Hi Ella. Paid loss is generally from the insurer's perspective, that is, what the insurer paid on the claim. That said, a loss triangle can be prepared based on other definitions of paid loss, as long as the definition is consistently represented in the historical data.
Thank you for this video, this is exactly what I was looking for. I am still a bit unsure about the 'Loss Date'. When I first looked at these triangles I thought the 'Claim Lodgment Date' would be the best date to go by, since it will be each row will be fixed to contain only claims made in 2016 for example. I was thinking that using 'Loss Date' would make each row a little harder to interpret as there could be some quite large jumps year on year. For example if a claim is lodged in 2018 for an injury which occurred in 2016, there may be a substantial jump in the numbers between 24 and 36 months. I am thinking that such a jump in the numbers will cause a misrepresentation of the loss development for claims with a loss date and lodgment date in 2016.
Hi Linda. I am not familiar with the term "lodgment date", but it sounds like it may be the date the claim is reported to the insurer. Loss date (or "accident date" or "date of loss") is generally used to create a triangle because it organizes the data in a manner that matches the insurer's obligation to carry reserves. Insurer's are generally required to carry reserves for all accidents that have occurred through a certain date, regardless of whether they were reported by that date. As you point out, late reported claims can lead to upward development in a triangle organized by accident year; however, this is a desirable feature from the standpoint of reserving. To the extent late reported claims are present, their influence is reflected in the development inherent in the triangle. The resulting development patterns can be used to anticipate future development for the more recent (less developed) years. You are correct that a report year triangle would be more stable but it is not as useful for standard loss reserving purposes. Sorry, it is a bit difficult to describe why this is the case verbally. Let me know if you still have questions. Thank you!
It's a well-explained video but I have a doubt about other claim metrics and different types of periods. Specifically, when should we consider each type of claim, and how do we select the corresponding periods? Could you provide more insights on these topics, including loss trending?
Hi @Noob-qf4yo. The most common triangle structure uses annual evaluations (the columns are evaluated at 12, 24, 36 months, etc) on an accident year basis. The structure may vary based on the data you have available and the goals of your analysis. A reason the "standard" structure is useful is that many industry triangles (and resulting statistics) are available on this basis. This can be useful when trying to compare company data to industry benchmarks. There are an endless variety of combinations of metrics that can be created in a triangle. Think of it this way: any metric that changes over time can be represented in a triangle form. Let me know if you have a question about a specific metric. I like your suggestion about loss trending and put that on my list of future videos. Have a great day! Don
Yep, I got that, but I'm confused about the specific scenarios where we look at accident year losses and in which cases we consider policy year, calendar year, or reporting year.
@Noob-qf4yo I see. It really depends on the circumstance. For loss reserving purposes, losses are generally arranged by accident year. In pricing, losses are often arranged by policy year. For claims-made policies, losses are organized by report year. In all cases, calendar year metrics can be calculated. Calendar year measures reflect the change in AY / PY / RY metrics within a specific calendar year. This video may help: czcams.com/video/OGvSV2aFV9s/video.htmlsi=rqq9yw9p8pc4WJCA The best advice I can give you is to attempt understand the difference between the different types and when it arises in a work context, it will make more sense.
Well, that makes sense. Now, another thing is confusing me. Let's consider conducting loss development based on paid losses, leading to ultimate losses. However, I'm confused with the equation ultimate losses = paid + unpaid, where unpaid includes case reserves and IBNR. How does the development on paid losses differ from that on the unpaid portion? Or is it through paid losses that we derive the remaining part of the ultimate loss equation, meaning unpaid losses through ultimate losses developed from paid losses - actual paid losses = IBNR + Case reserves? I'm not sure if I'm heading in the right direction.
Hi @Noob-qf4yo. I think you are heading in the right direction logically. Here are some formulas to summarize your thinking: Estimated unpaid loss = future expected development on paid loss, Estimated unpaid loss = case + estimated IBNR, therefore: Future expected development on paid loss = case + estimated IBNR. Does this help clarify the issue?
@Archer Actuarial Consulting, Thanks for the videos. Is there any way an online seminar can be organized to exactly explain what Actuaries or Actuarial analyst do in the day to day work? I meant real life example.
Hi Collins. That is an interesting idea. I will think more about it. In the meantime, I am happy to share my knowledge of the profession over a short phone call. You can find my email on my website. Regards, Don
interesting video! my question is if the column is incurred loss at 12,24,36,48,60 months after inception (expiry) while the row is period of insurance(eg. 01/01/15 to 31/12/2015), shall i take the first loss run data from the beginning of 01/01/15 or the end of 31/12/15? thank you
Hello! For a triangle where the first row is accident year (or policy year 2015), the first entry would be from the loss run at December 31, 2015. Any loss run before January 1, 2015 would have no values associated with accident year 2015. Hope that helps!
Hello Nowera. The loss runs are generally provided by a claims administrator and are based on actual loss amounts (paid loss in this example). In your example, the subtotal of the 2016 claims changed because the cumulative paid loss amounts change over time. The first evaluation was at 12/31/2016 and the second at 12/31/2017.
Hello and thanks for the feedback! You can create an incurred loss triangle in a similar way described in the video for the paid loss triangle. The only difference is that instead of using paid loss, use the sum of paid loss and case reserves.
This is the best educational video on triangles!! Thank you so much for this!. Thank you.
I think this is one of the most beautiful videos I have ever seen on youtube. Thank you so much!
This really is fantastic info. Beautifully presented, clearly explained. Thanks (from a non-actuary 😊)
I usually never give comments. But this is amazingly explained. Thank you very much
I'm happy you found it useful - thanks for the encouragement!
Fantastically explained. A very big than you.
Nicely explained. Very helpful. Thank you for taking the time to create this and other videos. Hope to see more.
Thank you Rajesh, I appreciate the comment!
This is a great video. Does the paid loss mean what the company paid or the insurer paid on the claim?
Hi Ella. Paid loss is generally from the insurer's perspective, that is, what the insurer paid on the claim. That said, a loss triangle can be prepared based on other definitions of paid loss, as long as the definition is consistently represented in the historical data.
Thank you for this video, this is exactly what I was looking for.
I am still a bit unsure about the 'Loss Date'. When I first looked at these triangles I thought the 'Claim Lodgment Date' would be the best date to go by, since it will be each row will be fixed to contain only claims made in 2016 for example.
I was thinking that using 'Loss Date' would make each row a little harder to interpret as there could be some quite large jumps year on year. For example if a claim is lodged in 2018 for an injury which occurred in 2016, there may be a substantial jump in the numbers between 24 and 36 months. I am thinking that such a jump in the numbers will cause a misrepresentation of the loss development for claims with a loss date and lodgment date in 2016.
Hi Linda. I am not familiar with the term "lodgment date", but it sounds like it may be the date the claim is reported to the insurer.
Loss date (or "accident date" or "date of loss") is generally used to create a triangle because it organizes the data in a manner that matches the insurer's obligation to carry reserves. Insurer's are generally required to carry reserves for all accidents that have occurred through a certain date, regardless of whether they were reported by that date. As you point out, late reported claims can lead to upward development in a triangle organized by accident year; however, this is a desirable feature from the standpoint of reserving. To the extent late reported claims are present, their influence is reflected in the development inherent in the triangle. The resulting development patterns can be used to anticipate future development for the more recent (less developed) years.
You are correct that a report year triangle would be more stable but it is not as useful for standard loss reserving purposes. Sorry, it is a bit difficult to describe why this is the case verbally. Let me know if you still have questions. Thank you!
Best explanation ever…..
Awesome, thank you!
It's a well-explained video but I have a doubt about other claim metrics and different types of periods. Specifically, when should we consider each type of claim, and how do we select the corresponding periods? Could you provide more insights on these topics, including loss trending?
Hi @Noob-qf4yo. The most common triangle structure uses annual evaluations (the columns are evaluated at 12, 24, 36 months, etc) on an accident year basis. The structure may vary based on the data you have available and the goals of your analysis. A reason the "standard" structure is useful is that many industry triangles (and resulting statistics) are available on this basis. This can be useful when trying to compare company data to industry benchmarks.
There are an endless variety of combinations of metrics that can be created in a triangle. Think of it this way: any metric that changes over time can be represented in a triangle form. Let me know if you have a question about a specific metric.
I like your suggestion about loss trending and put that on my list of future videos.
Have a great day!
Don
Yep, I got that, but I'm confused about the specific scenarios where we look at accident year losses and in which cases we consider policy year, calendar year, or reporting year.
@Noob-qf4yo I see. It really depends on the circumstance. For loss reserving purposes, losses are generally arranged by accident year. In pricing, losses are often arranged by policy year. For claims-made policies, losses are organized by report year.
In all cases, calendar year metrics can be calculated. Calendar year measures reflect the change in AY / PY / RY metrics within a specific calendar year. This video may help: czcams.com/video/OGvSV2aFV9s/video.htmlsi=rqq9yw9p8pc4WJCA
The best advice I can give you is to attempt understand the difference between the different types and when it arises in a work context, it will make more sense.
Well, that makes sense. Now, another thing is confusing me. Let's consider conducting loss development based on paid losses, leading to ultimate losses. However, I'm confused with the equation ultimate losses = paid + unpaid, where unpaid includes case reserves and IBNR. How does the development on paid losses differ from that on the unpaid portion? Or is it through paid losses that we derive the remaining part of the ultimate loss equation, meaning unpaid losses through ultimate losses developed from paid losses - actual paid losses = IBNR + Case reserves? I'm not sure if I'm heading in the right direction.
Hi @Noob-qf4yo.
I think you are heading in the right direction logically. Here are some formulas to summarize your thinking:
Estimated unpaid loss = future expected development on paid loss,
Estimated unpaid loss = case + estimated IBNR, therefore:
Future expected development on paid loss = case + estimated IBNR.
Does this help clarify the issue?
@Archer Actuarial Consulting, Thanks for the videos.
Is there any way an online seminar can be organized to exactly explain what Actuaries or Actuarial analyst do in the day to day work? I meant real life example.
Hi Collins. That is an interesting idea. I will think more about it.
In the meantime, I am happy to share my knowledge of the profession over a short phone call. You can find my email on my website. Regards, Don
interesting video! my question is if the column is incurred loss at 12,24,36,48,60 months after inception (expiry) while the row is period of insurance(eg. 01/01/15 to 31/12/2015), shall i take the first loss run data from the beginning of 01/01/15 or the end of 31/12/15? thank you
Hello! For a triangle where the first row is accident year (or policy year 2015), the first entry would be from the loss run at December 31, 2015. Any loss run before January 1, 2015 would have no values associated with accident year 2015. Hope that helps!
Hello, how do you calculate the loss runs? for example as in 2017 the subtotal of 2016 changed how do we calculate this?
Hello Nowera. The loss runs are generally provided by a claims administrator and are based on actual loss amounts (paid loss in this example).
In your example, the subtotal of the 2016 claims changed because the cumulative paid loss amounts change over time. The first evaluation was at 12/31/2016 and the second at 12/31/2017.
Very useful..Thank you. 🤝
Thank you!
thank you very much it is very explanatory. how do i get the video of incurred loss triangle?
Hello and thanks for the feedback! You can create an incurred loss triangle in a similar way described in the video for the paid loss triangle. The only difference is that instead of using paid loss, use the sum of paid loss and case reserves.