Hacker News Reader

Comment by JumpCrisscross | Hacker News Reader

Not the Atlanta Fed, the Atlanta Fed’s GDPNow model [1].

It’s pretty good [2]. And we are seeing a flattening of 2024’s aggregate wage growth of 4.15%. But the difference in wages is like 0.42% which is indistinguishable from noise. (GDPNow predicted a phantom recession in 2022.)

In this case, the model is probably recording a surge in January imports without “an offsetting increase in inventories,” as “that is a lagging indicator” [3].

[1] https://www.atlantafed.org/-/media/documents/research/public...

[2] https://caia.org/blog/2024/08/15/increased-accuracy-gdp-mode...

[3] https://www.calculatedriskblog.com/2025/03/a-comment-on-gdpn...

View on HN

> GDPNow is an excellent tracking model, however, the January surge in imports - especially for gold - caused the model to move negative. As the Atlanta Fed noted: "the contribution of net exports to first-quarter real GDP growth fell from -0.41 percentage points to -3.70 percentage points".

Anyone know what's up with the surge of gold imports? We don't have gold tariffs, right? Are people hedging against the dollar's instability?

Drive by comment, I saw: https://unusualwhales.com/news/the-race-to-move-gold-bars-ou...

> The sudden shift is fueled by fears of tariffs as Donald Trump threatens a trade war that could disrupt the gold market. Traders, worried that gold imports may soon face tariffs, are rushing to get their bullion into the U.S. before the policy takes effect.

These models all seem like they are lagging. Are there any models that are more leading?

That's important because it's a pure mathematical model which may be wrong as it cannot take in all influences that drive it.

Ding ding ding

"All models are wrong, some are useful"; Why should I care what the ATL Fed's GDPNow model predicts? By all accounts a mathematical model is only as good as its assumptions and we know even the best economic models have spherical cow level assumptions

Mathematical models are so inflexible and unresponsive to real-world, commonsense facts, that as far as I know the value of what's in my pocket goes up at the grocery store. What we need is a grounded, grassroots approach to banking that eschews the elitism of sums involving numbers that, as far as anyone knows, are no more real than Plato's perfect spheres.

Some of these numbers are nearly impossible to measure because:

- tech gets better - products move upmarket as companies target richer customers - sometimes, products move downmarket and gets worse - products fall out of use and new ones take their place

With something like TV's, we can clearly see deflation at work. Your dollar buys vastly more TV than in 1990.

But what about education? That's nearly impossible to measure because we're learning different things. So if the price goes up by, say, 100%, how do we offset that against the nearly unmeasurable change in value?

You can see why economists resort to measuring Platonic, spherical eggs.

This is a GDP "nowcast," which only involves measurable numbers. The main criticism of GDP is that by focusing on the exchange of currency, it underestimates barter and social exchanges such as are prevalent in developing countries and Bethesda's Fallout series.

Fallout is getting more relevant every day, it seems.

If you're concerned by inflation in XP points earned doing post-apocalyptic errands for people, you've got bigger problems than a GDP decline.

> But what about education? That's nearly impossible to measure because we're learning different things.

At least at the college I attended, what you studied had zero correlation to the cost, every full-time student paid the “same” tuition as everyone else.

Without using any elitist numbers, can you express the odds that somebody will miss this parody?

I’m missing the source (and don’t know if I know it), but am curious :P

I don’t know if it is a quote or meme or anything like that, it might just be some bespoke hand-crafted parody.

If you have a better idea, a literal Nobel prize awaits.

Ah, but you're wrong. The "Nobel Prize" for Economics is actually not a literal Nobel Prize but one established by the Nobel committee after Alfred Nobel's death which has some distinctions making it different from the others.

But that's all just me being pedantic :)

https://en.wikipedia.org/wiki/Nobel_Memorial_Prize_in_Econom...

A literal one but perhaps not an original, authentic one :)

GDP itself is a mathematical model too - just a much more sophisticated one which utilizes more data.

> GDP itself is a mathematical model

GDP is a metric. The specific way in which we construct it is based on a small zoo of models, each working to estimate a myriad of parameters.

It's estimated via a model. We don't actually know the value of all economic activities happening in the country. We just have statisticians survey some fraction of businesses on how they're doing, and plug those result into a model to estimate how the entire country's doing.

> It's estimated via a model

So is every physical constant. The gravitational constant is a metric. We then measure it by way of models.

GDP isn’t a model per se. GDPNow is.

> it's a pure mathematical model which may be wrong

Sort of. It means it requires interpretation.

An altimeter is an analog mathematical transformer. It takes a trained pilot to properly read it in cold weather. Similarly, GDPNow is not a headline number, but a tracking model. Someone struggling with the definition of GDP isn’t its intended user.

No model can take in all influences that drive the economy. The more important aspect of it being a pure mathematical model is that there's no fudge factor for how much they like the current government. So it's not like the Doomsday Clock where the number is intentionally adjusted to drive the news cycle.

Rather, the GDP nowcast dropped because of the release of economic data that was historically associated with a drop in GDP. So if the model is wrong in this particular instance, there would have to be some reason why the historical association no longer holds.

Without knowing the detail, it's almost certainly a statistical model. Call it ML if you like. GDP nowcasting models take contamporaneously observable variables like foot traffic as features and GDP - which is observable only with a several-month lag - as the target.

The same caveats as with all statistical models apply. We only see the relationship as observed in the past and it only captures correlations (clearly, foot traffic in and of itself does not meaningfully cause GDP).

All models are wrong, but some are useful.

https://en.wikipedia.org/wiki/All_models_are_wrong

Reply on news.ycombinator.com