Spring 2011

Greetings!

Invotex® Group is pleased to share insights about current trends and issues of interest to litigators and counsel, particularly those with which we have recent experience. We hope you find this information informative, and we welcome your feedback.

In this Issue

1. Patent
   Patent Litigation Damages: Apportionment Yes, But How?
2. Economics
   Regression Analysis and Credibility in Legal Proceedings: Part I - Parsimony
3. Valuation
   The Tax Value of Personal Goodwill in the Sale of a Company
4. Complex Litigation
   The Value of Decision Trees in Complex Litigation Settlement Negotiations
5. Continuing Professional Education
   Upcoming Events

Patent

Patent Litigation Damages: Apportionment Yes, But How?
by Scott J. Williams and Marylee P. Robinson

The intellectual property world has been buzzing since the recent Court of Appeals for the Federal Circuit’s decision in the Uniloc USA, Inc. and Uniloc Singapore Private Limited v. Microsoft Corporation case – with the majority of the attention focusing on the demise of the 25 Percent Rule in the calculation of patent infringement damages. However, the Court’s ruling also included continuing support for stricter guidelines in the use of entire market value (EMV) theory and experts and attorneys alike should be mindful of this.

Relying on Lucent Techs., Inc. v. Gateway, Inc.,(1) Uniloc argued that “the entire market value of the products may appropriately be admitted if the royalty rate is low enough.”(2) The Court responded saying, “The Supreme Court and this court’s precedents do not allow consideration of the entire market value of accused products for minor patent improvements simply by asserting a low enough royalty rate.”(3) Further citing the same Lucent case, the Court reinforced, “For the entire market value rule to apply, the patentee must prove that the patent-related feature is the basis for customer demand.”(4) Lastly, the Court stated, “This case provides a good example of the danger of admitting consideration of the entire market value of the accused where the patented component does not create the basis for customer demand.”(5)

This ruling is consistent with another Court of Appeals’ decision in the case of Cornell Univ. v. Hewlett-Packard Co. Here the Court clearly stated that, “An over-inclusive royalty base including revenues from the sale of non infringing components is not permissible simply because the royalty rate is adjustable.”(6) The Court added, “The entire market value rule in the context of royalties requires adequate proof of three conditions:

1. the infringing components must be the basis for customer demand for the entire machine including the parts beyond the claimed invention,
2. the individual infringing and non-infringing components must be sold together so that they constitute a functional unit or are parts of a complete machine or single assembly of parts, and
3. the individual infringing and non-infringing components must be analogous to a single functioning unit.”

The Court gives further guidance in determining an acceptable royalty base in this case, “The logical and readily available alternative was the smallest salable infringing unit with close relation to the claimed invention - namely the processor itself.”(7)

These recent decisions emphasize the need for damages experts and IP valuators to carefully consider whether the patented invention is the basis for customer demand and, if not, what portion of product revenues are relevant. However, apportioning product revenue (or costs) in consideration for the patented technology is often challenging. In many cases there is no separate charge to the customer or other accounting record for the patented feature leaving the expert/valuator to determine its relevant economic contribution.

Information used to make this determination can be found in product marketing materials, internal business plans and forecasts and other internal corporate communications. If that type of information has not been produced in discovery or does not prove useful, then interviews, additional discovery document requests, interrogatories or searching of publicly available marketing information can also provide useful insights. Publicly available sources for data include public company annual and quarterly reports, industry reports, analyst reports and analyst calls. Further, there are various methods an expert/valuator can employ such as determining ratios based on component cost, estimating revenue contributions based on financial reports and using customer surveys. The use of surveys is becoming more commonplace although they can be costly and should be conducted by experienced, credible firms.

As the case law regarding EMV continues to evolve, the courts are requiring greater precision in the calculation of infringement damages. This higher threshold combined with the challenge of a lack of available data, emphasizes a need for diligent analysis and creative thinking by today’s damages experts and IP valuators.

For more information, contact Scott J. Williams.
______

(1) Lucent Techs., Inc. v. Gateway, Inc., 580 F.3d 1301, 1337 (Fed. Cir. 2009).
(2) Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292 (1st Cir. R.I. 2011).
(3) USA, Inc. v. Microsoft Corp., 632 F.3d 1292 (1st Cir. R.I. 2011).
(4) Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292 (1st Cir. R.I. 2011).
(5) Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292 (1st Cir. R.I. 2011).
(6) Cornell Univ. v. Hewlett-Packard Co., 609 F. Supp. 2d279; 2009 U.S Dist. LEXIS 28125.
(7) Cornell Univ. v. Hewlett-Packard Co., 609 F. Supp. 2d279; 2009 U.S Dist. LEXIS 28125.

Back to top.

Economics

Regression Analysis and Credibility in Legal Proceedings: Part I - Parsimony
by Michael J. Petron and Emraan S. Khan

Regression techniques, long established as central to the field of econometrics and statistics, have become increasingly important to lawyers and their clients in adversarial proceedings. Regression analysis—the statistical method for describing the relationship between two or more variables—has since moved from being viewed as unreliable “black magic” to a serious scientific attempt to distill essential truths from complex problems.(1) This trend, dating back to the 1970s, is no doubt due in part to a gradual acceptance by judges and legal policy makers. Regressions have been used as evidence to prove liability under Title VII of the Civil Rights Act of 1964, violations of the Voting Rights Act of 1965, racial bias in death penalty litigation, “but-for” damages in contract disputes and damages in antitrust litigation matters.(2) As electronic databases and recording instruments improve our ability to collect data and as the complexity of business increases in an interconnected world, the reliance on regression techniques and the need to draw accurate conclusions from them will only continue to grow.

The purpose of this series will be to help non-economists better understand some common mistakes of regression models, thereby allowing closer scrutiny of the conclusions they draw. Each segment will introduce the reader to a very basic principal or guideline to use in evaluating a regression analysis in a court-room setting. Due to the fact that legal proceedings may incorporate a statistical analysis whether or not it is adequately performed, a keen attorney should always stay attuned to such insights. We begin the series with a discussion of the concept of parsimony, a principle equally supported by common sense and econometric experience.

Part I - Parsimony
Generally stated: all else equal, the simpler (more parsimonious) the model, the better. It often may seem fitting, when trying to describe a complex world, to use a complex model. For example, when attempting to model the price of coffee mugs, it may be tempting to throw in the kitchen sink: price of tea cups, consumer price index, exchange rates, housing price index, GDP, weather patterns and so forth. One could also rationalize that the coffee mug model warrants an exponential trend, a dummy for each retail shopping season or a lagged indicator of past coffee mug prices. But, while all these added variables may improve the fit of the model, as University of Pennsylvania economics Professor Francis Diebold notes, “Decades of professional experience suggest that… simple, parsimonious models tend to be best for out-of-sample forecasting in business, finance and economics.(3) He explains a number of reasons why smaller, simpler regression models are better than larger, more complex ones:

First, by virtue of their parsimony, we can estimate the parameters of simpler models more precisely. Second, because simpler models are more easily interpreted, understood, and scrutinized, anomalous behavior is more easily spotted. Third, it’s easier to communicate an intuitive feel for the behavior of simple models, which makes them more useful in the decision-making process. Finally, enforcing simplicity lessens the score for “data mining”—tailoring a model to maximize its fit to historical data.(4)

The parsimony principle is also connected to some very practical implications. Experts tend to agree that non-parsimonious models have a much poorer forecasting performance than equally equipped parsimonious models.(5) This means that while the price of coffee mugs regressed on the price of tea cups, GDP, inflation and weather conditions may fit the past data well, it will not provide a very good forecast of the price of coffee mugs next year. As one can imagine, non-parsimonious models may impress the court by recreating what has already happened, but they are potentially flawed when trying to provide a practical forecast of damages or “but-for” estimates. In later installments, after introducing a few more important concepts, we will attempt to show through example the specific detrimental effect that a lack of parsimony has on out-of-sample forecasts.

It should be noted, however, that while simpler models are preferable, they are only preferable all else being equal. The omission of relevant variables can be even more detrimental to the validity of an analysis than the inclusion of irrelevant variables, and there are many other fundamental statistical issues that may require additional complexity to address. On the whole, however, courts do consider the importance of parsimony and the fact that equations cannot include all potential variables. In EEOC v. Sears Roebuck, the court writes:

If too many factors are added that do not significantly affect the dependent variable, the model can become distorted and then may not accurately estimate how much the independent variables influence the dependent variables. For a regression analysis to be meaningful, it is therefore important to strike a balance by including all factors which significantly affect the dependent variable, and excluding those variables which do not significantly affect the dependant variable.(6)

As we will see in later installments, there are specific metrics (above and beyond the basic R-squared) that factor in the inherent tradeoff between a model’s sophistication and parsimony and allow the economist to judge the overall balance of a model.

In Part II of the series, we will discuss particular hazards of ignoring the parsimony rule: namely, data-mining and over-fitting. A logical continuation of the theme of model complexity, the next segment will bring to the reader’s attention the subtle dangers of over-including factors that are considered when conducting a regression analysis.

For more information, contact Michael J. Petron.
______

(1) Fisher, Franklin M. "Statisticians, Econometricians, and Adversary Proceedings." Journal of the American Statistical Association 81.394 (1986): 227.
(2) See, e.g. EEOC v. Sears Roebuck, 628 F. Supp. 1264 (1986); McClesky v. Kemp, 481 U.S. 279 (1987); Cotton Brothers Baking Co. v. Industrial Risk Insurers, 941 F.2d 380 (5th Circ. 1991); State of Georgia v. Ashcroft, 195 F. Supp. 2d 25 (2002); and Norman Law v. NCAA, 94-2053-KHV (1998).
(3) Diebold, Francis X. Elements of Forecasting. 4th ed. Mason, OH: Thomson South-Western (2007): 46.
(4) Diebold, Francis X. Elements of Forecasting. 4th ed. Mason, OH: Thomson South-Western (2007): 46.
(5) See for example, Evans, Michael K. Practical Business Forecasting. UK: Blackwell (2003): 227-228, 256.
(6) EEOC v. Sears Roebuck, 628 F. Supp. 1264 (1986).

Back to top.

Valuation

The Tax Value of Personal Goodwill in the Sale of a Company
by Joseph S. Estabrook and Charles W. Rains

If you are planning for a possible sale of a Personal Service Corporation (“PSC”) or another C Corporation, you may find that a buyer will want the transaction to be an asset deal versus a stock deal. An asset deal is more advantageous to the buyer because it eliminates the possibility of acquiring unknown liabilities. In addition, an asset deal allows the buyer to receive a step-up in the basis of the assets acquired. What this means to the seller is the possibility of double taxation: the corporation will recognize a gain on any appreciated property, and then the shareholder will pay tax on any dividends or liquidating dividends received. Considering that the tax on a corporate gain can be 35%, and then the shareholder pays another 15% on the liquidating dividend, the net effective tax rate can be as high as 45%. Forty-five percent is a substantial tax burden when the majority of the value in a PSC is likely to be goodwill, which typically has no basis. One way to circumvent a portion of this problem is to establish the goodwill as the shareholder’s personal goodwill. Under this scenario, each shareholder can individually sell their personal goodwill, and, in most cases, the shareholder will only pay a 15% capital gains tax on any gain. Since personal goodwill likely represents the majority of the assets’ value in a PSC, the resulting tax savings can be substantial.

The following case study illustrates how such a transaction can be crafted:

A privately held insurance agency was contemplating a merger with/sale of its assets to a larger agency. As in most insurance agencies or professional corporations, there was the potential for the allocation of a substantial portion of the value to the owner’s personal goodwill.(1) To quantify the amount of personal goodwill, the fair market value of the stockholders’ equity of the company was determined. For the purpose of this study, we’ll set the fair market value of the company at $5,000,000. The $5,000,000 needs to be allocated into three buckets: tangible assets, identifiable intangible assets and goodwill.

We first determined that the fair market value of the tangible assets, consisting of current and fixed assets, equaled $200,000. We then identified three intangible assets: trained and assembled workforce, agency relationship (with the national insurance companies) and customer lists and relationships. Our estimate of the fair market value of those intangible assets was $1,800,000. The remaining balance of $3,000,000 represented the goodwill of the company. We then had to allocate the goodwill between the corporate goodwill of the company and the personal goodwill of the owner. In performing this type of analysis, there are a number of factors that can be considered, including but not limited to:

1. The age and health of the professional
2. Professional's demonstrated earning power
3. The professional's reputation in the community for judgment, skill and knowledge
4. The professional's comparative professional success
5. The nature and duration of the professional's firm, either as a sole proprietor or as a contributing member of a partnership or professional corporation
6. Type of service offered
7. Type of client served
8. Length of time at the current location
9. How the fees/services are billed
10. Source of new clients (referral base)
11. The individual professional’s amount of production
12. The number of employees and their length of service
13. Economic and demographic information on the community where the firm is located (firm location)
14. The number of other professionals in the community offering the same service or specialty (level of competition)

After performing our analysis, we determined that 25% of the $3,000,000, or $750,000, was corporate goodwill, and the remaining 75%, or $2,250,000, was personal goodwill.

By analyzing the company’s assets and portioning out corporate and personal goodwill from the other tangible and intangible assets and considering the difference between individual ordinary income tax rates and the 15% capital gain tax rate, the owner benefits from a substantial tax savings by breaking out the sale of their personal goodwill from the overall transaction. Since every company is different, each situation must be analyzed based on specific facts and circumstances. Also, company owners should consult an attorney for guidance.

For more information, contact Joseph S. Estabrook or Charles W. Rains.
______

(1) A word of caution. If a  non-compete or employment agreement between the owner and the company currently exists, all or a portion of the personal goodwill may already have been transferred to the company.

Back to top.

Complex Litigation

The Value of Decision Trees in Complex Litigation Settlement Negotiations
by Edward A. Gold

In the Spring 2010 edition of Litigation Perspectives, Scott Williams discussed the value of developing an early analysis of damages to assist with settlement negotiations. This article extends that train of thought by discussing a particularly helpful analytical tool for negotiating settlement of complex patent and other litigation matters.

It’s not uncommon to have a patent case with multiple patents issued at different times, uncertainty surrounding the interpretation of the claims of those patents, and challenging arguments regarding invalidity of one or more of those patents. What are the damages? The answer depends on the outcome of many findings by judge or jury as to how the claims should be interpreted, whether any patents are invalid, which patents are infringed and therefore over what timeframe the infringement occurs. A decision tree can help evaluate the overall value of the potential outcomes.

Let’s look at a very simple example to get started (see Figure 1). Imagine a litigation matter where the Plaintiff has a 75% chance of winning and a 25% chance of losing. If the Plaintiff wins, she will be awarded $10 million in damages. If she loses, the award is zero. The weighted average of the two possible outcomes is $7.5 million (75% times $10 million plus 25% times zero equals $7.5 million). That simple calculation is the essence of a decision tree. There is an event (the trial) that has some uncertainty associated with it (the chance of winning or losing) and a value (the damages award) that is dependent on the outcome of the event.

Figure 1


Our simple decision tree provides us with a weighted average value of $7.5 million over the two possible outcomes. Notice that if the case goes to trial, the Plaintiff will never win $7.5 million. The only post-trial outcomes in our example are $10 million and zero. But prior to trial, $7.5 million is the amount that takes into account the risk of winning and losing and the amounts that would be awarded with a win or a loss. This weighted average is often called an expected value. Any settlement offer over the expected value is one which the Plaintiff can feel comfortable accepting. Even offers somewhat below $7.5 million might be acceptable if the Plaintiff is risk adverse.

We can also adjust the above calculation to consider litigation costs (see Figure 2). In our example, if the Plaintiff will spend $2 million in litigation costs, win or lose, then the two possible outcomes are $8 million with a trial win and negative $2 million with a loss. The expected value is now 75% x $8 million + 25% x $-2 million = $5.5 million. Note that in this case it does not matter if we subtract the litigation costs directly from the outcome values or we subtract at the end from the expected value. This option is available because the costs were the same regardless of the outcome of the trial. However, if the Plaintiff also had to pay the Defendant’s costs in the event of a loss, then the expected value could only be calculated if the appropriate litigation costs are subtracted from each damages outcome before multiplying by the probability of winning or losing.

Figure 2



In each of these examples, the “tree” consists of two branches emanating from one event. But one can easily imagine more possible outcomes than simply win or lose. In one appellate decision, Judge Richard Posner recommended the use of a four-pronged decision tree analysis when he rejected a class action settlement for various deficiencies including a lack of analytical rigor regarding potential damages outcomes.

A high degree of precision cannot be expected in valuing a litigation, especially regarding the estimation of the probability of particular outcomes. Still, much more could have been done here without (what is obviously to be avoided) turning the fairness hearing into a trial of the merits. For example, the judge could have insisted that the parties present evidence that would enable four possible outcomes to be estimated: call them high, medium, low, and zero. High might be in the billions of dollars, medium in the hundreds of millions, low in the tens of millions. Some approximate range of percentages, reflecting the probability of obtaining each of these outcomes in a trial (more likely a series of trials), might be estimated, and so a ballpark valuation derived.(1)

However, even Judge Posner’s proposed analysis doesn’t begin to convey the power of a decision tree. In more detailed analyses, multiple outcomes can stem from multiple events creating a complex tree-like diagram. For example (see Figure 3), assume a positive outcome in a Markman ruling results in our earlier Plaintiff improving her odds of winning infringement to 90% (with a corresponding 10% chance of a loss) while a negative Markman outcome leads to a 50% chance of winning infringement. Now there are two events, the Markman ruling and the trial itself. We can compute the expected value by multiplying the odds of every combination of events by the damages value for that outcome and add all the probability-weighted values together. So, if the Markman ruling has a 60% chance of a positive outcome, then the expected value (including litigation costs of $2 million) is:

(60% x 90% x $8M) + (60% x 10% x $-2M) + (40% x 50% x $8M) + (40% x 50% x $-2M) = $5.4M

Figure 3



Clearly, a decision tree becomes increasingly complex with each additional event. And as Judge Posner indicates, there will always be a degree of imprecision in the estimation of the probabilities and awards no matter how close to reality one tries to make the tree’s structure. However, with the assistance of a damages expert, modeling a handful of events can be a very cost effective exercise. As the case evolves, or the negotiations progress, the attorneys and damages expert can update the model’s probabilities or outcome values. Changes can be made to one assumption at a time which teaches the negotiation team where the high impact assumptions lie. The team can even try to estimate the opponent’s view of the case by entering the perceived values the opponent would ascribe to the probabilities and damages outcomes. Such an analysis often reveals which differences in opinions about the underlying issues lead to significantly different views in the expected value of the case. With this knowledge in hand, the negotiators may be able to better focus the discussion between the two sides thus bringing the dispute to an amicable and cost effective conclusion.

For more information, contact Edward A. Gold.
______

(1) Cheryl Reynolds, et al. v. Beneficial National Bank, et al., United States Court of Appeals, Seventh Circuit, April 23, 2002, paragraph 20.

Back to top.

Upcoming Events

Back to top.

Disclaimer: The opinions expressed in this newsletter are the opinions of the individual author(s) and may not reflect the opinions of the firm or any other individual associated with the firm.