Litigation Support in Antitrust Situations

By Lester E. Heitger and D. Larry Crumbley

Since Daubert [509 U.S. 579, 113 S.Ct. 2786 (1993)], trial judges have had a special responsibility to ensure that testimony is not only relevant, but also reliable. Before permitting an individual to testify as an expert, a judge must determine whether the expert’s reasoning and methodology can be appropriately applied to the facts of the dispute [Stagl v. Delta Airlines, Inc., 117 F.3d 76 (CA-2, 1997)]. In addition, judges must take into consideration an expert’s background and practical experience when deciding whether the expert is qualified to render a report and testify [McColluck v. H.B. Fuller Co., 61 F.3d 1038 (CA-2, 1995)]. A judge may decide that an expert’s expertise is too general or too deficient to qualify [Trumps v. Toastmaster, 969 F. Supp. 247 (S.D. N.Y. 1997)]. A judge also may decide that research and studies cited are too dissimilar to the facts involved in the litigation [General Electric Co. v. Joiner, 522 U.S. 136, 118 S.Ct. 512 (1997)].

Overview of Antitrust Laws

U.S. antitrust laws are an outgrowth of the reaction to business practices during the early years of the country’s industrial development. Some businesses were seen to be using any tactic at their disposal to force competitors out of business and create a monopoly, then raising prices to relatively high levels. This practice was particularly common in the railroad and oil industries.

Federal legislation prohibited the formation and continuation of monopolies except when in the best interest of the public (for example, many local electric, gas, and water utilities).

It is easier to oppose a monopoly, however, than to agree upon whether one exists. In the recent antitrust litigation brought against the Microsoft Corporation, a court found the company to be a monopoly, but the final settlement over its business practices did not satisfy its critics. Similarly, widespread deregulation in formerly monopolistic industries has not always resulted in marketplaces as competitive as critics have desired.

Actions to Be Taken

The antitrust process begins with a complaint filed with the U.S. government by an individual who believes that a company is in violation of antitrust laws. If the government agrees with the complaint, an attempt will be made to remedy the situation by negotiations and other measures. If this attempt fails, the U.S. Justice Department may start proceedings against the business, to force compliance with antitrust laws through the courts.

Sometimes monopoly complaints are not accepted. The complaint might not provide enough information or evidence to pursue the issue, or the Justice Department may not have the resources to pursue all valid complaints. Individuals or other companies have the recourse of filing a lawsuit themselves under the antitrust laws.

The Role of Accountants

In antitrust disputes, accountants may be called upon to determine whether there is liability under the antitrust laws. The primary issue that forensic accountants address is whether a defendant has engaged in predatory pricing. Pricing below an appropriate measure of a defendant’s costs is a requirement to predatory pricing liability [Brooke Group Ltd. v. Brown and Williamson Tobacco Corp., 209 U.S. 209 (1993)]. Accountants may also be asked to calculate the damages a party has sustained as a result of a violation of the antitrust laws.

For example, in an antitrust suit in Indianapolis, a grocery chain was accused by another grocery company of engaging in predatory pricing activities during an 18-month period. The accused grocery chain had gone from a profitable position in earlier periods to significant losses during the alleged predatory pricing period. The plaintiff alleged that the defendant’s change from profitable operations to major losses was evidence of predatory pricing. The defendant countered by saying its losses were due to efforts to compete on a price basis with another major new grocery retailer that had entered the area.

The defendant hired a forensic accountant to demonstrate that the company was not engaged in predatory pricing. The accountant noted that experiencing business losses was not a clear sign of predatory pricing. The forensic accountant ran regression analyses on store data from both parties. He found that the defendant’s stores had in all periods priced its products above average variable costs and therefore, in his opinion, had not engaged in predatory pricing. This expert found that the plaintiff in the dispute had also experienced major losses during the period and, in his opinion, the plaintiff was closer to the predatory pricing benchmark than was the defendant.

The plaintiff then changed the geographic area covered by its complaint to smaller geographic areas in an effort to find some areas of operation in which the defendant had priced products below average variable cost. The defendant’s forensic accountant analyzed store cost data for each of these new areas and found in each instance that the allegations were false. This analysis provided by the forensic accountant was instrumental in the court’s decision for the defendant.

Actions that constitute predatory pricing. Predatory pricing is the act of pricing a product low in order to drive competitors out of business. This definition, although descriptive, does not give a judge or jury much guidance in determining whether predatory pricing has occurred. Because of that, the courts have used a simpler definition to apply the concept of predatory pricing in actual situations. The operational definition is whether a company prices its products or services below “average variable cost.”

Knowing what types of cost behavior patterns are common in business is not enough to allow an expert or analyst to merely look at a company’s ledger accounts and identify which costs are fixed, variable, or mixed. Most organizations have few costs that are captured, recorded, and reported by their cost behavior pattern. Instead, costs typically are identified and recorded by their functional characteristics, such as product cost and period cost. For example, period costs include most operating costs, such as marketing, customer service, personnel, product warranty, accounting, and administration. Product costs include direct materials, direct labor, and manufacturing overhead. Management, investors, and lenders find functional classifications useful, but they are of little use in determining whether pricing is predatory.

Without more information about individual costs, it is necessary to determine the cost behavior pattern of each cost. An accountant can use a variety of methods to estimate cost behavior patterns. A working knowledge of a firm’s accounting system often provides some insight into the nature of cost behavior in the company’s accounting system. Sometimes account titles are misleading, or accounts are used to record items not originally intended. A quick way for an accountant to become familiar with an accounting system is to spend time talking to one or more seasoned veterans with the company that know about its history and how its accounting system works.

Graphic Analysis

Although experience and intuition are valuable in giving an expert insight into an accounting system, they are seldom enough. A better approach to gain insight into cost behavior patterns is graphic analysis. Plotting data and creating scatter diagrams can be useful in showing an expert the general trend or nature of costs. Analysts can easily plot data from a spreadsheet and use that information to form an initial understanding of cost relationships and cost behavior patterns.

Another cost analysis method frequently used by consultants is the high-low method. With this approach, the highest and lowest costs are identified, along with their related activity levels. The difference between the two costs and the two activity levels is calculated and then divided by the difference in activity levels to determine an estimate of the variable cost per unit. This variable cost is then inserted into the total cost equation to find the fixed cost component.

Example. A company is analyzing 20 months of overhead cost data. The highest overhead cost for any month was $300,000, at an activity level of 100,000 units. The lowest overhead cost for any month was $280,000, at an activity level of 90,000 units.

Variable cost = $300,000 - $280,000
                          100,000 - 90,000
                          units            units
                     =  $20,000
                           10,000 units
                     =  $2 per unit
Fixed cost = $300,000 -- (100,000 units x $2 per unit)
                   = $100,000
Total cost = $100,000 + $2 per unit

The high-low method of cost analysis is a quick and easy approach to obtain some understanding of cost behavior information in a dispute. But this approach does have potential flaws. For example, this approach uses only the highest and the lowest values of the cost data in the analysis. If these values are outliers, representing unusual data, the resulting cost behavior information will probably be distorted. Business phenomena such as wildcat strikes or malfunctioning machinery can cause such anomalies, and cost analyses or projections should not be based on an atypical level of activity. In addition, there are only two data observations used in the analysis; the rest of the data are ignored.

Both graphic analysis and the high-low method are intended to help an analyst understand the nature of an accounting system and the characteristics and nature of the company’s costs. The high-low method is particularly useful early on, when an expert may have only highly aggregated data. Financial information gathered and analyzed in this way is preliminary and provides only initial observations and guidelines. Sometimes such analyses are useful in determining what documents should be requested during the discovery process, or in determining in a general sense what damages might be present in a dispute.

Regression/Correlation Analysis

Regression or correlation analysis is a statistical technique for measuring the nature and strength of the relationship between two variables. When applied to cost accounting data, regression/correlation can be used to determine the relationship between specific cost data and some measure of activity, such as sales dollars or production volume. Regression analysis is used to determine the nature (direct or inverse) of the association between two or more variables in the analysis. Correlation analysis is used to measure the strength of the association between the variables.

Simple Linear Regression

The most common type of regression/correlation analysis used in cost behavior analysis is simple linear regression. This method uses two sets of variables: a dependent variable, which usually is cost, and an independent variable or predictor variable, which usually is the measure of the volume of activity. In simple linear regression, an accountant enters cost and activity data into the regression model. The regression model will use the data to compute a regression line, which is represented by the formula:

Y = a + bX

where Y is the computed value of the dependent variable obtained from a specific value of the independent variable; a is the constant, or intercept, where the regression line crosses the vertical axis (the value of Y when X = 0); b is the slope of the regression line, describing the change in the value of Y for each unit change in the value of X; and X is the value of the independent variable.

A positive value for b indicates that the dependent and the independent variables are positively correlated (moving in the same direction). A negative b value indicates that the dependent and the independent variables are moving in opposite directions. The Exhibit shows sample cost and activity data and the resulting regression line.

The regression line is often referred to as the least squares regression line or the line of best fit. The regression line is the line which does the best job of minimizing the squared distance between the regression line and the individual data observations used in computing the regression line.

Standard Error of the Estimate

Of course, business cost data seldom show perfect correlation. Most cost categories are not created to include only costs that are fixed, variable, or some other characteristic. When actual cost data are regressed against some measure of activity, that data likely will have some amount of variability. A measure of the variability of the cost data in the regression analysis is called the standard error of the estimate. The greater the variability of the data, the less precise the estimates can be.

Typically, many estimates are point estimates made using the regression parameters. For example, an expert may use the regression to estimate the total costs that a business should have incurred at a given volume of activity. Because there is some variability in the regression data, an expert may incorporate a measure of the regression variability and some desired confidence level in determining a range of possible values.

To do this, an expert can use the standard error of the estimate and values from either a T table or a Z table for the desired confidence level. For example, assume the point estimate for a cost is $256,000 and the standard error of the estimate is $25,500. An expert wants to find the range of costs within which the actual cost should fall 95% of the time. The sample size used in the regression analysis was 60 months of data. Either the T table or the Z table gives a value of 1.96 for a 95% confidence interval. So the range of projected cost is $256,000 + ($25,500 x 1.96), or $206,020 to $305,980. (T tables are used to find values for sample sizes less than 60. The smaller the sample, the larger the T value, which adjusts for sampling error. Z tables are used for large sample sizes. Because many business regressions are run using small samples, T tables are commonly used.)

Correlation Analysis

Correlation analysis measures the strength of the relationship between the variables. There are several measures that describe the strength of the relationship between the two variables. The first is called the coefficient of correlation, or r, which measures the strength of the association between the dependent and the independent variable. The coefficient of correlation ranges in possible values from –1 to +1, indicating perfect negative or perfect positive correlation between the two variables.

If the coefficient of correlation is squared, it becomes the coefficient of determination, which can range from 0 to +1. The coefficient of determination measures the amount of variance explained by changes in the independent variable. For example, a coefficient of determination of .84 means that 84% of the dependent variable’s total variance can be explained by changes in the independent variable. High r-squared values mean that there is a strong linear relationship between the variables.

The coefficient of nondetermination (1 – r squared) is the portion of the relationship between two variables that is not explained by the coefficient of determination. This unexplained variance may be random fluctuations or it may be a variation that is explained by some other independent variable.

Association Versus Causation

Regression and correlation analysis shows the degree of association between variables, but it does not prove causation. There is a tendency to think that if an analysis shows a high degree of correlation, movements in the independent variable cause the changes in the dependent variable. However, one must look to other supporting information before making such a claim.

If regression/correlation results show relationships that seem illogical, factors affecting the results may need to be evaluated. The relationship may be merely a spurious correlation or a coincidence. Sometimes in regression and correlation analysis the numbers happen to be ripe for a relationship that just does not make any logical sense. For example, a California state representative reputedly ran a regression of teacher salaries in the state of California against gambling revenues in Las Vegas and found a high coefficient of correlation. He then concluded that teachers were all squandering their raises at the Las Vegas casinos. Although teachers may have been doing just that, the fact that teachers’ salaries and gambling revenues were moving in the same direction did not prove that one caused the other.

Reasons for the Unexpected

There are many reasons why accounting data regressions may not turn out as expected. Some of the accounting reasons that regressions may yield unexpected or perplexing results are:

• Allocations. Many cost analyses contain costs allocated from shared services. How these costs are allocated and timed can make a big difference in the cost behavior results from the analysis.
• Transfer prices. When products or services are transferred (sold) between various parts of a company, a transfer price or a chargeback must be determined. A transfer price based on the market price, when regressed against units produced, will appear to be a directly variable cost, even though the creation of the product or service will entail a mixture of fixed and variable costs.
• Entity concept. Transfer pricing issues bring up the question of which entity the forensic accountant should study to determine the cost behavior pattern for the transfer item. The answer depends upon the objectives of the cost analysis.
• Accounting policies. Many accounting policies have an important impact on an analysis of cost behavior. If the accounting policy is to charge depreciation to units only once each year rather than each month, a cost analysis will have different results. Similarly, accounting policy may state that various costs are allocated based on the number of units receiving the allocation rather than on dollars of sales. A difference in allocation method can have a major impact on cost behavior.

The Need to Become Familiar with Accounting Systems

Forensic accountants must be adequately informed about the nature and operation of the accounting system for each and every business under evaluation. If two companies differ in their accounting treatment of a particular type of cost, an analysis may suggest different cost behavior for the same cost.

For example, consider two similar companies that allocate corporate marketing costs to all of their stores on a weekly basis. One company allocates weekly marketing costs on a per-store basis. The weekly sales dollars achieved by each store has no impact on the amount of cost allocated to it. The other company also allocates weekly marketing costs to its 50 stores, but its allocation is based on each store’s actual sales for the week as a percentage of the company’s total sales. If a store’s relative sales volume changes, the store’s allocation of corporate marketing costs will change accordingly.

This example illustrates how a company’s accounting policies can have a significant impact on the cost behavior estimates that result for the regression/correlation analysis. Many accounting policies have an impact on accounting data analysis. Company policies on when to write off prepaid items, at what level the firm records depreciation and in what timeframes, and on how to record employee fringe benefit costs can all impact the results of a regression/correlation analysis.

Close