Recent studies by investment banks Credit Suisse First Boston and Bear Stearns found that return on pension plan assets boosted the financial performance of many large corporations. For fiscal 2000, almost half the S&P 500 companies with defined benefit pension plans reported pension income rather than pension expense, and in many cases this pension income contributed significantly to operating income. In Canada, we conducted a study of the 27 largest Canadian corporations (based on the value of their pension plan assets) and found similar results, with 41 per cent of our sample reporting pension income. However, these results, achieved during a bull market and buoyant economy, could easily be reversed in the coming years, particularly if the current bear market and tougher economic conditions prevail. Companies that based their expected return on pension assets on overly-optimistic assumptions are most likely to see future operating earnings dragged down by this reversal, an effect that investors may not have anticipated.

In the U.S., the studies found that aggregate operating income of the entire S&P 500 would have been approximately 3 per cent lower in 2000 if pension income had been excluded. Further, 35 of the companies reporting pension income (including IBM, DuPont, Boeing and U.S. Steel) obtained more than 10 percent of their operating earnings from pension income, even though in many cases the actual return on their pension plan assets was negative. In fact, for defence contractor Northrop Grumman, pension income reached 49 per cent of operating earnings. In our survey of Canadian companies, we found that 11 out of 27 companies reported pension income rather than pension expense, and for these companies, income represented an average of 5 per cent of operating earnings. Notable examples were: Maple Leaf Foods, with pension income accounting for 11.8 per cent of its operating income in fiscal 2000, Hudson’s Bay with 9.4 per cent, Dofasco with 8.6 per cent and BCE with 6.3 per cent.

These pension earnings do not legally belong to shareholders, nor do they contribute to a company’s cash flow. Rather they result from the application of complex accounting rules for defined-benefit plans. And while they can improve earnings, they are not reported as a separate line item on the income statement, but rather are typically used to offset other operating expenses. However, U.S. accounting rules, and similar rules introduced in Canada in 2000, require companies to disclose the components of pension expense (income) separately, in the notes to the financial statements.


Reporting pension income is permitted under Generally Accepted Accounting Principles (GAAP). In a defined-benefit pension plan, a company promises an employee a certain amount of pension benefits upon retirement, based on the number of years of service and the annual salary of that employee. While the employee is active, the company is responsible for contributing to a pool of assets from which these payments will eventually be drawn. What is included as pension expense on the income statement is not the annual cash contribution but rather an amount comprised of three key components:

  • Service cost: Corporations annually report a service cost to reflect an incremental year of service by their employees, which entitles the employees to more future benefit payments and increases the company’s pension obligation.
  • Interest cost: This reflects the growing time value of the future payments owed to employees.
  • Expected Return on Plan Assets (ROA): Companies carefully invest the pool of assets in a combination of equities and bonds. Some of these pools are tremendous —BCE has $14.3 billion in pension plan assets and Air Canada has over $9.1 billion (all figures in this article are in Canadian dollars). Companies hope the markets will provide returns that increase the size of the asset pool enough to meet future obligations to employees; if so, it relieves the corporation from contributing additional cash to the pool.

The above three components are combined to arrive at a pension expense or pension income that appears on the income statement. If the combined service cost and interest cost is larger than the expected ROA, the company will report a pension expense. If not, the company reports pension income.


The issue, however, is that while the service cost and interest cost are influenced by strict actuarial assumptions, the expected ROA is based entirely on management’s estimate of the long-run returns on their invested pool of assets. It does not depend on what the actual returns on the assets were for that year. For example, in 2001 Schneider Inc. reported pension income of $2.8 million, representing 7.5 per cent of its operating income. This pension income was driven by an expected ROA of 8.5 per cent, or $26.8 million, offsetting the combined service and interest costs of $24 million. Actual ROA, however, was only $8 million, or 2.6 per cent. Had the company been required to use actual ROA, Schneider would have reported a pension expense of $15.9 million. By holding its expected ROA steady at 8.5 per cent, Schneider is able to enjoy a continued increase in its operating earnings from its pension plan, despite the declining real-world performance of those assets.

Companies are permitted to use an expected long-run return on pension plan assets, rather than report the actual annual returns, because U.S. and Canadian accounting bodies believe it is important to “smooth” the effect of pension accounting on the income statement. Since actual market returns are quite volatile from year to year, reporting them would translate into volatile annual pension expense/income and cause companies’ operating results to fluctuate with the performance of the stock market.

Indeed, actual ROA has differed significantly from expected ROA in recent years. The Bear Stearns report indicates that in fiscal 2000, the median S&P 500 company had an expected ROA of 9.2 per cent (this compares to 8.01 per cent across the Canadian companies we surveyed), while the average actual return was only 5.0 per cent (7.24 per cent for the Canadian companies). In fiscal 2001, the gap will be at least as wide, since most companies left their expected ROA unchanged while the Dow Jones and TSE 300 ended down 6 per cent and 11 per cent, respectively, on the calendar year. Surprisingly, 63 companies in the S&P 500 raised their expected ROA for 2001, by an average of 0.44 per cent. Even worse, 34 of these companies did so despite the fact that their actual ROA in 2000 was lower than their expected ROA for the same year. In Canada, five major Canadian banks recently reported fiscal 2001 results, and as Exhibit 2 illustrates, the average ROA was 7.7 per cent while actual ROA dropped to -2.3 per cent. Only Bank of Montreal reduced its expected ROA (from 8.4 to 8.2 per cent) while its actual assets returned -5.3 per cent. And Bank of Nova Scotia actually raised its expected ROA from 7.5 to 8.0 per cent, after earning a strong 20.2 per cent on its assets in fiscal 2000. (This upward revision seems hasty given the low performance of 2.6 per cent in 2001.)

To evaluate how much a company’s operating earnings depend on its pension accounting, we analyzed the sensitivity of those operating earnings to a 1 per cent change in the expected ROA. As reported in Exhibit 1, among the Canadian companies surveyed, the average impact on operating earnings would be a 6.9 per cent decrease for every 1 per cent decrease in the expected ROA. However, for some companies with large pension plans and weaker operating performance, the effect is much more dramatic. For instance, Stelco’s operating earnings would drop a whopping 54.9 per cent for every 1 per cent decrease in expected ROA; Algoma’s would drop 36.0 per cent; Air Canada’s would plummet 23.8 per cent; and five other companies’ would fall more than 5 per cent. Financial statement users should also be aware that the effect works equally in the opposite direction—a simple increase in the expected ROA assumption dramatically improves some companies’ earnings.

Exhibit 1: The Impact of Pension Income (Expense)


A fundamental issue lies in the reasonableness of a company’s expected ROA assumption. In a recent interview with Fortune magazine, legendary investor Warren Buffett decried expected ROA assumptions as “heroic” given historical market returns and current government bond yields. Since most companies invest 30-50 per cent of their pension assets in high-grade bonds that currently yield about 5 per cent, generating an 8 per cent ROA on overall plan assets (the average expected ROA of the Canadian companies surveyed) implies they are expecting returns of over 10 per cent on the equity component of their plans, going forward. Many market experts would say this is an unreasonable expectation given the late-‘90s bull market run-up. Furthermore, it is questionable how certain companies can justify their current expected ROA, given their own historical assumptions. For instance, Imperial Oil had an expected ROA of just 6.0 per cent in 1982, when 10-year Canadian Government bonds were yielding 16.75 per cent. This means that Imperial Oil could simply have invested their entire pension asset portfolio in government bonds and guaranteed a return of 16.75 per cent. In 1989, when the same bonds were yielding 10.0 per cent, Imperial Oil had only raised its expected ROA to 8.5 per cent. Yet, for 2000, when these bonds were yielding only 6.5 per cent, Imperial Oil had an expected ROA of 10.0 per cent.

Another way to assess the reasonableness of a company’s ROA assumption is to examine the off-balance sheet account entitled “unamortized experience gain/loss.” Here a company is required to account for the cumulative difference between actual and expected ROA. Yet again, the accounting standards provide a “smoothing” mechanism. Companies are only required to recognize these losses as pension expense when they reach a “critical value” (10 per cent of the greater of: overall pension plan assets or the pension obligation). But even then, only the excess (above 10 per cent) is typically amortized over a period of 10-15 years, and recorded as part of pension expense. Thus, while companies can immediately improve operating earnings through increased pension income (or reduced pension expense) by their choice of expected ROA, the downside repercussions of an unrealistic choice (the eventual amortization of losses) lag years behind, and are fractional compared to the original improvement in earnings.

Growing cumulative “experience losses” over time suggests unrealistically high expectations for ROA. The Canadian data suggest these losses are just beginning to accumulate, in part because companies captured “experience gains” during the bull market of the late 1990s. Recent filings by the Canadian banks provide significant insight: While the five major banks had an aggregate “unamortized actuarial gain” (which includes these experience gains as well as changes in actuarial assumptions) of $2.46 billion at the end of 2000, this was dramatically reversed and became an aggregate loss of $1.65 billion by year-end 2001 (see Exhibit 2). Even on an individual basis, every single bank switched from a gain in 2000 to a loss in 2001. Three of the banks have already reached the “critical level” and will be forced to recognize a portion of these losses as an expense in 2002. If the banks are a good indicator, and if the stock market continues to falter, many Canadian companies will soon be reaching their “critical level” and be forced to expense a portion of their accumulated losses, and adjust their expected ROA downward to prevent this from continuing. Either of these actions will result in decreases in operating earnings.


Another important assumption of the defined-benefit pension plan is the rate at which future retirement benefit payments are discounted. While this rate affects the income statement through the service and interest cost components of the pension expense, it has a much more significant impact on the pension obligation, a balance sheet account. In 1992, the U.S. Department of Labor estimated that a 1 per cent decrease in the discount rate would increase this liability by 10 per cent. Under the new Canadian accounting rules, companies are required to use a rate reflecting the current yield on long-term high-quality corporate bonds. As interest rates have steadily dropped in both the U.S. and Canada, companies have been required to report increasing pension obligations and deterioration in the “funded status” of their defined-benefit pension plans. The funded status of a plan is the extent to which plan assets exceed the pension obligation.

For example, in 2001 Bank of Montreal dropped its discount rate from 8.1 to 6.7 per cent. As a result, its pension obligation grew from $2.2 billion in 2000 to $2.9 billion in 2001, and the plan went from being over-funded by $892 million in 2000 to being under-funded by $87 million in 2001. In fact, all five major banks reported a decline in their funded status, and in aggregate their over-funded status fell from $3.5 billion in 2000 to just $567 million in 2001 (see Exhibit 2). Both the poor returns on plan assets, as well as the lower discount rates, contributed to this drop. Such changes are perceived by analysts and creditors as a deterioration of the health of the plan, and could negatively impact future cash flows if companies are required to increase their contribution to the plan to eliminate its under-funded status.


The complex accounting rules for defined-benefit pension plans were crafted to provide detailed information about pension plans, and at the same time protect companies from excess volatility in their earnings. The result is a set of rules that are not widely understood and that permit companies considerable leeway. As Canadian companies release their annual financial results, it will be worthwhile to delve into the pension plan disclosures to understand how much of a company’s earnings are the result of optimistic expectations for the ROA on the pension assets, and to consider how long these effects can be sustained.

Exhibit 2: The Big 5 banks