You are currently browsing the tag archive for the ‘Goodhart’s law’ tag.
GDP is “the size of our economy,” the sum total of goods and services produced by “our economy” consolidated into a single dollar figure. In case you didn’t already know this, how big it is, and how fast its growing, is considered by many observers to be important. When it doesn’t seem to be growing as fast as it was in the past, people write books.
It’s also probably not surprising to anyone who reads this blog (BREAKING: it still exists) that how we measure GDP is, when you dig into it, at least a little weird. For starters, the border between what is and isn’t “the economy” as opposed to “stuff people do with their time” is a little fuzzy. Plus, stuff that seems to definitely be party of “the economy” is occasionally hard to measure. This ends up with the rent that homeowners don’t actually pay themselves but are “imputed” to pay themselves to work out to be 5-10% of all of GDP!
When you think about it, that isn’t wrong—we obviously spend a lot of money building houses, buying new and used houses, fixing broken houses, and renting houses we don’t own, in a way that suggests a) houses are definitely part of the economy and b) if we don’t measure the benefits homeowners get from owning very larger economic objects then GDP will look weird as a result. The folks who do this for a living explain it better and more thoroughly than I can.
But when you think about it, it does suggest that the legal and economic structure of relationships between people and institutions can matter a lot in deciding what does and doesn’t go into GDP. To wit, let’s visit a parallel universe, one where America is just the same as it is today except for one big difference.
In this universe, there is a very popular thing called The Netflix Organization that millions of Americans use to stream content. Everything physically and institutionally about The Netflix Organization and how people use it is the same as Netflix in our universe, but legally it’s structured just a little differently:
- The Netflix Organization is collectively owned by its members, not its shareholders.
- Its shareholders are actually all creditors who just have a set of unusually-structured debt contracts with the Netflix Organization.
- The monthly fees that owners pay to The Netflix Organization are actually collective contributions by the owners to
- pay organization staff and other costs;
- service debt payments (actual debt + dividends in our universe)
- cover depreciation; and
- engage in capital improvements (improving streaming performance and UIs, creating and buying rights to new content).
So here’s what would be weird about this universe—if you just left it here, GDP would be exactly the same as it is in our universe. But you wouldn’t just leave it here. Because the owners of Netflix don’t seem to be making any income! Yet they’re collectively paying billions every year to support this capital that they collectively own, which wouldn’t make any sense if it wasn’t producing economic value to its owners. So you need to impute the value of streaming Netflix content to users, and consider that economic output that would be added to GDP.
And depending on how you calculate it, that’s a lot of value. Netflix claims American users streamed 42.5 billion hours of content last year. How would you impute a dollar value onto that? Well, an average movie ticket last year cost $8.43. The average movie is around 2 hours; so you could impute a value of $4.215/hour streamed. But of course each hour streamed is probably viewed by more than one person; let’s just stipulate that the average works out to around two.
You would then impute a value of Netflix Organization income to its owners of $358 billion—which would add around 2% to 2015 GDP! And not only that—Netflix streaming is growing rapidly, from 29 billion hours in 2014. That figure would’ve only added 1.3-1.4% to 2014 GDP; put another way, the growth in Netflix streaming alone boosted nominal GDP growth by 0.6-0.7 percentage points last year.
Now, the assumptions I used to impute economic value to Netflix streaming are more than challengeable. But the point is that, in this parallel universe, you would very likely have to go through the exercise and impute something. We don’t do that in our universe because Netflix is considered to be selling a product to consumers, and therefore the product is automatically valued at the purchase price when it’s considered for addition to GDP. Which is fine! Fundamentally what I’ve done above is to rearrange a bunch of activity not considered economic in our current framework into something different purely on paper, with no real world change, and yet prompt a potential—and potentially large—reevaluation of our core economic metric.
The point of this exercise—and this post—isn’t that “GDP is bad” or “GDP isn’t accounting for disruptive tech, bro” or “the Lucas critique/Goodhart’s law/Cambell’s law,” although it necessarily includes a little of all those things. It’s mostly just that there’s an inherently arbitrary nature to measuring anything, and that if you want to measure something, you should probably measure it in a lot of different ways.