6-Month T-Bill 2024 Year End Projection Review

This Opinion post first appeared in my blog here: https://pwlcm.wordpress.com/2024/12/20/6-month-t-bill-2024-year-end-projection-review/

Disclaimer: This post is just for educational sharing purposes. Please do your own due diligence on any products mentioned in this post.

__

As mentioned at the end of my blog post 6-Month T-Bill Mid Year Projection Review in 2024, I would do a year end review. With the last T-bill of the year closed its auction yesterday (19 December 2024), now is the time!

Here is the tabulation of the projection results in second half of 2024:

On average the Yield Premium Method gave a projected yield pretty close to the actual cut-off yield, and it had a higher hit rate (84.6%) of capturing the actual yield within its projected yield range.

On the other hand with the Polynomial Trendline Method, the first 5 projections during the second half of 2024 were increasing overestimating the cut-off yields, which triggered a switch from 3rd Order to 2nd Order Polynomial Trendline Projection as described in my blog post BS24118Z 6-Month T-Bill Yield Projection.

The switch resulted in the Trendline Method underestimating the cut-off yields, initially seemingly with decreasing underestimations, but soon the deviations widen and returned to 0.5x% level. As a result of the larger deviations, the Polynomial Trendline Method had a low hit rate (only 7.7%).

In view of the less reliable projections of the Polynomial Trendline Method, I resorted to only pick the projected yields from the Yield Premium Method in subsequent T-bill issues.

Combining the projection results from the first and second halves of 2024, here is the projection statistics for 2024:

Clearly, the Yield Premium Method has better projection power than Polynomial Trendline Method, with smaller deviations and higher hit rate.

This makes logical sense because the former method is based on the institutional sentiments reflected from the MAS Bill auction results just 1 or 2 days prior to the T-bill auctions.

Whereas setting the order of polynomial in the latter method is very much dependent on the dynamic of interest rate environment. While experimenting with the polynomial order in Excel, I found out that only 3rd order has a lift at the end of the curve, and the rest of order 2,4,5 will project a drop to 0%, which is unlikely with the Fed slowing down the rate cuts next year. So I intend to switch back to 3rd polynomial order for this method.