant increase in credit risk. Has your PD increased by more than double here? No. It has gone from 16% to 23%. So, we will not consider this a significant increase in credit risk. What could be another criterion? If your rating grade has gone below a certain notch. Right? For example, if the rating grade goes to six or below, then we will consider it a significant increase in credit risk. Has my rating grade gone below this certain rating grade? Has it gone below this threshold? The answer is no. So, according to that, it has not gone into Stage Two. So, it means there is no significant increase in credit risk. So, if there is no significant increase in credit risk, then it will remain in Stage One, right? And if your rating grade becomes nine, meaning it has already defaulted, then it will directly go into Stage Three. Again, I am saying, you have to find the current rating grade of this borrower. You would have re-evaluated this borrower based on behavioral information and updated the rating. Now, I have to categorize this account into either Stage One or Stage Two or Stage Three. So, how will I categorize? I will say all newly originated accounts go into Stage One. So, this is not newly originated. Then you have to check if there is very low credit risk in your account. Is your credit risk very low? If your credit risk is very low, then put it into Stage One without thinking. I have taken the low benchmark as, say, 3%. But for this rating grade, the one-year PD is coming to 0.522%. Is it below the threshold? Below the low threshold? The answer is no. According to that, it will not go into Stage One. If it does not go into Stage One, then I have to check if there is a significant increase in credit risk for this borrower. For a significant increase in credit risk, you will have to establish some internal guidelines. For example, what I did is, I looked at what the PD was at the time of origination over the entire lifetime and what the PD is today over the entire lifetime. If that PD has increased by more than double, then I will consider it a significant increase in credit risk. Or if the rating grade has gone to six or below, then I will consider it a significant increase in credit risk. So, if either of these criteria is met, then you will consider it a significant increase in credit risk. If neither of these criteria is met, then in that case, it will remain in Stage One, right? And if your rating grade, obviously, you have gone into the defaulted grade, then you will directly go into Stage Three. So, you can see the formula here. If I am directly in rating grade nine, then I am Stage Three. If my PD is below the low threshold, then I am Stage One. Otherwise, I will check if my PD has relatively increased by more than double, or if my rating has gone below the threshold, then it is Stage Two. Otherwise, it will remain in Stage One. So, in this way, I will do staging classification. What is the benefit of this? If it is in Stage One, then you have to make provisions for one-year ECL, not lifetime ECL. You do not need to be conservative. If it is going into Stage Two, you have to keep more provisions. You have to be conservative. You have to provide for lifetime ECL. Right? And if it is going into Stage Three, then obviously there you have to consider PD directly as 100%, and your ECL will be very large. Right? So, this was your ECL calculation. You can see the diagram below. You will see losses in different scenarios: base scenario, up scenario, down scenario. In this plot, these losses are visible here. I have also plotted cumulative PD, that for different rating grades, the bank computes cumulative PD in this way, given different macroeconomic scenarios and weightages to those macroeconomic scenarios. Here, you can see the diagram below. This red plot is your base economy. I am shifting that economy based on G. On the left-hand side, meaning the economy is deteriorating. I am shifting the economy on the right-hand side, meaning the economy is improving, or it is around the base. So, these three diagrams show that you have to shift the economy in different scenarios, towards the good situation, towards the downturn scenario, and only towards the base scenario, right? Okay, here I have used some levers. You can see in different scenarios, as I have shown you, if you increase the weight of the downturn, then your ECL will accordingly increase, right? Second, I have also kept provision for some hypothetical scenarios here. If I have to shift the economy by 0%, then my ECL will be equal to the baseline scenario. If I shift my economy by 20 percentile on the positive side, right? Then you can see my ECL is higher in the base scenario, and the hypothetical scenario has lower ECL. If I shift my economy by 20 percentile on the lower side, then my ECL is lower in the base scenario, one year as well as lifetime, and my hypothetical scenario has higher ECL, right? So, we can make different plots to demonstrate how this entire tool is computing my PDs at a point in time and calculating ECL based on that, based on which banks will maintain provisions, right? Here, we have modeled many components. We understood the staging criteria. We learned to calculate point-in-time PD. We learned to calculate point-in-time LGD. Good, the PD that I am computing at different times is called PD term structure. The LGD that I am computing at different times is called LGD term structure. So, I calculated the PD term structure. I calculated the LGD term structure. I calculated the effective interest rate. I considered different scenarios. In those scenarios, I took the macroeconomic index. Right? Besides that, I am assigning probabilities to the scenarios, and ultimately, I am computing one-year ECL and lifetime ECL, right? According to your bank, if you need any customized solutions, then you can definitely come in contact with us, where we can model different components by looking at your data, because there are many methodologies for each component. There are many ways to adjust PD based on macroeconomic indices. So, based on what data your bank has, we can choose the method together. So, if you want any guidance or customized solution in that regard, you can definitely come in contact with us, right? So, this was all from my side. I hope the session was useful. I explained the entire framework of how we can implement IFRS 9 and how to calculate expected credit loss within a bank, right? If you want to come in contact with us, you can see the details in the thumbnail below, right? This calculation is only for one loan. If you have to do it for the entire portfolio, then you will have to create such a model for every loan. For that, you will have an engine. The ECL of all those loans will be added up, and ultimately, your bank's ECL will be calculated, right? So, this is all from my side. Thank you very much.