Skip to content
All library documents

Building USD Discount and Projection Curves from Market Data

Article Quant Q&A · Author: MikeRand

Summary

The document discusses how to construct USD yield curves from publicly available data for valuing interest rate swaps and FX forwards. It distinguishes a discount curve from a LIBOR projection curve: the response recommends bootstrapping an OIS curve for discounting, then bootstrapping LIBOR swaps while discounting with that OIS curve. It also notes that FX forward valuation requires market forward points. For a government curve, another response suggests using zero-coupon Treasury prices instead of constant-maturity yields, which may provide more maturity points but could need smoothing.

The question centers on whether Federal Reserve H.15 series alone are adequate, and the answers point to additional swap and Treasury price sources. These are practical suggestions rather than a full curve-building recipe: the text gives no instrument conventions, interpolation choices, market quotes, or validation results. The discounting setup also assumes a USD portfolio collateralized in USD, so the appropriate curve depends on the portfolio’s collateral terms and market conventions.

Key ideas

  • An OIS curve can serve as the USD discount curve when the portfolio is collateralized in USD.
  • A LIBOR projection curve can be bootstrapped from LIBOR swaps using the OIS curve for discounting.
  • FX forward valuation requires market data for forward points.
  • Zero-coupon Treasury prices can provide more curve points than constant-maturity yields, though smoothing may be needed.
  • The discussion does not specify curve conventions or demonstrate valuation accuracy.

Tags

Full text
# Is it possible to derive a reasonable USD yield curve using only publicly/freely available data sources?


# Is it possible to derive a reasonable USD yield curve using only publicly/freely available data sources?












Background

I'm a corporate financial analyst with a small derivatives portfolio (amortizing interest rate swaps and FX forwards) looking to value these derivatives "properly" (which, in my case, means "without gross error" as opposed to say the multiple decimals of precision required for real quants).

Question

Using only publicly available free data sources (e.g. the H.15 from the Fed), should I be able to reconstruct both a risk-free and a LIBOR yield curve (risk-free for discounting, LIBOR for swap cash flow forecasting)?

What I've figured out so far

1) One approach to the risk-free would be to bootstrap the treasury constant maturities series in the H.15.

2) In theory the LIBOR should come from bootstrapping the interest-rate swap series in the H.15.

But it doesn't feel like either of these are consistent with approaches discussed on this site advocating using the OIS (which I believe is just the effective federal funds on the H.15) to derive a risk-free curve.

## Answer by Randor (score 2)

https://quant.stackexchange.com/a/21017

bootstrap fedfunds (ois ) swaps to get your discount curve (asuming your portfolio is usd, and is usd collateralised). strangely i dont see the data on the fed site. i see data on LCH's site: http://www.lchclearnet.com/asset-classes/otc-interest-rate-derivatives/volumes/settlement-prices-swapclear-global#usd

to get libor projection curve, you need to bootstrap libor swaps, keeping in mind that they are discounted on the curve you got just now from the ois.

for the fx forwards, you need to get fx forward points market data...

## Answer by Alex C (score 1)

https://quant.stackexchange.com/a/21028

Instead of the constant maturity series (which IMO would give only a few points), you could use the prices of ZCB to get the USD curve. They are available here http://www.wsj.com/mdc/public/page/2_3020-tstrips.html It might require some slight smoothing to get a clean curve. This is the best way I know to get a US Govt curve for free.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.