For use in subset of my thesis, I’ve been given some exchange market data for several exchange-listed products, including Eurodollar rate futures as well US treasury futures and Fed Funds futures. I would like to fit an approximate yield curve / multi curve (govt + libor) while explicitly omitting OTC data (e.g. swaps) and cash US treasury data. My goal is to fit these curves to some lower dimensional space then observe deviations of my projected prices from true market prices.

Some questions I have:

  1. I can pull implied repo data at the start of each day. Is this stable enough to reproduce real market pricing? To pivot my first credit-risk inclusive future (ED) and/or my first riskless (FF? 2yr treasury future?) off of, I will take the respective spot rate at the start of the day as well. I understand FF is uncollateralized (domestic/reserve-holding) corporation-based yields, which are surely not government. However I don’t really know how else to get short end into my risk-free approach. Is there a better way given the data?
  2. Because I am not using swaps, it’s unclear how I might consider credit-based libor premium. My earliest treasury is 2 years, but I also have Fed funds. I’m not sure if there’s an stripped down modern equivalent that operates under the no-OTC constraints?
  3. Assuming stable implied repos, I can use intraday futures to imply cash treasuries then determine yields from these implied cash treasuries. Does this seem reasonable?

Any guidance or advice from the practical side (I’ve never done this before) would be highly appreciated.

  • 1
    $\begingroup$ What are you going to use for the long end of the curve? Liquid futures only trade out to basically 2yr $\endgroup$ – AlRacoon Jan 25 '19 at 16:28
  • $\begingroup$ For clarification, I am talking about the LIBOR curve. In yield curve generation, many yield curves are generated for different currencies and rates. In USD, the 2 basic curves are USD LIBOR and UST. $\endgroup$ – AlRacoon Jan 25 '19 at 16:46
  • $\begingroup$ I was under the impression Eurodollar futures were decently liquid out for 5 years, and had planned to use that. I didn’t really have a good plan for further out than that however, where I think OTC would usually kick in. Naturally, that creates an issue since the swap spread goes negative at the far long end, but mostly only care risk inclusive curve to at most 10yr. Still an issue to be sure $\endgroup$ – Michael Clinton Jan 25 '19 at 20:32
  • 2
    $\begingroup$ What is your aversion to using OTC instruments? At the end of the day, you are constructing a yield curve to price things. The key is that the rate (price) be liquid enough that you should be able to price and subsequently trade at that price. If there are sufficient participants (users and dealers) in OTC markets, it shouldn't matter whether it trades on an exchange or OTC. (Unless you somehow do not think the OTC markets are a fair market; or maybe you don't have access to trades and levels in the OTC markets. $\endgroup$ – AlRacoon Jan 25 '19 at 21:01
  • 1
    $\begingroup$ Agree with your points but basically: (1) I don’t have access to the data; (2) if I did somehow acquire it, it would not be time aligned which could problems; (3) if time aligned, it still isn’t always firm quotes / necessarily actionable. Surely practitioners must do this (not sure what sources people use—Bloomberg?), but seems very tough for me. My uni does have a terminal for what it’s worth. But I’m trying to study some high frequency features, so to me, use of OTC data seems problematic for believability mostly from for timing reasons. $\endgroup$ – Michael Clinton Jan 26 '19 at 17:19

There is a Fed Funds/UST basis. You can't just 'splice' FF futures and Treasuries - they're two different markets. e.g. over last ~5y, 2y UST/OIS spread has ranged from ~-10 to +30.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.