On some days (a lot of days, including the first day of this week) Nvidia monopolizes financial media coverage. It beckons seductively to journalists, to say nothing of investors.
Jensen Huang’s turned everyone into groupies. Like Rick James, he gestures hypnotically through a neon haze. (“I seen like, a green — his aura or whatever. I seen it. It was green.”)
But if you look past Nvidia, there’s an entire stock market back there somewhere, and behind that, an entire economy. All of that presumably matters somehow, although it’s admittedly difficult to say just how. At best, the stock market bears only a glancing resemblance to the real economy, and Nvidia single-handedly accounts for a quarter of global equity performance in 2024.
To the extent you think the rest of it (i.e., the news ex-Nvidia) still matters, it’s worth noting that US economic surprises have rolled over definitively even if the economy itself hasn’t.
I don’t care much for “surprise” indices, but the “slippage” illustrated above is meaningful. And ongoing.
“The magnitude of downside surprises across the data are the largest since March of 2019,” Nomura’s Charlie McElligott remarked. That’s “iterat[ing] some of the slowdown / normalizing vibe.”
Bonds have noticed, paring losses to nearly pull even for 2024, despite looming questions around fiscal sustainability and political dysfunction.
“Yes, a few high-profile US data points have come in stronger than expected, such as May’s payroll data and June’s PMIs, but the vast slew of data have ‘missed’ in recent months,” SocGen’s Albert Edwards wrote Tuesday. “And while the ‘web’ has been very excited by Citi’s Economic Surprise Index, somewhat below the radar the precipitate fall in the Bloomberg measure of Economic Surprises is much more extreme,” he added, referencing the measure illustrated above.
Albert also pointed to the Chicago Fed’s National Activity Index. I don’t consult it myself, but as a weighted average of seven dozen monthly indicators, it’s ostensibly useful. A flat reading indicates trend growth. Negative 0.7 would be a recession.
As the figure shows, the CFNAI and GDP have decoupled. If you ask Edwards, that’s a bad omen. Or at least it could be.
Of course, Albert could show you a recession harbinger in a coffee table ring. Or in an oily puddle. Or in a cumulus cloud formation. And when he says, as he did on Tuesday, that a given divergence is “unprecedented” outside of the GFC, you shouldn’t take it literally (the chart plainly shows GDP diverged from the CFNAI consistently for half a decade from 1995 through 2000, a period many economists suggest is a good analogue for the current environment).
But he’s correct to suggest the disparity illustrated above may “call into question the supposed strength of the US economy,” particularly when considered with the succession and scope of recent data misses.
If you wanted to — which is to say if you were me, and you were looking for an article hook — you might ask “Who’s right?” The buoyant NFP headline and cap-weighted equity benchmarks (which, notwithstanding Nvidia’s three-day, $430 billion rout, remain perched near record highs), or the rest of the data and various equal-weighted equity market gauges, which are dramatically underperforming.
And what of the signal from bonds? As Edwards wrote, “the US bond market has woken up and smelt the weakness in the coffee.” Check that coffee table ring. It looks pretty recessionary to me.
The economic surprise index is more informative than a lot of people think. Intuitively, the idea that economists are often wrong doesn’t tell you anything new. But the index is helpful because of a simple truism for almost every economic data point you can care to imagine: the number one predictor of next month’s data point is this month’s data point. The number two predictor is last month’s.
As a consequence, every forecasting shop out there–with their big fancy models built by Ph.D.s using dozens of inputs, gathering their own real-world real-time data, conducting their own surveys–can’t help but overweight the most recent data. As they should! It’s the single best predictor! (You could prop up an economic forecasting shop that just took the last few prints of every data point out there, slapped on a trend factor, and call that your forecast. You’d be close enough to right to be mistaken for an economist).
In consequence, when the underlying state of the real economy changes, all forecasters are wrong all at once and in the same direction. If it’s just noise, the effect doesn’t last, but when there’s been a real change, the error persists. With a persistent, significant surprise index, the economy is screaming, “Something has changed!”
I spent some time looking at a chart of one of the surprise indicies (Citigroup), 2004-present. Two things jumped out:
– The index oscillates in a pretty constant range (-80 to +80) with a period of about 1 year. It isn’t perfectly periodic or consistently seasonal, but in almost every year the index will hit the top of the range, decline to the bottom, and rise again.
– The index correlates directionally with change in Treasury yield. When the index is rising [falling], 10Y yield is usually rising [falling]. However, the relative magnitude of moves changes a lot – i.e. the sensitivity of yield change to index change is not constant.
That’s it.
An indicator that regularly makes highs and lows every year seems not very informative. Unless you’re a Treasury trader, maybe.
It seems to me that surprise indicies merely reflect forecasters’ over-extrapolation whip-sawing around short-term fluctuations around a trend, but are too noisy to reveal much about whether the trend is changing.
JL – nice work. It’s no surprise that I agree with your conclusion.
This basically tracks with what I was saying. Things change, then after a few months, models adjust to the “new normal,” potentially followed by an over-correction. Fundamentals shift at a pretty glacial pace, so an oscillation per year sounds about right.
That it tracks treasury yield confirms it. Since the 10y is a pretty good proxy for the bond market’s expectations for how the economy is developing. If the economic surprise index is negative, then the zeitgeist regarding the state of the economy is negative, so 10Y yield would be expected to fall.