Monday, February 23, 2009

A History of Mark to Market Accounting

Accounting for Derivatives and
How bonuses are paid.
© Raymond May 2008

It was a Monday morning in 1995, at that time I was the head trader of US fixed income derivatives in New York for JP Morgan. I arrived at my trading desk at 60 Wall Street later than usual, just after eight having battled through my commute from New Jersey to find a post-it attached to my computer screen to "call Bill". I soon got the message that I was "done"- "oh not to worry the trade had been done at very good levels and a printout was on my desk". "Done" meant a trade had been executed and I owned the resulting position. I quickly scanned the one page and realized it was a very BIG trade, the biggest I had ever seen, at least $15 billion! This was the first time I had ever had a trade executed on my book without consulting me for pricing information. The bond market was open, futures opened in 10 minutes but were trading on the electronic platform. Shock! the futures were printing red, very red, indicating a very large downward market movement. A big market movement on an ordinary day was 10 basis points or one 10th of a percent. This Monday the screen was RED and down big, the futures were showing down more than 50 basis points, the market was not yet open, there was nothing I could do but wait. We were down $25,000,000 before we started, why was I the last person on the planet to know about this trade? As the futures opened the prices continued to fall, now down 75 basis points – I felt ill?

Come bonus time, the bank had lost 25 million but the great and good Investment Bankers that put the deal together claimed to have made 25 million. Now it was all politics and Bill was king and his word was all that counted. So the bank loses all the way round, 25 million on the deal and then pays out a bonus! I of course am blamed for losing 50 million as if it was separate and no one was interested in the truth or it ensuring it did not happen again.

This is a small example of the bonus situation in 2008. And this example occurred in 1995! it has got much worse since then. The best way to get a bonus is to arbitrage the accounting system! There are many ways to take advantage of the accounting system but the simplest is to record a profit based on some model today that will not be realized until sometime in the future.

If there was ever a case when the fixed / variable explanation for high investment banker bonuses 2008 was it, this is the true test. The argument put forward is that investment bankers receive low salaries and a majority of the remuneration is paid in the form of a bonus based on performance. In 2008 all these institutions have lost huge amounts of money and the variable piece should be "zero" but that is not what is happening. "I made money, not my fault someone else lost it" goes the cry. There is some truth to that and anyone truly adding value should be paid, but who is the judge of this value? they certainly don't have my confidence based on my experience. Very few people add much above the benefits offered by the organization (systems, network, processes, brand).

Many good analyses are available on this crisis including in January 09's Economist, which highlights how well the financial models in interest rates, foreign exchange and equities performed. And that it was the poor performance of the CDO (collateralized debt obligation) that lies at the root of this crisis. Although this is a good analysis one element of the investment bank that stands out from my own experience of 13 years in the City and Wall Street which is the true root cause is the accounting model, which revolves on the "bonus pool". Or as I like to say, "getting close to the pig trough".

I remember someone saying to me in the 80's "why invest in Merrill Lynch? if the company makes money the management takes it and if Merrill loses money you take it" it really made me think, and it has disturbed me every since. And now the whole world knows.

When an investment banker claims to have contributed a "$50mm" profit, this is very debatable, but until this year no one debated it. First the 50mm is usually "revenue" not profit and second the revenue is based on very unclear mark to market models and respective allocations.

The basic aim in the accounting model it to align the employees self interest with that of the organization. Make them make money and the rewards are shared. In a classic sales role a salesman is rewarded with commission on successful sales. This is a model well understood in all businesses, but in the financial markets when you are talking CDO or Swaps, what is the "profit"? It all depends on the accounting method. In many businesses "commission" depends on "cash-in" and not just profit. This has not been the case on Wall Street.

Lets go back to the beginning. At JPM we began executing "swaps" in the early 1980's based out of London. A swap (or interest rate derivative) is a relatively simple concept - combine a floating rate note with a fixed rate note allows only the net coupons payment over time between a fixed interest rate and the later market interest rate. This introduced the now popular LIBOR index. This allowed Treasurers to manage interest rate payments without having to call and re-issue bonds and debt. When I joined the accounting group at JP Morgan in London the few deals that had been executed were accounted for as two loans! One fixed and one variable. Each month the accounting department calculated the interest receivable and payable and the net total was the monthly earnings... nice and simple. Accept we did it all by hand, can you imagine that today! 1000 deals computed one at a time! It took all month, so we calculated the monthly earnings just in time to start again. The month earnings number did not vary much month to month, it was very stable.

There is a funny story here, each month we had to report earnings to New York on the 4th day of the month, we never had a chance – we knew the big number – e.g. 4 or 5 million, it was something close to last month, but in order to get the rest of the number we would need many more days, so we simply made them up by taking the last 6 figures from a random 10 pound note!

The average length of the deals at that time was approximately five years. Even with no computers this business generated stable positive income. This was a result of how the business was managed, simply and conservatively - by executing new deals in groups. First a transaction was executed and later one or more equal and opposite transactions where grouped together, leaving a closed group with a positive earning stream. There was no market to market at this time.

My first task was to develop a computer system to help manage this portfolio. It took six months and many long nights. It was at this time that I was approached by the banks external auditors to confirm that the portfolio did not hold a "future loss". Oh what a question? What was the present value of this portfolio? At this time we had no model, the zero curve was a thing of the future. It took 12 months to answer the question. The first model we developed we named 1 plus i, or principal plus interest.

The One plus i model:
From the observed yields on treasury bonds for all maturities and credit spreads we were able to define a current yield curve. All cashflows from all deals are calculated and complied into a single list of date and cashflow amount, the sum of all cashflows by day – as a result each day had either a positive or negative daily total. Moving to the most far out cashflow and dividing it my the appropriate current market rate (i) giving a P amount and an I amount (example - cashflow of 20mm and i =10% on12/31/2020, gives a P = 18.18mm and I = 1.82mm. The P is saved and the I amount is added/subtracted to each original amount listed on each 12/31 annual anniversary. This is performed for each date back to the present date stepping forward one day at a time. The sum of all the P's is the present value of the portfolio. I hope you followed that! Quite simple really and required no complex models. The difficulty was to generate an accurate list of all cashflows. At this stage all the deals where nothing more than paper deal confirmations.

The first run of our model took place on the 8th August 1988 (8.8.88) and we valued 66,666 cashflows! we did this on a mainframe and those two numbers were printed on the top of the green striped computer paper! very chilling. But the value was minus a few trillion and needed much further reconciliation. Between August 1988 and the final audit in January 1989 we were able to calculate the correct number with much certainty, including accounting for future hedging and operational costs. This opened the debate on mark to market accounting, which was easily won and implemented at JP Morgan in 92 based on that first 1988 model. The business head had told me before we started out that “there was over 250mm in NPV”, he was right on the mark!

Once we had developed a method for calculating the mark to market or current value we needed a model to do this daily not just annually. This was accomplished by reversing the 1+i model and creating the zero coupon curve which was first accomplished in that same year 1988, and now a market standard. We reached the zero curve in two places quite independently. In the late summer of 1988 I received a visit from Bob Barker a New York based researcher that had been given the take of developing a pricing model for a swap. He wanted to compare his model to mine. It was amazing we had reached exactly the same conclusions and our models matched exactly. I would have to assume that at the same time others around the street were reaching the same answers. I still have my first zero coupon curve workings – hand written proofs. (Bob’s spreadsheet model later became know as 3 + I and was used through the organization until the late 90’s)

The computation of the net present value of this complex portfolio allowed us to develop two further advances. The first was model position and risk management, the representation of this historically portfolio in terms of today’s market – and allow for the accurate hedging of the whole portfolio and no need to use "matched groups" and secondly the accurate reporting of daily profitability – the change in mark to market over time.

The current positions where calculated using two different methods - one was simply to add the P's that resulted from the 1 + i calculation into annual buckets and the second was to "tweak" the current market rates used as inputs one at a time, recalculating the NPV each time. By changing or tweaking just the 10 year treasury yield by one basis point and re-calculating the NPV the portfolio sensitivity to the ten year rate could be calculated and therefore the correct hedge amount. This allowed large portfolios to be accurately managed.

In the summer of 1989 I now had six month data on the mark to market and was able to monitor the performance on a monthly basis of the business - I was shocked they were making a LOSS! I decided to bring this up with management, who informed me I must be mistaken - they were making record profits! Now it was true that the portfolio was making money, but that was being generated by the historical portfolio, all new business was all being added at a loss. My information was not helpful and certainly could not be true. Sound familiar? My word against all those more knowledgeable managers. Eventually after what was years I was able to win the argument but not before many bonuses were paid along the way. But the power of mark to market accounting was demonstrated for all time.

The next question to resolve was how to allocate profitability (or rather new revenue). From 1989 the best model I developed was the "change report" this allocated the change in NPV from one day to the next to all the possible reasons. Classical interest accrual, movement in market rate, new business etc. This needed the market conditions at the time of all new deal to be captured. This was done in a spread sheet and required a "mid market rate" to be agreed, from which new revenue or bid.offer to be calculated. How should this be allocated between the "marketer" and the "trader" to ensure they were incentives to help each other. It was decided to generate a normal bid.offer margin this was subtracted first and allocate this 60% to the trader and 40% to the marketer, excess profit would be allocated 60% to the marketer.

As you can see the trader was incentivized to "shade" his mid market and the marketer to gain an attractive mid market.

Back to the 1995 trade example: The accounting revenue allocation model was set up for normal business and required a tacit agreement between the trader and the marketer. However the trade executed in 1995 was huge, so huge that the ability to hedge was very uncertain in normal times and certainly very tricky when it was public knowledge and the simple revenue allocation set out above would and could not apply, certainly not based on a prior Friday end of day closing prices when by Monday the market knew of the distress the trade was causing and any hedge could only be executed at very reduced pricing. This irresponsible was small compared to the events that occurred in 2006-2007 but indicated what was likely to happen!

The second issue is the use of revenue, uncertain revenue rather than fully costed profit and loss in setting investment bankers bonuses. It is amazing how little interest senior investment banks have on costs and processes allowing operations to run hog wild in these big organizations. Costs don’t affect how senior peoples bonuses are calculated, so why care? If you have even been in a meeting with senior investment bankers and operations managers and watch the bankers eyes glaze over. No they like everything to key of revenue only.

This is my perspective from inside JP Morgan, at that time nothing or very little was shared across the industry. Today these models are available to be purchased.

1 comment:

Lassi Kokko said...

Very good&informative blog.
Keep on writing!