Lognormal frequency distribution and Modified convexity: Difference between pages

From ACT Wiki
(Difference between pages)
Jump to navigationJump to search
imported>Doug Williamson
(Classify page.)
 
imported>Doug Williamson
(Categorise the page and align with ACT glossary.)
 
Line 1: Line 1:
A lognormal distribution is one where the logarithm - for example log(X) or ln(X) - of the variable is normally distributed.  
(MC).  


Lognormal distributions have a minimum - usually 'worst case' - value, whilst having an infinitely high upside.
Broadly speaking, modified convexity measures the curvature of an instrument’s price function, as yields change - from a given starting point - by a small amount.  


A simplified illustration is set out below.
More strictly, it is the rate of change of modified duration with respect to yield - at the given starting yield.




A simple (non-symmetrical) lognormal distribution includes the following values:
Modified convexity can be calculated from Convexity as follows:


0.01, 0.1, 1, 10 and 100.
'''Modified Convexity = C<sub>MOD</sub> = Convexity / (1+r)<sup>2</sup>'''


The median - the mid-point of the distribution - being 1.


Where:


This distribution is skewed: most of the values being in the lower (left) part of the distribution, the upside being infinitely high, and the downside limit being 0.
r = periodic yield = [[Nominal annual rate]] / compounding frequency per year


The logs - for example to the base 10 - of these values are:


log(0.01), log(0.1), log(1), log(10) and log(100)
The estimation of price change for a given small change in yield can then be calculated as follows:


= -2, -1, 0, 1 and 2.
Price change estimation using Modified Duration (MD) only:


= - Price x MD x Change in yield


When the parent values are lognormally distributed, the transformed (log) values follow a (symmetrical) normal distribution.
Price change estimation using Modified Convexity (C<sub>MOD</sub>):


So for example the mean, mode and median of the log values above (including -2, -1, 0, 1 and 2) would all be the same, namely the middle value 0.
= - [Price x MD x (Change in yield)] + &frac12; x [Price x C<sub>MOD</sub> x (Change in yield)<sup>2</sup>]
 
 
Because the value v yield relationship is a curve and not a straight line (values do not change linearly as yields change) the estimate of change in value using only modified duration will generally underestimate the new value (because the curve lies above its tangent). 
 
Therefore the modified convexity adjustment is always positive - it always adds to the estimate of the new price whether yields increase or decrease.
 
It is also possible to estimate the MD and the C<sub>MOD</sub> from given observations of Price and Yield, by rearranging them to solve for MD and C<sub>MOD</sub> - effectively running the price change estimation formulae in the other direction.




== See also ==
== See also ==
* [[Frequency distribution]]
* [[Convexity]]
* [[Leptokurtic frequency distribution]]
* [[Matching]]
* [[Lognormally distributed share returns]]
* [[Modified duration]]
* [[Median]]
* [[Normal frequency distribution]]


[[Category:The_business_context]]
[[Category:Interest_Rate_Risk]]
[[Category:Managing_Risk]]

Revision as of 08:59, 29 July 2014

(MC).

Broadly speaking, modified convexity measures the curvature of an instrument’s price function, as yields change - from a given starting point - by a small amount.

More strictly, it is the rate of change of modified duration with respect to yield - at the given starting yield.


Modified convexity can be calculated from Convexity as follows:

Modified Convexity = CMOD = Convexity / (1+r)2


Where:

r = periodic yield = Nominal annual rate / compounding frequency per year


The estimation of price change for a given small change in yield can then be calculated as follows:

Price change estimation using Modified Duration (MD) only:

= - Price x MD x Change in yield

Price change estimation using Modified Convexity (CMOD):

= - [Price x MD x (Change in yield)] + ½ x [Price x CMOD x (Change in yield)2]


Because the value v yield relationship is a curve and not a straight line (values do not change linearly as yields change) the estimate of change in value using only modified duration will generally underestimate the new value (because the curve lies above its tangent).

Therefore the modified convexity adjustment is always positive - it always adds to the estimate of the new price whether yields increase or decrease.

It is also possible to estimate the MD and the CMOD from given observations of Price and Yield, by rearranging them to solve for MD and CMOD - effectively running the price change estimation formulae in the other direction.


See also