For all that, permit yourself a sigh of satisfaction—but not much more. Wrapping up college days, I recall dropping off a final paper—the final paper, in fact—with a teaching assistant. I let loose a theatrical sigh of completion; college was really done. The TA snickered. “Your parents are allowed that sigh,” he said.
“You haven’t finished anything.” Two months later my sighs were more of the anguished variety as I scanned the employment pages and fretted landing an interview.
Your handful of models is really much like that sheepskin: (well done, by the way, for both, but . . .) only a start. Numbers assembly is merely the first step in modeling. Once the basic framework is constructed, the real work—calibration— begins. Eventually, what you are modeling is less the number itself and more the degree to which your process derives a value that varies from the number.
Once you determine consistency for that variance, you’re positioned to make the necessary adjustments. Now multiply this task by the line items in the individual model, and multiply again by the number of individual models. The finished model finds you not at the finish but at the starting line.
The role of modeling in asset analysis is significant, but it is far from the only element. When new analysts start at our company Argus, we tell them that the asset analysis process has four broad buckets: (1) financial statement modeling, (2) valuation analysis, (2) company knowledge, and (4) industry knowledge.
Yet even that represents only a few tools in the tool kit. Beyond individual asset analysis lies the interrelations of assets: the balance of buy, sell, and hold ratings for the analyst; the asset management process for the portfolio manager. Simultaneously, investment professionals are charged with interacting with clients, a healthy dose of marketing, and maneuvering through the office politics and the back-office minutiae that never make it onto the job description but somehow eat up big chunks of the day.
In Michael Pollan’s excellent The Omnivore’s Dilemma (Penguin, 2007), the author introduced a lay audience to the concept of the Holon (which Wikipedia attributes to Arthur Koestler). The Holon is something that is a complete and integrated system in its own right yet simultaneously a subsystem or component of a greater whole. Koestler referred to Holons as autonomous, self-reliant parts. We’ve tried to approach modeling with the goal of creating a self-contained system, a “stable form able to withstand disturbances,” but always within the knowledge that a model is an intermediate form contributing to “the proper functionality for the larger role” —that is, the analyst’s role.
Throughout this process, we’ve been putting numbers on just about everything.
So, here’s the final question: what percentage of the analyst’s job is modeling? For once, we defer. The financial services industry is simply too open ended for any one answer to suffice. I know a hedge fund trader who plays, not the bounce on the news, but the next day’s rebound off the bounce; that’s all he trades. I interviewed a prospective analyst who in his current job started each day in cash, traded equities all day, and ended in cash; worn out by his daily grind, he was 31 years old. I also know portfolio managers who change their holdings much less frequently than Standard & Poor’s changes the constituents in the S&P 500, and others who still make pencil marks on charts.
Given the changes in the financial service industry in recent years, including the increase in high-velocity program trading and quant strategies based on complex algorithms, the meticulous modeler can feel a bit like Bartleby Scrivener, dipping his quill in the inkwell while computers whir in the background. The inkwell set may have felt some malicious glee when the quant “rocket scientists” drove their collateralized rockets straight into the mountainside—“without letting off the throttle,” as one wistful PM said to me—in the summer and fall of 2008. Such smugness is out of place, as old-fashioned financial managers cannot point to much better performance in that historically bad period.
The nature of the game has changed, and a mere 58 percent market decline is no more likely to dislodge growing reliance on computer-driven trading and quant strategies than the slide rule is likely to take back the desktop from the personal computer. For all that, meticulous modeling is not just vital to the market; we’d argue that it is secure in the market.
Most quant strategies have an exhaustive backlog of data but only a wispy forward element. Dig through the algorithm for that forward element and you’ll find the consensus—which even now is built on individually modeled expectations.
There is the risk that cost cutting could squeeze the last few humans out of the process, and that digital trend compilation will replace the necessarily subjective mix of hard modeling and industry assessment that informs the analysis process. Should the outlook become purely dependent on machine-generated trend analysis, then no mountainside may be safe.
Our outlook is not so dire, if only because the bookends of the industry— greed and fear—need to find, respectively, confirmation and succor in a human face. The financial services industry—with trillions of dollars at stake, and even now with hundreds of thousands of employees—will remain a multifaceted world with a wealth of styles, approaches, theories, and gimmicks. Even though the financial data stream is now a binary blur, we think the industry will always find a place for those with a feel for the numbers.