The latest drama: Complexity, predictions, and proxies
Not about the tariffs, just about the arguments
Did you get caught up in the tariff drama this month? How much time did you spend on it? Did it affect your work, sleep, or general wellbeing? As much as I try to avoid the news, I got pulled in too, and would like to share a couple of thoughts.
To be clear, I don’t know whether the current tariff situation is net good or net bad, but I am pretty confident that:
No one knows what will happen in a complex adaptive system like the global economy or international politics
Short-term stock market performance is a terrible proxy for whether a policy is good or bad
I saw a lot of arguments from generally thoughtful people fail the above two points. Per usual, my spiel is to be a little more skeptical of commonly held beliefs.
1. Complex adaptive systems
Many people have made bold, definitive predictions — in favor of, and against — about what will happen as a result of the tariffs. Here’s a good rule of thumb:
Anyone who claims to know what will happen in a complex adaptive system is probably wrong
This is because complex adaptive systems have:
Non-linear reactions: A 1% rate cut might barely change borrowing, while the next 1% unlocks a wave of high-risk lending
Feedback loops: A tariff raises steel prices → automakers raise prices → wage demand rises → input costs rise again, feeding the cycle
Causal opacity: A shutdown of chip factories in Asia can end up causing the Fed to hike rates (shutdown → global car shortage → increased used car prices → increased inflation → Fed policy)
These attributes combine in ways that can’t be reliably predicted. The 2008 financial crisis is a great example. Sub-prime mortgage defaults rose from 7% to just 9%, yet AAA-rated investments lost 90% of their value, triggering a global crisis1.
2. Short-term stock market performance
I also heard a lot of arguments like:
“The markets went down, therefore ABC tariff decision MUST have been bad”
To believe that short-term stock market performance can tell us whether a large policy decision was good or not requires some extreme assumptions, like:
“Investors accurately predict the future.” I think we sufficiently debunked this one in my point about complex adaptive systems. But consider just how wrong the stock market was before the dot-com bubble popped in 2000.
“Investors’ interests are aligned with what is good for society.” Investors care about their return on investment over their time horizon. Policies often help them without helping society and vice versa. Look at the corporate lobbying industry.
“The stock market should always go up.” It is healthy for the stock market to have periods of recession. Unnaturally suppressing downturns disrupts the natural economic cycle.
To bring that third point to life, consider how catastrophic the California wildfires have been over the past decade. A century of aggressive fire suppression has allowed debris and fuel to build up, resulting in less frequent but more uncontrollable megafires. Fuel builds up in the economy too:
Banks that are too big to fail2
A national debt of $36T+3
Credit card balances of $1.2T+4
Equity valuations at 2x historical average5
The more we suppress downturns with policy interventions, the more fuel accumulates. Better to allow for less severe but more frequent controlled burns.
All that to say, don’t use short-term stock market performance as a proxy for whether a policy is good or bad.
What to do about it
Don’t get caught up in bold predictions. The most significant outcome of a shock to a complex adaptive system will probably be unforeseen. It’s OK to not always know what is going to happen.
Be skeptical of proxies. Just because something is easy to measure doesn’t mean it’s a useful measurement. Try to falsify the goodness of the proxy. Maybe ask your favorite LLM, “Why might XYZ not be a good proxy for ABC?”
Control the controllable. Living within your means is generally a good way to lower anxiety about the economy. It’s a great asymmetric bet: a small lifestyle adjustment can often protect you from significant downside risk.
App idea: Automatically log predictions, verify them when the appropriate time comes, and report back to everyone who heard the original prediction. Not because all predictions need to be right but because without this, there may be more incentive for people to try to sound smart than to be right, even if it harms others.
https://www.congress.gov/crs-product/IF12755
https://fiscaldata.treasury.gov/datasets/debt-to-the-penny/debt-to-the-penny
https://resources.newyorkfed.org/householdcredit/hhdc-iframe
https://www.gurufocus.com/economic_indicators/56/sp-500-shiller-cape-ratio