Please start any new threads on our new
site at https://forums.sqlteam.com. We've got lots of great SQL Server
experts to answer whatever question you can come up with.
Author |
Topic |
yipchunyu
Yak Posting Veteran
80 Posts |
Posted - 2003-01-25 : 11:43:46
|
anyone wrote a function to calcuate internal rate of return? |
|
ValterBorges
Master Smack Fu Yak Hacker
1429 Posts |
|
yipchunyu
Yak Posting Veteran
80 Posts |
Posted - 2003-01-25 : 11:59:31
|
yes, i learned the calclation method but anyone already wrote a udf for it? |
|
|
ValterBorges
Master Smack Fu Yak Hacker
1429 Posts |
Posted - 2003-01-25 : 12:31:13
|
post the formula you want and someone will write a udfIs it to calculate irr for a single investment or multiple?Edited by - ValterBorges on 01/25/2003 12:31:21Edited by - ValterBorges on 01/25/2003 12:48:23 |
|
|
yipchunyu
Yak Posting Veteran
80 Posts |
Posted - 2003-01-25 : 12:44:57
|
stated in the link u provided, IRR is defined by the equation: NPV(C, t, IRR) = 0. In other words, the IRR is the discount rate which sets the NPV of the given cash flows made at the given times to zero.In general there is no closed-form solution for IRR. One must find it iteratively. In other words, pick a value for IRR. Plug it into the NPV calculation. See how close to zero the NPV is. Based on that, pick a different IRR value and repeat until the NPV is as close to zero as you care. I searched through the web and find one function written by a guy named BobI don't get a sql server to test it. but is this the function i need ?/*************************/-- net present value-- npv = sum(cf(t)/(1+r)^t) for t=0 to n-- where cf(t) is the cash flow at time t-- and r is the discount rateif exists (select * from dbo.sysobjects where id = object_id('dbo.npv') andxtype in ('FN', 'IF', 'TF'))drop function dbo.npvGOcreate function dbo.npv (@rate real) returns realbegin declare @npv real -- return value declare @t int declare @cf money set @npv=0 set @t=0 declare cur cursor for select cf from test open cur fetch next from cur into @cf while @@FETCH_STATUS = 0 begin set @npv = @npv + @cf * power(1+@rate, -@t) set @t = @t+1 fetch next from cur into @cf end close cur deallocate cur return(@npv)endgo/*************************/-- internal rate of return-- irr is defined as the discount rate at which the npv of the cash flows isexactly zero-- the only way to solve for irr is through iteration-- the irr can be multivariate or undefined, therefore a guess value isrequired-- irr and npv are inverse functions-- a good test is the npv of the cash flows at a discount rate equal to theirr should-- equal zero (or very close to zero)if exists (select * from dbo.sysobjects where id = object_id('dbo.irr') andxtype in ('FN', 'IF', 'TF'))drop function dbo.irrGOcreate function dbo.irr (@rateguess real) returns realbegin declare @delta real -- rate delta in 2-point formula set @delta=.0001 -- .0001 equals one hundreth of one percent declare @epsilon real -- criteria for success, npv must be within +/-epsilon of zero set @epsilon=.005 -- .005 equals one half cent declare @maxtry smallint -- number of iterations allowed set @maxtry=10 declare @irr real -- return value set @irr=null -- assume failure declare @rate1 real declare @rate2 real declare @npv1 real declare @npv2 real declare @done smallint declare @try smallint set @done=0 set @try=0 while @done=0 and @try<@maxtry begin set @rate1 = @rateguess set @npv1 = dbo.npv(@rate1) if abs(@npv1) < @epsilon begin -- success set @done=1 set @irr=@rate1 end else begin -- try again with new rateguess set @rate2 = @rate1 + @delta set @npv2 = dbo.npv(@rate2) set @rateguess = @rate1 - @npv1*(@rate2-@rate1)/(@npv2-@npv1) set @try = @try + 1 end end return(@irr)endgo/*************************/-- setup test table of cash flows, first cash flow at t=0if exists (select * from sysobjects where id = object_id('test') and sysstat& 0xf = 3) drop table testGOcreate table test (cf money not null)goset nocount oninsert test (cf) values (-100)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)insert test (cf) values (10)set nocount offgoselect dbo.npv(.1)goselect dbo.irr(.05)go-- the net present value of the internal rate of return should be very closeto zeroselect dbo.npv(dbo.irr(.05))go |
|
|
ValterBorges
Master Smack Fu Yak Hacker
1429 Posts |
Posted - 2003-01-25 : 12:53:05
|
This looks good npv can also be written without a cursor this might be worth doing if your table has lots of records for multiple accounts.Returns-35.0493970.016231281.879624E-5Simplified the npv function however table must have a sequential column./*************************/ -- net present value -- npv = sum(cf(t)/(1+r)^t) for t=0 to n -- where cf(t) is the cash flow at time t -- and r is the discount rate if exists (select * from dbo.sysobjects where id = object_id('dbo.npv') and xtype in ('FN', 'IF', 'TF')) drop function dbo.npv GO create function dbo.npv (@rate real) returns real begin declare @npv real -- return value SELECT @npv = SUM(cf*power(1+@rate,-pid))FROM testreturn(@npv) end go /*************************/ -- internal rate of return -- irr is defined as the discount rate at which the npv of the cash flows is exactly zero -- the only way to solve for irr is through iteration -- the irr can be multivariate or undefined, therefore a guess value is required -- irr and npv are inverse functions -- a good test is the npv of the cash flows at a discount rate equal to the irr should -- equal zero (or very close to zero) if exists (select * from dbo.sysobjects where id = object_id('dbo.irr') and xtype in ('FN', 'IF', 'TF')) drop function dbo.irr GO create function dbo.irr (@rateguess real) returns real begin declare @delta real -- rate delta in 2-point formula declare @epsilon real -- criteria for success, npv must be within +/- epsilon of zero declare @maxtry smallint -- number of iterations allowed declare @irr real -- return value set @delta=.0001 /*-- .0001 equals one hundreth of one percent */set @epsilon=.005 -- .005 equals one half cent set @maxtry=10 set @irr=null -- assume failure declare @rate1 real declare @rate2 real declare @npv1 real declare @npv2 real declare @done smallint declare @try smallint set @done=0 set @try=0 while @done=0 and @try<@maxtry begin set @rate1 = @rateguess set @npv1 = dbo.npv(@rate1) if abs(@npv1) < @epsilon begin -- success set @done=1 set @irr=@rate1 end else begin -- try again with new rateguess set @rate2 = @rate1 + @delta set @npv2 = dbo.npv(@rate2) set @rateguess = @rate1 - @npv1*(@rate2-@rate1)/(@npv2-@npv1) set @try = @try + 1 end end return(@irr) end go /*************************/ -- setup test table of cash flows, first cash flow at t=0 if exists (select * from sysobjects where id = object_id('test') and sysstat & 0xf = 3) drop table test GO create table test (pid int not null, cf money not null) go set nocount on insert test (pid,cf) values (0,-100) insert test (pid,cf) values (1,10) insert test (pid,cf) values (2,10) insert test (pid,cf) values (3,10) insert test (pid,cf) values (4,10) insert test (pid,cf) values (5,10) insert test (pid,cf) values (6,10) insert test (pid,cf) values (7,10) insert test (pid,cf) values (8,10) insert test (pid,cf) values (9,10) insert test (pid,cf) values (10,10) insert test (pid,cf) values (11,10) set nocount off go select dbo.npv(.1) go goselect dbo.irr(.05) go -- the net present value of the internal rate of return should be very close to zero select dbo.npv(dbo.irr(.05)) go Edited by - ValterBorges on 01/25/2003 15:16:16 |
|
|
yipchunyu
Yak Posting Veteran
80 Posts |
Posted - 2003-01-25 : 22:18:02
|
ValterBorges, thx a lot for your help. I will try it when I back to office tomorrow. BTW, what do u mean table must have a sequential column?besides, what parameters should i pass to the funcion to calculate?sorry for my ignornamce |
|
|
ValterBorges
Master Smack Fu Yak Hacker
1429 Posts |
Posted - 2003-01-25 : 22:53:54
|
Notice that in my post there is a pid from the formula in the link it's the t[i] valuequote: We first have the net present value or NPV: NNPV(C, t, d) = Sum C[i]/(1+d)^t[i] i=0where: C[i] is the i-th cash flow (C[0] is the first, C[N] is the last). d is the assumed discount rate. t[i] is the time between the first cash flow and the i-th. Obviously, t[0]=0 and t[N]=the length of time under consideration. Pick whatever units of time you like, but remember that IRR will end up being rate of return per chosen time unit.
You would pass it your best guess at the irr and it will find it because it uses an approximation technique Looks like newton's method to me.http://groups.google.com/groups?q=newton%27s+approximation+method&hl=en&lr=&ie=UTF-8&oe=UTF-8&selm=RL%25v9.170362%24md1.36573%40sccrnsc03&rnum=6Make sure you read the part about f'(x) being close to 0.Because if @npv2-@npv1 is zero you'll get a divide by zero errorwhich means you need a better initial approximation.set @rateguess = @rate1 - @npv1*(@rate2-@rate1)/(@npv2-@npv1) until if finds a rate where the npv is 0.Take a look at this thread.One guy mentions that starting out with a guess of 10% and using newton's approximation which has quadratic convergence will give you good results most of the time. http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&threadm=868045948.9041%40dejanews.com&rnum=12&prev=/groups%3Fhl%3Den%26lr%3D%26ie%3DUTF-8%26oe%3DUTF-8%26selm%3D868045948.9041%2540dejanews.com%26rnum%3D12It's been a long time since I took a numerical analysis course but let me try to explain convergence. One simple method to find a value in a liner function is the bisection method. Image you have a point on a line or an array of values and you have 100 items and they are sorted. It will take at most 10 tries to find the value you're looking for because every time you split the search space in half. It's like guessing for a number a someone telling you higher or lower. Ex: Let's say I can pick a number from 1 to 100 and I pick 1 however you don't know that and you guess the following50I say lower and you guess25I say lower and you guess13I say lower and you guess6etc...so at most it would take you ten times for 100 values. 10^2=100Well newton's method is more sophisticated and it would find the correct value in 4 tries. 4^4 > 100 that's why they call it quadratic convergence. I also remember that some method's work better than other's in different situations and that sometimes these approximations never converge but you'll have to pick up a numerical analysis book to get a better explanation.Wow I can't believe I forgot so much.All you mathematicians please forgive me and please feel free to correct me if i'm wrong.Edited by - ValterBorges on 01/26/2003 00:12:02 |
|
|
nerellav
Starting Member
1 Post |
Posted - 2011-08-11 : 07:55:55
|
@yipchunyuHave you written a UDT for this? if yes, can you please post the code. when I tried to implement this IRR function I am getting a divide by zero exception. Thanks in advance.Vamsi. |
|
|
|
|
|
|
|