Varying Assumptions
While some firms have explored market-based methods, the vast majority of other companies likely will use financial models to calculate option values. These standard models require several input assumptions, key among them an estimate of the underlying share price's future volatility. But more importantly, the same problems that bedevil market valuation of employee stock options--nontransferability, vesting periods, forced exercise and forfeitures--pose significant challenges for model-driven valuation. While option theory dictates that options should almost always be held to expiration, the reality is that many employees exercise grants early due to termination, tax considerations, or beliefs about future returns, and employees forfeit unvested shares. Any pricing model must account for these factors, either by including them in the model or making adjustments to key assumptions.

Together, these factors leave companies with much flexibility in the calculation of option prices. If a company uses the Black-Scholes pricing model, all these adjustments are lumped together into a single assumption about the expected term of the options. This is, needless to say, a difficult estimate to make, as historical experience provides almost no insight into this estimate. The SEC acknowledged this difficulty in Staff Accounting Bulletin No. 107 in March 2005, where it offered a "plain vanilla" formula.

While transparency and improved disclosure is the goal of Statement 123R, all this flexibility leaves an investor in an unusual situation: two companies with nearly identical characteristics (size, stock price volatility, industry sector, employee base) might make exactly the same stock option grants, but they could end up calculating very different values. How can an investor make comparisons between these two companies?

ISS Findings
To better understand this problem, ISS reviewed the first mandatory disclosures under Statement 123R from 36 companies with fiscal year end dates from August to October 2006. (To date, about 60 companies have filed financial statements with Statement 123R disclosures.) ISS also developed a standardized methodology for calculating stock option values, based on the FASB-preferred Hull-White lattice model that explicitly accounts for suboptimal employee exercise patterns, and option exercise and forfeitures due to pre- and post-vesting employee termination. ISS also developed standard model inputs: a uniform risk-free rate assumption and the consistent use of historical volatility with the same look-back period for all companies.

Initial analysis of these disclosures indicates that companies are consistently, and in some cases significantly, understating the expense of employee stock options, even under the new accounting regulations. Eighty-nine percent of the 36 companies reported option values that were lower than the adjusted value that was calculated under the Hull-White model. The average company's reported value was 29 percent lower than the adjusted value; when the handful of companies reporting higher option values are excluded, the average disclosed option value was 33 percent lower than the adjusted value. The size of the difference between the reported and adjusted values ranged from 84 percent higher than reported value all the way down to 7 percent lower than reported.

The financial impact of these differences is not trivial. For the 36 companies, ISS used option values derived from the Hull-White model and recalculated net income and earnings per share (EPS) to determine an adjusted EPS. ISS found that the average company's reported EPS is 2.3 percent higher than the adjusted EPS. Here, too, there is much variation, with differences ranging from 30 percent less to 5 percent more.

Model Choice and Model Inputs
Why are ISS' values higher than company-reported values? In aggregate, there weren't significant deviations between company volatility assumptions and historical volatility. In the ISS sample, the average volatility assumption made by companies is almost identical to the average historical volatility experienced, and companies were equally likely to estimate a higher-than-historical volatility (which would yield a higher option expense) as to estimate a lower-than-historical volatility. The average size of the difference between the estimated and historical volatility was relatively small in any event, at 3.62 percent.

What, then, is driving the difference between companies' reported and the ISS-adjusted option values? It comes down mainly to model choice. Recall that while the Hull-White model makes empirically verifiable assumptions for each the distinguishing features of employee stock options (exercise behavior, post-vesting termination, vesting periods), the Black-Scholes model requires that all these assumptions be lumped into an opaque, hard-to-estimate "expected term" assumption.

Much of the protest by companies in 2004 and 2005 against expensing stock options was directed at the Black-Scholes model, which in unmodified form significantly overstates the value of employee stock options. It might come as a bit of a surprise, then, to find that nearly three quarters of the 36 companies in the sample are still using Black-Scholes to calculate option expense. This choice is more understandable, because the model allows companies to adjust the expected term of the options to account for employee stock options' distinguishing characteristics (vesting periods, nontransferability, and suboptimal exercise behavior). The average company's expected term is 56 percent of the contractual term length, which pushes the option values down significantly.

It is entirely appropriate that companies adjust expected term length downward--they must reflect the reality of employee stock options. But investors should pay close attention to the scale of this adjustment; of the Black-Scholes input variables, term length has the greatest impact on the final value estimation, more so than volatility.

These results indicate that institutional investors must pay special attention to the series of assumptions that companies are making in their expected option-term length disclosure. Not only does this assumption propagate across other assumptions (volatility, risk-free rate, and dividend yield assumptions, for example), but the number itself is opaque, difficult to estimate directly, and not particularly amenable to historical estimation methods. Binomial and trinomial lattice models, such as the Hull-White model, offer far better transparency into the most important model inputs and their impact on option value.

This article was derived from an ISS white paper. For more information on this topic, go to the ISS Web site: http://www.issproxy.com/index.jsp.