0
$\begingroup$

I could find lots of stuff online for IR derivatives but it seems there isn't too much on FX for this specific adjustment.

$\endgroup$
2
$\begingroup$

The futures/forward convexity adjustment comes from the covariance between rates and the index. For a future/forward that settles on an index $I_T$ on expiry $T$ the future price is $F_{\text{fut}} = \mathbb{E}^{\mathbb{P}}\left[I_T \right]$, where $\mathbb{P}$ is the risk neutral measure, and the forward price is $F_{\text{fwd}} = \mathbb{E}^{\mathbb{Q}^T}\left[I_T \right]$, where $\mathbb{Q}^T$ is the $T$-forward measure. Applying the change of measure $d\mathbb{P}/d\mathbb{Q}^T = e^{\int_0^T r_t dt} D(T)$ you get $$ F_{\text{fut}} = \mathbb{E}^{\mathbb{Q}^T}\left[e^{\int_0^T r_t dt} D(T) I_T \right] = \mathbb{E}^{\mathbb{Q}^T}\left[ I_T \right] + \mathbb{E}^{\mathbb{Q}^T}\left[(e^{\int_0^T r_t dt} D(T)-1) I_T \right] \\= F_{\text{fwd}} + D(T)\mathbb{COV}^{\mathbb{Q}^T}\left[e^{\int_0^T r_t dt}, I_T \right] $$ When $I_T$ is an FX index or an equity index market practice seems to disregard the covariance term.

You can however get an estimate of its magnitude using a simple Hull & White model with volatility $\sigma_r$ and no mean reversion for $r_t$, and an exponential brownian motion for $I_T$ with volatility $\sigma_I$ and correlation $\rho$ between the two brownians, the formula above becomes $$ F_{\text{fut}} = F_{\text{fwd}} e^{\sigma_r \sigma_I \rho T^2/2} $$ With say $\sigma_r = 50$ bps, $ \sigma_I = 10\%$, $\rho = 25\%$ and $T=5$ you would get $F_{\text{fut}} = F_{\text{fwd}} \times 1.0006$, so it is not entirely negligible.

$\endgroup$
4
  • $\begingroup$ thanks! Makes sense from a mathematical prospective, I was wondering what is the economic/financial reason why futures rates are usually higher than forwards? $\endgroup$ – Student Mar 26 at 17:20
  • $\begingroup$ if positive correlation between index and rates, gains on upward move of index are more likely to be reinvested at a higher rate, losses at a lower rate. $\endgroup$ – Antoine Conze Mar 26 at 17:40
  • $\begingroup$ Usually interest rate futures have price given by 100-R, so if the underlying rate decreases I get excess margin and I can invest it at lower rates in this case. Can you clarify which IR futures contracts you have positive price-rates correlation so that when price goes up I can invest the extra margin at a higher rate? thanks $\endgroup$ – Student Mar 27 at 16:38
  • $\begingroup$ your OP was on FX futures. Obviously for an IR future that settles on 100-R the index and IR are negatively correlated. $\endgroup$ – Antoine Conze Mar 29 at 7:02
1
$\begingroup$

The futures/forward convexity adjustment for non-interest rate futures only tends to matter for futures with maturities greater than a year (which tend to be part of bespoke structures and not traded in size on screen). You can get a closed form solution in a GBM-Ho-Lee hybrid model if you don't mind grinding though some partial differential equation work you will find that the convexity adjustment is proportional to $\frac{1}{2}\rho\sigma_r\sigma_xT^2 + \frac{1}{3}\sigma_r^2T^3$

So we need a convexity adjustment even if the two Brownian motions are uncorrelated, since the short rates still drives the drift of the underlying process which means there will be a terminal correlation between it and the cash account.

$\endgroup$
3
  • $\begingroup$ Hi. Just a comment, here $\sigma_x$ is the volatility of the spot price in the hybrid model, and needs to be calibrated from market implieds $\sigma_I$ which are Black & Scholes volatilities of forward prices. Hence the additional term in the convexity adjustment formula. $\endgroup$ – Antoine Conze Mar 25 at 9:53
  • $\begingroup$ We have the closed form solution for that calibration though (dbl check me as I did this on scrap paper now)? $\sigma_I^2 = \sigma_x^2+2\rho\sigma_r\sigma_x(T-t)+\sigma_r^2(T-t)^2$. I don't think that removes the extra adjustment term? $\endgroup$ – river_rat Mar 25 at 15:17
  • $\begingroup$ the correlation between $r$ and the spot also has to be calibrated from the correlation between $r$ and the forward, or vice-versa $\endgroup$ – Antoine Conze Mar 25 at 15:25

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.