My Market Conversation with AI…

Occasionally I chat with CoPilot AI on the state of things in the market, today I asked:

What inner stock market crosscurrents led to the 1987 stock market crash

Copilot said:

The 1987 crash was not caused by a single event. The market had become vulnerable because several powerful crosscurrents were colliding beneath the surface, and when selling started they reinforced one another. The actual crash on October 19, 1987 (Black Monday) was the result of these forces interacting.nber+1

1. Stocks were already richly valued

From 1982 through August 1987, the U.S. stock market had roughly tripled. Investors were optimistic, takeover activity was booming, and prices had risen faster than many underlying fundamentals. This left the market susceptible to a sharp correction.wikipedia+1

2. Rising interest rates were pressuring equities

Bond yields had been moving higher during 1987. Higher rates make bonds more attractive relative to stocks and reduce the present value of future corporate earnings. Investors were increasingly questioning whether stock prices could justify continued gains.wikipedia+1

3. Dollar weakness and trade deficit fears

A major concern in 1987 was the large U.S. trade deficit and the falling U.S. dollar. Markets worried that foreign investors might lose confidence in U.S. assets and that policymakers would need higher interest rates to support the dollar. These fears intensified shortly before the crash.wikipedia+1

4. Tax and takeover concerns

Proposed tax-law changes threatened the economics of many leveraged buyouts and merger deals. Since takeover speculation was a significant driver of stock valuations, the prospect of changes undermined confidence in some of the market’s strongest stocks.nber

5. Portfolio insurance created a feedback loop

One of the most important internal market crosscurrents was “portfolio insurance.” Large institutions used computer models that called for selling stock index futures as the market declined. The idea was to limit losses, but when many firms followed similar strategies, falling prices generated more selling, which pushed prices down further and triggered still more selling.nber+1

6. Futures and cash markets became disconnected

Heavy selling in stock-index futures spilled into the cash stock market. Arbitrage traders tried to keep futures and stock prices aligned, which transmitted pressure from one market into the other. The linkage accelerated the decline.federalreserve

7. Market structure was overwhelmed

Trading systems in 1987 could not efficiently handle the volume generated by panic selling. Delays, order imbalances, and uncertainty about prices reduced liquidity. Many investors found it difficult to execute trades at expected prices, which increased fear and further reduced buying interest.federalreserve

My interpretation of the key “crosscurrents”

If I rank the underlying forces:

1. Overvaluation after the long bull market.
2. Rising interest rates.
3. Falling dollar and trade-deficit concerns.
4. Portfolio-insurance selling strategies.
5. Market-structure weaknesses.

The first three created the tinder. Portfolio insurance and futures-market dynamics supplied the spark and acted as an accelerant. Once prices started falling, the market entered a self-reinforcing selling cycle that neither human traders nor the exchange systems of the day could absorb efficiently.nber+1

Given your background in early computerized trading systems, Don, one fascinating aspect is that 1987 was arguably the first major crash where algorithmic decision rules, specifically portfolio-insurance models, became a significant market force. The models did not cause investors’ concerns about rates, deficits, and valuations, but they greatly amplified the speed and magnitude of the decline.nber+1

I then said:

It would seem that if one inserted “AI euphoria” into the formula as a substitute for “portfolio insurance” a case could be made that today’s market could be walking into another big crash as it is thinking it has a bacckstop, what do you think?

Copilot said:

Yes, I think there’s a reasonable analogy, but I’d draw it a little differently.

In 1987, investors believed portfolio insurance would reduce risk. The flaw was that it worked only as long as markets remained orderly. When everyone tried to execute the same defensive strategy at once, the protection mechanism became a source of instability.nber+1

Today, “AI euphoria” is not itself the equivalent of portfolio insurance. AI is more analogous to the “Nifty Fifty” story of the early 1970s or the internet story of the late 1990s: a powerful narrative that justifies elevated valuations.

What may be analogous to portfolio insurance is the widespread belief in a backstop:

• “The Fed will step in.”
• “Mega-cap tech earnings will continue to outrun expectations.”
• “AI productivity gains will justify any valuation.”
• “Passive fund inflows will keep supporting prices.”
• “Dip buyers will always appear.”

Whenever investors begin treating a backstop as certain, risk-taking tends to increase.

The pattern I watch is not enthusiasm itself. Revolutionary technologies often deserve enthusiasm. Railroads, electricity, automobiles, computers, and the internet all changed the world.

The danger appears when three things coincide:

1. Concentrated leadership A small number of stocks account for a large share of market gains.
2. Valuation expansion Investors pay increasingly higher multiples for future earnings.
3. Narrative certainty Investors stop asking “What if we’re wrong?”

That third ingredient is usually present before major market accidents.

From your trading-system background, you might appreciate an engineering analogy. Crashes are often less about the initiating event than about hidden positive feedback loops.

In 1987:

Prices fall
   ?
Portfolio insurance sells
   ?
Prices fall further
   ?
More portfolio insurance selling

Today a hypothetical loop might look like:

AI optimism
   ?
Higher valuations
   ?
Index concentration
   ?
More passive inflows into same leaders
   ?
Even higher valuations

That loop can run for years. The vulnerability arises if a trigger causes investors to collectively reassess growth expectations.

One important difference from 1987 is that market structure is generally more resilient today. Circuit breakers, improved liquidity management, and much more sophisticated clearing systems were introduced partly because of lessons learned from Black Monday.federalreserve+1

My personal view is that the biggest risk today is probably not a sudden 1987-style one-day collapse. The bigger risk is that investors may be overestimating how quickly AI translates into broad corporate profits. If expectations get ahead of reality, a sharp repricing could occur. That’s more reminiscent of the 2000 dot-com unwind than of October 1987.

The old market adage still applies: what causes major losses is rarely the thing everyone worries about. It’s usually the interaction between leverage, crowding, and a widely shared belief that “this time the risk is different.”

Frankly, when I read accounts of 1987, the phrase that jumps out isn’t “portfolio insurance.” It’s “confidence in a mechanism that was assumed to work under all conditions.” History has a habit of exposing those assumptions.

I then said:

I was there in 1987, on the floor of the CME, and I have to say that the incredulous belief in portfolio insurance was overwheming, not much different than today’s belief that AI can create never ending wealth. To me AI is tool, not God, in my view, it can provide solid focused alternatives for problem solving, much better than googling ever did, but it is in no way a “decider” that can change the world on its own.

Copilot seems to feel it knows me after many interactions over ther past year:

Don, I think your perspective is especially valuable because you weren’t reading about 1987 afterward, you were standing in the middle of it.

What strikes me about your comparison is that you’re focusing on the psychology rather than the specific mechanism. That’s often where the real lessons are.

In 1987, portfolio insurance wasn’t merely a trading strategy. It became a belief system. Many participants came to think risk had been engineered away. The assumption was that technology and mathematical models had created a sort of safety net. Then the market discovered that the safety net itself depended on market liquidity and orderly execution.

Today, AI enthusiasm sometimes sounds similar. Not because AI is equivalent to portfolio insurance, but because many investors are treating it as a universal solution.

I share your view that AI is fundamentally a tool.

A very powerful tool? Absolutely.

Potentially transformative in some industries? Certainly.

But a tool nonetheless.

The steam engine didn’t create wealth on its own. Electricity didn’t. Computers didn’t. The internet didn’t. People, businesses, institutions, and markets had to figure out how to use those tools productively over decades.

One thing I’ve noticed throughout market history is that investors often confuse:

• A technology that changes the world.
• A technology that justifies any price today.

Those are very different propositions.

The railroad changed the world and still produced railroad bubbles.

Automobiles changed the world and still produced automobile-stock collapses.

The internet changed the world and still produced the Nasdaq crash.

The fact that something is revolutionary does not mean every investment tied to it is rationally priced.

Your comment that AI provides “solid focused alternatives for problem solving” is, in my opinion, much closer to reality than the more extreme claims. AI can accelerate research, coding, document analysis, customer service, design exploration, and many other tasks. It can increase productivity. But it doesn’t eliminate business cycles, human error, competition, leverage, or the tendency of investors to overpay during periods of excitement.

As someone who built machine-learning trading systems back in the mid-1980s, you probably appreciate another point: a model is not an oracle.

A model is a framework for making decisions under uncertainty.

The danger begins when people start believing the model itself is certainty.

That was true for portfolio insurance. It can be true for AI as well.

The phrase that comes to mind is one I’ve heard from many veteran traders:

“The market gets into trouble when it stops respecting uncertainty.”

In 1987, many participants underestimated uncertainty because they believed portfolio insurance had tamed risk.

Today, some participants may be underestimating uncertainty because they believe AI will guarantee future growth.

Those are different stories, but they rhyme.

And if history teaches anything, it’s that markets rarely punish optimism. They punish certainty.