Just enough capital, and not a dollar more

Before payments, I spent around two years as a Quantitative Strategist at Morgan Stanley, on a desk in the Fixed Income Division most people outside a bank have never heard of: Macro Valuation Adjustments. This is what that work is, and what I built there.

What a valuation adjustment is

A bank's balance sheet carries a number nobody trades and everybody watches: the capital it has to set aside to satisfy its regulators. A valuation adjustment is a piece of that: a reserve held against the firm's positions so that its books, and its capital, line up with what the rules require.

What makes it interesting is that it cuts both ways. Set it too low and you are in breach: you have not reserved enough, and the regulator has a problem with you. Set it too high and you have parked capital that could have been earning; every dollar over-reserved carries an opportunity cost. So the job is not "make the number big and safe." It is to make it exactly right: just enough to satisfy the rules, and not a dollar more.

Why it's hard

Three things, stacked on top of each other:

• It is not a single number. There are different kinds (a model VA, a portfolio VA, and others), and they vary again by the underlying product: a swap, an option and a swaption each carry their own. "The VA" is really a whole family of adjustments, each with its own methodology.
• The rules are not one set of rules. Different markets regulate differently, so the same adjustment has to be computed differently depending on where the desk sits: a customisation per trading location, not a single formula.
• Historically, a lot of it was done by hand. And hand-computed regulatory numbers go wrong. A mistake here is not a P&L error you book and move on from; it is a compliance failure, with the regulatory and reputational consequences that come with one.

What I built

I was on the Macro Valuation Adjustment team, and my job was to take these off manual footing, on an in-house, Scala-based graph-execution framework, with internal libraries layered on top. The shape of every one of them was the same once you saw it: read some data, run a set of complex calculations over it, and push the results into the downstream systems that book and report them. The plumbing was never the hard part; the methodologies were, the actual math of each adjustment, which differed by product and by desk.

I built and supported more than ten of these, across the FX desk in Brazil, the interest-rate desks in London, Hong Kong and New York, and the structured-notes desk in London and New York, each with its own method and its own jurisdiction's rules.

Automating them meant more than "run the calc." It meant a tool the people who owned these numbers could actually operate:

• Run it from a UI. The people who owned the number could operate it themselves, on demand.
• Validate the data before anything is pushed. A bad input is caught before it becomes a bad regulatory number.
• Send the figures to stakeholders. The relevant numbers go out, by email, to the wider set of people who need them.

The validation is the part that mattered most: the whole point was a number people could trust.

The payoff

Tighter controls, and roughly 25 engineer-hours a month handed back to the Valuation Control and Strats teams. But the hours saved were the least of it. The real win was moving a regulatory number off a manual process that could quietly go wrong and onto one that validated its own inputs and left a trail anyone could follow.

In a bank, the reliability of a number the regulator trusts matters more than the hours it saves.

A manager worth working for

The thing that has stayed with me most from Morgan Stanley isn't a valuation adjustment; it's my manager, Byron Warner: someone I could talk to about anything, work or not, and the kind of person you are simply glad to have in your corner. In my first-year review, he wrote that I was "rookie of the year."

I told him I wanted to do more than my role gave me, and what he did with that is what I remember. He asked me to move to a team that fit my goals better than his own, knowing it would leave him short-staffed and hurt in the near term. I only learned later how many strings he had pulled to make that move happen; he never mentioned any of it. A manager who will trade his team's short-run performance for what is right for you is rare.

He set the bar for what I expect from a manager, and gave me a great deal of confidence. I cherished working for him, and I'd do it again in a heartbeat.