Unrealized gains

[Charles Day]

On Mon, Jun 30, 2008 at 11:34 AM, Charles Day <cedayiv at gmail.com> wrote:

> On Mon, Jun 30, 2008 at 9:16 AM, Derek Atkins <warlord at mit.edu> wrote:
>
>> Quoting Charles Day <cedayiv at gmail.com>:
>>
>>  Recording only the realized gains will put the books out of balance. When
>>>> you sell, you have a realized gain on the quantity sold and an
>>>> unrealized
>>>> gain on the quantity remaining. You have to record them both.
>>>>
>>>> Try removing the unrealized gain splits from either of the examples I
>>>> attached previously and look how it throws off the trial balance.
>>>>
>>>>
>>> Just to (hopefully) clarify, unrealized gains that occur PRIOR to a sale
>>> don't have to be recorded until a sale actually occurs. I think that
>>> might
>>> be what's throwing you off.
>>>
>>
>> I still don't think I'm being thrown off here.  Honestly, I think the
>> Trial Balance report is broken in that it's not properly using the PriceDB
>> entries to compute the UNREALIZED gains.  The only time you should need
>> to add splits are for actual realized gains (or losses, of course).
>>
>
> Ah, so you're saying that the Trial Balance report ought to be figuring out
> ALL unrealized gains, as opposed to just those that have occurred since the
> last exchange (which is what it appears to be doing). So then I wouldn't be
> forced to do any revaluation by hand.
>
> That would certainly make the bookkeeper's job easier, but I'm wondering
> how the report would actually be able to compute unrealized gains correctly.
> The report would have to know the basis of each unit of non-home-currency
> and each unit of non-currency held. Since GnuCash doesn't keep track of
> *which* units are sold when a sale occurs, how would it know the basis of
> the unsold units? (Perhaps there is a way, but it's not coming to me at the
> moment.)
>

I did some reading of the Trial Balance report code... the first thing I
noticed is that when it figures out the basis it always does so using an
average weighted exchange rate that is computed from buys AND sells, rather
than just buys. The net effect is that unrealized gains that occurred prior
to the most recent sale are not included. If the algorithm was changed to
exclude sells, the report might include all unrealized gains correctly - but
only if we assume that the user always uses the average cost method when
calculating realized gains. Not a valid assumption.

It seems to me that one way to get the Trial Balance to correctly compute
all unrealized gains, without changing any code, is to record sales
differently from what is recommended in the Concepts Guide. For example, the
sale occurring at t2 in your example (sell 10 shares @ $3) can be recorded
like:

Memo               Account                Qty Price  Buy   Sell
================== ====================== === ===== ====== ======
Basis              Assets:Broker:XYZ Corp -10     1         10.00
Realized Gain      Income:Realized Gains                    20.00
Deposit proceeds   Assets:Broker                     30.00

When recorded this way, the computation method for realized gains doesn't
matter any more and the trial balance actually balances. But of course, now
execution price isn't recorded.

Thoughts?

-Charles

With trading accounts, all unrealized gains are tracked directly in the
> books in a way that makes entering revaluation splits unnecessary. The
> report system wouldn't need to calculate the unrealized gains.
>
> -Charles
>
> Let me try to prove my point, and I'll try to do this by using an
>> example.  Assume we start out by buying 100 shares of X at date T0
>> for price P0.  Then on date T1 we see the price changed to P1.  On
>> date T2 we sell 10 shares for price P2.  On T3 we see the price
>> is now P3.  Then on T4 we sell another 10 shares for P4.  Just to
>> make numbers easy, let's say P[i] = $i+1 (e.g. P0 = $1, P1 = $2, etc).
>>
>> So we have the following data points:
>>
>> Date     Action     Txn Shares  Txn Price   Txn Value   Gain/Loss
>> -----    ------     ----------  ---------   ---------   ---------
>> T0        Buy         100           $1        $100         N/A
>> T1        None         0            $2                     N/A
>> T2        Sell        -10           $3         $30         $20
>> T3        None         0            $4                     N/A
>> T4        Sell        -10           $5         $50         $40
>>
>> Note that there are NO GAINS at times T1 and T3 because it's JUST
>> a pricedb entry.
>>
>> Similarly, you don't need to RECORD the unrealized gain of the
>> 90 shares at time T2 or the 80 shares at time T4 because you can
>> compute that FROM the pricedb entry.  The only gain you need to
>> record is the actual realized change from the actual exchange.
>>
>>  -Charles
>>>
>>
>> -derek
>>
>> --
>>      Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
>>      Member, MIT Student Information Processing Board  (SIPB)
>>      URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
>>      warlord at MIT.EDU                        PGP key available
>>
>>
>