Could you please explain the following for me? In some cases, the data being projected may develop with a negative tail. In these situations, we may consider the assumed percentage developed unsuitable as a weight in this credibility method. We may use different weights, or where the incurred development displays a negative tail, we may consider it is more appropriate to apply the Bornhuetter-Ferguson method to the paid development.
Recall that the BF method is a credibility weighted average between the chain ladder method and the initial expected loss ratio method, where the weights are the expected percentage developed according to the chain ladder method. If you expect a negative tail, then the current development percentage is above 100%, so it is no longer a true credibility estimate. Hence you might consider using paid instead of incurred.