Here’s how two analysts uncovered conceptual risks – flaws in a firm’s assumptions and inferences about risk – that elude detection but may be as consequential as potential losses from external events.
“We have to get a better understanding of what they are doing and whether it’s working correctly,” said my boss’s boss, looking at me intently. “Starting today, that is your principal assignment, for as long as it takes.” The problem, he explained, was that the 1986 Federal tax reforms had created something called the Alternative Minimum Tax, or AMT, that penalized firms with tax-exempt income that exceeded a specified threshold. Insurance companies like ours were particularly affected, since we all typically invested heavily in tax-exempt securities. At our firm, as in many others, the accounting department had for years carried out an extensive (and expensive) twice-yearly survey of all business units, aggregating their earnings forecasts to obtain a firm-wide forecast. This information now enabled the tax department to determine the relative proportions of taxable and tax-exempt investment income needed to maximize after-tax earnings. Once these calculations had been completed, the tax department called the head of portfolio management – my boss’s boss – and essentially “advised” him to adjust his portfolios to achieve that optimal mix of taxable and tax-exempt investment income. “The specific problem,” he informed me, “is that no one in portfolio management understands the tax law, the tax department’s calculations, or even what questions to ask. Your job is to learn what we need to know.”
“Don’t worry about the politics,” he assured me. “I’ve told the head of the tax department that we need to understand this better, and he agrees. He has assigned their brightest young star to work with you on this assignment. You and he can start tomorrow.”
That night I did a quick review of the AMT’s major features and implications. Before it became law, taxes paid by insurers like us decreased in proportion to the income derived from tax-exempt securities. So insurers bought mostly tax-exempts and substantially reduced their taxes. The AMT essentially created a second tax that increased in proportion to tax-exempt income and other calculations, and required insurers to pay the higher of the old tax or the AMT. Post AMT, the graph of taxes paid on overall corporate income was V-shaped, where the lowest point in the V depended on an insurer’s mix of taxable and tax-exempt income, as well as on other less controllable factors that varied from year to year. Every insurer wanted to be at the lowest point in this V, where taxes were minimized and after-tax earnings were maximized.
The next morning I met with Blake, the tax expert selected to collaborate with me. He was friendly and incredibly smart. The first thing we did was to have me, under his expert guidance, create a spreadsheet with the complete sequence of calculations needed to determine the firm’s tax situation and the proportions of taxable and tax-exempt income needed to maximize after-tax earnings. The calculations weren’t tremendously complicated, but sufficiently lengthy for me to readily visualize the eyes of my boss’s boss and other investment colleagues glazing over as Blake and I tried to explain them. But solving that problem could wait until later.
Now that Blake had guided me through the tax calculations and answered my many questions, we needed to focus on understanding their investment implications. I scheduled us to meet with traders who specialized in tax-exempt securities. They first explained that there are two types of tax-exempt bonds. General obligation bonds are issued by entities like states and municipalities that can levy taxes and use tax revenues to pay bond interest and principal. By contrast, revenue bonds are issued by entities such as water and sewer plants, airports, and toll roads that lack the power to tax but rely on operating revenues to pay bondholders. Despite these differences, credit quality is rarely an issue with either type, since nearly all bond issuers pay bond insurers to guarantee the timely payment of principal and interest. The cost of obtaining these guarantees is more than offset by the resulting lower interest costs on the bonds being issued.
Next, the traders told us that at issuance tax-exempt bonds usually have very long maturities – typically thirty years. When purchased at issuance these bonds have higher yields than the after-tax yields on comparable corporate bonds. By contrast, the short and intermediate maturity tax-exempts available for purchase in the secondary market have yields that are only slightly greater than the after-tax yields on comparable taxable bonds. So most insurance companies had traditionally bought long-maturity tax-exempts at issuance and held them to maturity.
The new Alternative Minimum Tax appeared to make that traditional strategy obsolete, and seemed to favor a strategy of adjusting tax-exempt holdings and income each year to maximize a firm’s after-tax income, since the tax calculation included components that varied from year to year. Given this potential need to buy or sell tax-exempts to maximize earnings, we asked about trading. What we learned next was crucially important. Except for general obligation bonds issued by huge states and municipalities such as California and New York City, tax-exempts are relatively illiquid. After issuance they trade infrequently and in small lots, and have relatively high bid-ask spreads. This spread is the difference, at any given time, between the purchase price and the sale price of a particular bond. This means that if interest rates haven’t changed, when you sell a bond you get less than what you paid for it. And if you were to buy it back, you would pay more than its earlier sale price. The bid-ask spread is essentially an unavoidable trading commission for buying and selling bonds of any kind. Bonds are considered liquid if they have low bid-ask spreads, and illiquid if their spreads are high.
The crucial fact was that illiquid bonds comprised the bulk of tax-exempt holdings for insurers like us. This means that significantly adjusting our portfolios to supposedly maximize after-tax earnings could actually be quite costly. This was something neither Blake nor I nor anyone in the tax department had incorporated in our tax calculations. The real question, we now understood, was whether the estimated benefits of adjusting the portfolio’s tax-exempt holdings and income exceeded the transaction costs of making those adjustments. It was time for more calculations.
To make our calculations simpler and memorable, we focused on a change, from one year to the next, of a dollar of forecast underwriting income. We first calculated the benefit to the firm’s after-tax income of adjusting the portfolio to maximize after-tax income. Next we calculated the trading cost of making such adjustment. First, we found that for every dollar change in forecast underwriting income, optimizing after-tax earnings would require buying or selling approximately seven dollars of tax-exempt securities – provided the forecast was made in January. If the forecast change occurred mid-year, it required a purchase or sale of fourteen dollars of tax-exempts. Second, we multiplied these trading volumes by the trading costs supplied by our traders to obtain a total transaction cost per dollar change in forecast underwriting income. All these computations assumed the then-current investment environment.
When we then compared the benefits of adjusting the portfolio to the costs of doing so, we found a very slight benefit for adjusting the portfolio to a January forecast, but a substantial net cost of adapting to mid-year forecast results.
The heads of portfolio management and of the tax department proved receptive to our results. They particularly found our summary statistics (on the benefits and component costs of responding to forecast changes in underwriting income) to be extremely useful as a way of grasping and remembering the relevant implications of otherwise eye-glazing lengthy calculations. Memorable rules of thumb are wonderful aids to understanding!
But the high point of our presentation was something we had learned only a few days previously. Blake and I had been searching for the historical results of the twice-annual survey of forecast earnings, and discovered that no official records had been kept of the survey results. Then we accidentally encountered a long-time employee who had received these reports for many years and had been filing them away. So we borrowed his reports and compared the twice-yearly forecast changes in underwriting income to the actual changes in underwriting income for a dozen years. We found that forecast and actual changes had a correlation of zero. In fact, the best predictor of year-end underwriting income was the prior year’s underwriting income, a forecast that certainly didn’t require a costly internal survey.
The forecasts had no predictive value at all! Consequently, adjusting the portfolio to reflect forecast changes would create a net cost, since it would trigger costly trades with no predictable benefits.
This tale of quant detective work has a number of implications. I’ve already mentioned the value of rules of thumb in explaining and simplifying complex calculations. But another implication is much more important: the need to identify hidden hazards in a firm’s decision processes.
In this case there were two hidden hazards. First, the tax department assumed, without evidence or inquiry, that there were no costs to adjusting the investment portfolio to optimize after-tax income. Second, they and the accounting department both assumed that the very costly twice-yearly survey accurately forecast underwriting income, an assumption they had never checked for accuracy. Such hidden assumptions create invisible risks – hidden hazards – since acting on them can, if they are incorrect, as in this case, dramatically worsen financial results rather than improve them. I suspect that risks of this sort occur in the decision processes of many corporations, and will be tremendously magnified as firms become increasingly swayed by conclusions derived from “big data,” “data mining,” and similar widely touted analytic methods for finding ways to potentially increase earnings. Unfortunately, most corporate executives are ill prepared to challenge or test the results of these new methods, which are based on sophisticated calculations far more complex and obscure than those involved in this particular case.
Hidden hazards are risks created by serious errors, omissions, or distortions in the thought patterns and decision processes that firms rely on in making corporate decisions. One might reasonably call them cognitive risks, although they involve numerous individuals or departments – in this case, the tax, accounting, and investment departments, each with its own unique role. Cognitive risks are internal, and so differ in kind from the catastrophic external events that are the principal focus of executives, Chief Risk Officers, and risk managers. And they encompass the processes that lead to major strategic decisions that are far more consequential, and potentially perilous, than the routine administrative processes lumped into the vaguely defined category of operational risk. Identifying and correcting hidden hazards is more like detective work than like an actuarial quantification of frequency and severity.
I believe that these cognitive risks are more pervasive, more difficult to identify, and more consequential than the risks posed by external events or routine processes. Why? Because we, the decision makers, are enormously reluctant to scrutinize and test the validity of our thought patterns and decision processes, or to admit or even suspect that we may be wrong, even and especially when the stakes are considerable.
The most serious risks to our firms’ survival may turn out to be conceptual rather than physical – significant but unrecognized errors or gaps in the mental maps and models that all corporations and their leaders need and use to navigate their way forward in an uncertain world.
William H. Panning is Principal and Founder of ERMetrics, an Enterprise Risk Management Consultancy focused on rigorously modeling and maximizing a firm’s value.