Passing the Test vs. Getting It

I remember this moment in JP Morgan's training class, when I realized I was lying to myself that I understood the concept of diversification. By then, I had passed the first level of the CFA examination and learned about Sharpe ratios, the Efficient Frontier and a bunch of other ostensibly more complex concepts. I could easily and quickly solve mathematical problems testing these concepts. But this should be a familiar feeling to most of us - one that accompanies being proficient enough in something to answer questions/solve problems/sound smart but knowing that knowledge rests on shaky foundations, built around a sneaky wager that no one will ask you four successive "Why's" when you confidently assert something.

Take my confusion about diversification for example which was really a gnawing sense that I didn’t really understand why people cared about risk-adjusted return/sharpe ratios:

"Diversification is good."

"Why?"

"Because it reduces portfolio volatility."

"But won't it also reduce expected return? After all, you're not putting everything into your single highest-return idea."

"Yes, diversification might lower your expected return somewhat, but typically it reduces volatility enough to compensate—meaning your risk-adjusted returns (like your Sharpe ratio) are actually higher."

"But why should we even care about volatility? If I'm investing for the long run, shouldn't I just pick whatever maximizes my expected return, regardless of fluctuations along the way?"

The answer:

Diversification works because it reduces the likelihood of experiencing severe, simultaneous drawdowns across your portfolio. If your investments move up and down together, you concentrate risk: the 'bad scenarios' in one asset become the bad scenarios in every asset. While this symmetry of large upsides and downsides doesn't necessarily affect your expected return (the average outcome across all possible scenarios), it does create scenarios with extreme outcomes—both good and bad.

The theoretical 'long-term'—in which volatility averages out—is largely an abstraction. In reality, no one truly invests in an infinitely long-term scenario; we invest with tangible life constraints and finite time horizons. Obligations like mortgage payments or living expenses don't disappear just because markets are down. Consequently, large simultaneous losses carry disproportionately severe real-world consequences, significantly increasing your risk of financial distress or ruin.

Meta-lesson

It could have been that I was too dense to have a clean picture in my head or that the lecturer was doing a bad job. But take a look at Investopedia's comprehensive piece on diversification. They cover correlation coefficients, standard deviations, and portfolio weightings in extensive detail. Yet, surprisingly, nowhere in this thorough explanation did they tie this back to why this matters to investors.

Needless to say, the people who wrote this probably know this deep in their bones, much more than I do. It’s likely because people who write finance books and teach seminars think it's too obvious to state. It's a breakdown in theory of mind. The second reason is that I was too lazy to think about it for more than a second, since I had no incentive to. If I could pass the CFA examination and get a job at JP Morgan without understanding diversification properly, that should worry everyone a little bit.

This phenomenon is everywhere, not just in finance. (in fact, finance has much better epistemic norms than other industries) In business environments, surface-level understanding often goes unchallenged - partly because that’s all you need to do your job nine out of ten times. Moreover, it selects for generalists and pragmatists that are in a rush to get things done, so it’s unlikely that groups will be particularly good about policing each other on this dimension. In fact, if you don’t understand something well, the preferred instinct is to be vague and high level, not press your colleagues on mechanics, lest you reveal your ignorance. Again, in most cases, someone with superficial knowledge navigates just fine through meetings and decisions. But then the status quo breaks down in some interesting way — and everyone who’s not a nerd is caught with their pants down.

Spending time around academics and the process of writing have been antidotes to this for me - helping rewire my brain after spending my early career around people who just didn’t care that much about what was true. (In fact, as if it to prove my point,

Rick

caught an error in my initial version of this post and pointed out that I’m confusing expected returns — which is a geometric mean — for an arithmetic average of annual returns. Damning for my my prospects of truly grasping diversification but proves my meta point about being forced to write and unpack your conceptual model). This is why I'm a fan of the written memo culture at Amazon . You can conceal a lot with decks and presentations, but in my company, I want to reserve the right to be annoying and ask as many “Why’s” as I’d like and force simple explanations in english, not formulae or slides.

If you’re hiring someone who will exercise significant agency and have significant leverage within your organization, it’s worth probing the foundations of their model of what it is that their job is trying to optimize and why - because it’s pretty easy for smart people to reference common knowledge and talk about trends and market dynamics with lucidity, while still missing the core intuition from which everything flows. More importantly, whether they are right or wrong, you’ll know if your inquiry is met with intellectual curiosity and course-correction or defensive evasion and doubling down on vague generalities.

For example, if I were an early stage VC fund hiring an associate, I’d consider this for an interview question:

Interviewer: “You know what power laws in VC are?"

Candidate: "Yes, they're very important - one massive winner can make up for all the losers, so really, focus on the winners."

Interviewer: “Okay, but since everyone is doing this model, why shouldn’t we try a different model - one where we pick ten moderate winners instead of one big winner?”

The answer is stupidly simple if you understand the fundamental intuition behind VC investing and impossible to answer if you don’t. In fact, even LLMs will give you only the superficial answer until you press them.