For release on December 10, 1996 1996 Twenty-Fourth Annual National Conference on Current SEC Developments December 10, 1996 Remarks by Robert C. Lipe Academic Fellow Current Accounting Projects Office of the Chief Accountant U.S. Securities and Exchange Commission As a matter of policy, the Securities and Exchange Commission disclaims responsibility for any private publications or statements by any of its employees. The views expressed are those of the author, and do not necessarily represent the views of the Commission or the author's colleagues on the staff. Welcome, to the AICPA conference on current SEC developments. I am honored to speak here today. My topic for today is to discuss how to determine whether hedge accounting is appropriate when initially entering into a futures contract. First, I will provide a little background on the topic. Second, I will discuss two criteria for determining if hedge accounting is appropriate at inception. Loosely speaking, the criteria are economic sensibility and evidence from prior years. And finally, throughout my remarks I will refer to how the issues I am discussing today played an important role in a registrant matter at the SEC. The accounting for buying or selling a futures contract as a hedge is covered in Statement of Financial Accounting Standards No. 80, "Accounting for Futures Contracts," as well as in a recent FASB exposure draft, "Accounting for Derivative and Similar Financial Instruments and for Hedging Activities." As Russell Mallett stated earlier in this session, hedging as defined in SFAS 80 refers to a company buying or selling a futures contract in order to reduce or alter a specified risk that the company faces. This risk can be a risk that market values may move in an unfavorable direction. Or the risk can arise from unfavorable movements in interest rates. If the company's usage of futures contracts meets the criteria set forth in paragraph 4 of SFAS 80, then the gain or loss on the futures contract is deferred until the company recognizes its gain or loss on the underlying asset or transaction. This practice of offsetting the gains and losses is commonly referred to as "hedge accounting." As you may know, the FASB is reconsidering SFAS 80. If the FASB goes ahead with the new exposure draft, then the rules will change somewhat. The gains and losses on the futures contracts would no longer be deferred. However, the exposure draft will continue the practice of hedge accounting by accelerating the recognition of the gains or losses on the underlying. Thus both existing and proposed accounting standards allow gains and losses on some futures contracts to be offset against the gains and losses on the underlying. With this basic background, I would like to turn to my specific topic for today. Under what conditions does being party to a futures contract qualify for this special hedge accounting treatment? It turned out that one of the first registrant matters that I faced at the commission hinged on just this question. Today I will try to provide some guidance on how the staff addresses this question. My remarks will focus on the importance of paragraph 4b of SFAS 80 in determining if a futures contract should qualify as a hedge. The paragraph requires that, at the inception of the hedge, it must be probable that changes in the market value of the futures contract will offset the changes in the fair value the hedged item. In determining whether future offset is probable, registrants should address two questions. First, is there a clear economic relationship between the prices of the hedged item and the futures contract? Second, was there a high level of correlation between these prices during the relevant past periods? Paragraph 4b requires an evaluation of correlation and offset both at inception and on an ongoing basis. Since Russ provided excellent guidance regarding ongoing assessments earlier in this session, I will focus on applying these two criteria at the inception of a hedge. Economic Sensibility First let's focus on the economics. For hedge accounting to apply, it must be economically sensible that some common factor or factors will have opposite effects on the values of the underlying and the futures contract. For example, if you own 100 pounds of gold, you can hedge the risk associated with that inventory by selling a futures contract for 100 pounds of gold. If the price of gold goes up, the value of the futures contract goes down. Thus, hedging an inventory of gold with a futures contract on gold meets the test of being economically sensible. Many of the hedges found in practice are similar to this plain vanilla transaction. However, for a variety of reasons, registrants will sometimes use strategies that are less straight forward. For example, they might try to hedge their gold by selling a futures contract to buy copper. Paragraph 4b of SFAS 80 mentions that hedge accounting may be applied to these more complicated transactions. However, there must be a clear economic relationship between the underlying and the futures contract. For this to occur in my example, the prices of gold and copper must move in similar ways over time. In contrast, hedge accounting would not be appropriate if the prices of gold and copper reflect different economic factors. This example of using copper futures contracts to hedge gold is similar to the facts concerning the registrant matter mentioned earlier. The registrant used a hedging instrument that had some similarity to the underlying asset, but the correspondence was not perfect. Thus the registrant faced numerous questions from the staff regarding whether the economics of this hedge made sense. We kept asking "Is a high level of correlation among price changes probable?" Without a convincing economic analysis, hedge accounting as described in SFAS 80 would not be appropriate. Evidence from Prior Years The second criteria for judging correlation at the inception of the hedge is evidence from past data. In addition to considering whether the hedge is economically sensible, the decision to use SFAS 80 also depends on demonstrating a high level of correlation in past data. How does one do that? Since a lot of my research involves analyzing accounting and price data, let me tell you how I would approach it. The first step is to collect data on past price changes for the underlying asset and the futures contract. To illustrate my points, some simulated data are presented in figure 1 which is attached at the end of this speech. Figure 1 contains two plots of data. On the vertical axis in each graph is the return on the underlying asset. The horizontal axis in panel A is the return on a specific futures contract, labeled "contract A." The horizontal axis in panel B is the return to a different hedging instrument, labeled "futures contract B." What patterns would we look for in these two plots to determine whether historical returns on the futures contract offset the historical returns from the underlying asset? There should be a downward slope to the scatter plot. In other words, large positive returns on the asset should be associated with large negative returns on the hedging instrument, and vice versa. Both graphs display a downward slope. Which futures contract will be the better hedge, A or B? Contract B should provide better offset because the points on the graph are less dispersed. They are closer to being on a straight line. Indeed, if contract B were a perfect hedge, the points would form a straight line. Looking at these scatter plots is fun, but it is difficult to draw firm conclusions from just eye-balling the data. It reminds me of a psychologist asking people what they see when they look at an ink blot; everyone views them differently. What you as preparers and auditors need is a tool to summarize the information in these plots in a simple, easy to understand manner. The most common tool in use today is regression analysis. Figure 2 summarizes the key information provided by a regression program. It also contains the exact same plots as figure 1, along with the fitted regression line. When you run a regression, the program looks for the single straight line that fits these plots the best. By best, I mean that the program minimizes the average squared difference between the data points and the fitted line. Now remember that the whole purpose of the regression is to assess the level of correlation between the futures contract and the underlying asset. Figure 2 contains two measures of correlation, one is the R2 and the other is the correlation coefficient. I want to first focus on R2 . For contract A, R2 is 63%. For contract B, it is 94%. Recall from the scatter plots that contract B looked like a better hedge. The regression analysis confirms this because the R2 is higher for futures contract B. For those of you who are not familiar with R2, the statistic can be as low as 0% and as high as 100%. An R2 equal to 0% means that the changes in the value of the futures contract are unrelated to the underlying asset. An R2 of 100% implies a perfect correlation, which would translate into all of the points in our scatter plot lying on a straight line. What does the R2 of 63% for contract A tell us? In essence, it means that 63% of the historical returns to the underlying asset could have been offset by this futures contract. Certainly, 63% is not a small number, but it is important to realize that 63% is less than 100%. Indeed, 37% of the change in value of the underlying cannot be offset by a hedge using contract A. Thus if past macro- economic factors are essentially repeated in the future, at best, contract A can only offset 63% of the gain or loss on the underlying. That does not seem like a sufficiently high level of offset to justify hedge accounting. What is the magic level of R2 which would allow registrants to apply SFAS 80? I do not want to draw a bright line. Correlation at inception is a judgment call based on more than just this single number. Some CPA firms use a minimum R2 of 80% as guidance, and staff has no objection to that guidance. In cases where the R2 is less than 80%, so that more than 20% of value changes are unlikely to offset, I think staff may question whether hedge accounting is the right answer. That concludes my analysis of the R2 statistic. What about the other regression results? Are they useful? The answer is yes, but it is difficult for me to describe how they would be used in my limited amount of time. So I will simply give you the highlights. The correlation coefficient you see in figure 2 is another measure of correlation. Note that R2 equals the correlation coefficient squared. In other words, looking at panel A, if you square the correlation coefficient of -.79, the result is .63, which by definition will equal the R2 of 63%. So in order to have an R2 of at least 80%, the correlation coefficient would have to be somewhere between -1.0 and -.9. The other output consists of the slope coefficient and the intercept. I used these to draw the line on the scatter plot. To a registrant, the slope coefficient will help determine how many futures contracts must be purchased in order to offset the risk of the underlying asset. If you want more information about this, you can ask me later or call me in the office. Conclusion I will close with an epilogue of what happened in the registrant case I mentioned before. The registrant's historical analysis showed a correlation coefficient of -.77. But a correlation coefficient of -.77 implies an R2 of only 59%. This seemed low to the staff, and we were concerned whether the transaction met the criteria in paragraph 4b of SFAS 80. Sure enough, as Russ described earlier, this particular hedging transaction ultimately fell a part. The gains and losses on the futures contract did not offset the gains and losses on the underlying asset. The staff concluded that these futures contracts did not meet the criteria specified in SFAS 80. Thank you for your attention. Hopefully these remarks will help you or your co-workers or your clients to better assess whether hedge accounting is appropriate at the inception of the transaction. [figures 1 & 2 omitted from electronic copy] Regression output: Contract A: R-square = 63%. Correlation coefficient = -.794. Slope = -.671. Intercept = -.04. Contract B: R-square = 94%. Correlation coefficient = -.968. Slope = -.959. Intercept = -.01.