With the yield curve deeply inverted, what's the bull case?
I recently published an idea where I laid out the bear case. But as I said in that post, although there are reasons for caution, I'm not really bearish on this market. In this post, I'll lay out why I do not believe now is the time to go full short.
The bear case in review
First, let's briefly review the bear case. First, the yield curve is deeply inverted. Historically, the time of greatest risk for stocks occurs as the curve uninverts. The market seems to expect the curve to uninvert over the next year. Second, energy prices and real weekly earnings have been climbing again, which suggests that inflation might not yet be dead. Third, there continue to be significant inflation risks connected to Russia, China, and the record temperatures and ice melt the world has seen this year. And fourth, global liquidity has been falling due to Central Bank balance sheet reduction and interest rate hikes.
And on top of all that, stock market valuations are really quite high. S&P 500 P/E is over 25. If I were blindly trading a macroeconomic model, I would probably be all cash or even short the market here.
So what could possibly be the bull case, then?
1. We may be on the cusp of the biggest productivity boom the world has ever seen
There are several reasons I'm not a bear right now, but the big one is automation and AI.
I know, I know; it feels like a bubble! It feels like hype! We get one of these every few years! We just went through one with crypto and NFTs!
But listen, I don't do hype. Except the occasional opportunistic scalp, I never bought into the crypto craze. I've never owned Nvidia or Tesla stock. So take me seriously when I say that when Dall-E 2 came out, I immediately recognized this was a different kind of thing. And when ChatGPT came out, I switched careers to work with these tools. I use them every single day, and they have already at least doubled my productivity. In my opinion, the AI field could stop making breakthroughs today, and we'd still see huge productivity gains for the next decade just from adoption of the tools we already have. And the breakthroughs haven't stopped. If anything, they've accelerated.
In my opinion, GPT-4 is already something very close to AGI. Its reasoning ability is astounding. Yes, you have to finesse the prompts a little to max out its reasoning ability. Yes, it hallucinates, and yes, OpenAI has nerfed it a little with all their context management and alignment work on the back end. Yes, it's hard to believe that reasoning ability emerges from next-word prediction. But the reasoning ability is undeniable when you see it solve novel problems before your eyes. As I've argued elsewhere, it appears that a "grammar of reasoning" is latent in the grammar of our language, and OpenAI has successfully created a statistical model of at least a portion of this grammar through matrix math.
The "BabyAGI" and "AutoGPT" efforts to bootstrap AGI with GPT-4 as an "LLM core" have so far been (mostly) unsuccessful. GPT-4 isn't good enough to get there with this paradigm. But a whole lot of programmers (including me) are working on AI agents that use an alternative "code core" paradigm. And in my opinion, GPT-4 is already powerful enough to get to AGI with this paradigm. It's just a matter of years. My prediction is that we will see AGI this decade. Maybe sooner rather than later.
Here's where I go out on a limb. I believe that we are on the cusp of the biggest productivity boom the world has ever seen. It's already started. In a quarter with very high interest rates, we just had a 3.7% productivity gain, smashing economists' expectations.
I don't think this will be the technological "singularity" that Kurzweil predicted, with exponential acceleration into an incomprehensible future. I don't think it will put all of humanity out of work overnight. But it's going to big. And it might very well put me out of work overnight, because I am a pure knowledge worker. Let's just say I'm not worried that productivity gains will fall short of my expectations. I'm way more afraid that they will exceed them, and that I'll be out of a job. And if that happens, then the only saving grace will be if I own a piece of the companies and technologies that put me out of work.
2. We've probably already beaten inflation
The rest of this post may be a bit of an anticlimax after that diatribe, but let's do it anyway. First of all, in my opinion, we've probably already beaten inflation. On an annualized basis, it's already back at the Fed's 2% target.
Yes, energy prices are rising again. That remains a pretty significant inflation risk. And we could also see food prices go up. But we may be on the verge of reversing some other key inflationary trends.
For one thing, the corollary of productivity gains is falling labor costs. With the higher-than-expected productivity gain last quarter, we had lower-than-expected unit labor costs (+1.6% vs. +2.5% forecast). This will be especially true for knowledge work. The cost of writing, marketing, and software services is going to plunge in the coming years. Service work and manual labor will become proportionally more valuable, so it's going to be a good decade to be a service worker. But with the labor force participation rate back to pre-pandemic levels, we're probably through the labor market crunch.
Next, let's look at shelter. We've had a decade of NIMBY derangement in US housing policy, with hardly any new multifamily housing being built. But the YIMBY movement is starting to change things. Housing supply reform bills have been passed in quite a few US states, including some of the worst offenders like Washington and Colorado.
These reforms are already paying dividends for US housing supply. There are almost a million new units of multifamily housing soon to come on the market, and over 500,000 new starts—the highest number since the 1980s. And with US population growth under 1%, there's not going to be a ton of new demand unless we loosen up immigration limits. (Admittedly, the declining marriage rate means you could see growth in the number of households even if there's no population growth, because you have to house more singles.) The payoff is that we should see real slowing or even reversal in inflation of shelter costs.
And finally, let's also consider supply chains.
3. We're about to have a US manufacturing boom
Another side-effect of AI (combined with dollar weakness and reshoring from China) is that the US is about to experience a manufacturing boom. The charts are genuinely incredible.
Highest factory spending since 1981! How often do you see this kind of capital goods investment after huge interest rate hikes? Mind you, most of this is electronics manufacturing, with US companies betting big on chip and GPU demand from crypto and AI.
4. Leading economic indicators look good
You probably don't want to bet against the US economy when the GDPNow forecast is at 4.1% and the ECRI Weekly Leading Index looks like this:
5. The credit and labor markets still look okay
One reason we've so far avoided a recession despite rising interest rates is that both consumers and S&P 500 firms are in pretty good shape credit-wise. Consumers have relatively little credit card debt relative to their income:
Plus, a whole lot of S&P 500 firms took advantage of low interest rates during the pandemic era to lock in a lot of debt at very low interest, so they're in no great immediate danger of having debt roll over at higher interest rates. Having said all that, my analysis of the credit situation could be wrong. Breakages in the financial system often come on unexpectedly. (I suspect, for instance, that consumer credit would look worse if you used the medians rather the means. I haven't done that analysis yet.)
It's also significant that, for the moment, continuing jobless claims are trending down, so the labor market looks okay. More than any other indicator, this is the one I'm watching as a recession signal. In particular, if Chris Moody's Ultimate Moving Average-Multi-TimeFrame indicator goes green on the monthly chart, I will consider that a strong signal of recession risk. This has correctly flagged the onset of recession for all recessions since the 1960s, with only one false positive. For the moment, though, we're still okay.
How to play it
So, there are a number of good reasons to be bullish! But how do we play it? Chase the Nasdaq and buy a lot of megacap tech, right? Right?
In my opinion, wrong. At 45 price-to-sales ratio for Nvidia, I'd say AI is already priced in for big tech. With all that semiconductor supply being built in the US, there's about to be a lot of competition in the sector anyway, so Nvidia might actually be a short. Besides, with AI, every company can afford to build proprietary software, so it erodes moats for all the big SaaS companies. Falling R&D costs will allow smaller firms to compete more effectively and build propreitary software. Every sector is the tech sector now. Wouldn't it be something if the biggest gains from AI came in sectors like utilities or food and beverages? I think that's the world we're headed towards. I think you buy the whole US market, and maybe the whole global market, and not just tech.
For me, the most difficult thing is to know how much to include fixed income in the mix. Over 4% yield on a 30-year Treasury at a time when I suspect inflation's been beat? That's hard to resist.
There may be a case for some kind of return stacking (i.e., leveraged Treasuries) ETF like NTSX. However, note that the Fed won't have a ton of reason to cut rates if growth is as strong as I think, so rates could very easily stay high. And with the yield curve inverted, leveraged strategies are expensive. Leverage costs are set by the short end of the yield curve, so if you leverage long-end Treasuries during a yield curve inversion then you're paying more for leverage than you're getting in yield. So I'm not sure what the right balance or the right use of leverage is, but I do think the traditional 60/40 allocation makes more sense now than it has in many years. This can help smooth out the volatility if we do see some of the recession warning signs play out.
The bear case in review
First, let's briefly review the bear case. First, the yield curve is deeply inverted. Historically, the time of greatest risk for stocks occurs as the curve uninverts. The market seems to expect the curve to uninvert over the next year. Second, energy prices and real weekly earnings have been climbing again, which suggests that inflation might not yet be dead. Third, there continue to be significant inflation risks connected to Russia, China, and the record temperatures and ice melt the world has seen this year. And fourth, global liquidity has been falling due to Central Bank balance sheet reduction and interest rate hikes.
And on top of all that, stock market valuations are really quite high. S&P 500 P/E is over 25. If I were blindly trading a macroeconomic model, I would probably be all cash or even short the market here.
So what could possibly be the bull case, then?
1. We may be on the cusp of the biggest productivity boom the world has ever seen
There are several reasons I'm not a bear right now, but the big one is automation and AI.
I know, I know; it feels like a bubble! It feels like hype! We get one of these every few years! We just went through one with crypto and NFTs!
But listen, I don't do hype. Except the occasional opportunistic scalp, I never bought into the crypto craze. I've never owned Nvidia or Tesla stock. So take me seriously when I say that when Dall-E 2 came out, I immediately recognized this was a different kind of thing. And when ChatGPT came out, I switched careers to work with these tools. I use them every single day, and they have already at least doubled my productivity. In my opinion, the AI field could stop making breakthroughs today, and we'd still see huge productivity gains for the next decade just from adoption of the tools we already have. And the breakthroughs haven't stopped. If anything, they've accelerated.
In my opinion, GPT-4 is already something very close to AGI. Its reasoning ability is astounding. Yes, you have to finesse the prompts a little to max out its reasoning ability. Yes, it hallucinates, and yes, OpenAI has nerfed it a little with all their context management and alignment work on the back end. Yes, it's hard to believe that reasoning ability emerges from next-word prediction. But the reasoning ability is undeniable when you see it solve novel problems before your eyes. As I've argued elsewhere, it appears that a "grammar of reasoning" is latent in the grammar of our language, and OpenAI has successfully created a statistical model of at least a portion of this grammar through matrix math.
The "BabyAGI" and "AutoGPT" efforts to bootstrap AGI with GPT-4 as an "LLM core" have so far been (mostly) unsuccessful. GPT-4 isn't good enough to get there with this paradigm. But a whole lot of programmers (including me) are working on AI agents that use an alternative "code core" paradigm. And in my opinion, GPT-4 is already powerful enough to get to AGI with this paradigm. It's just a matter of years. My prediction is that we will see AGI this decade. Maybe sooner rather than later.
Here's where I go out on a limb. I believe that we are on the cusp of the biggest productivity boom the world has ever seen. It's already started. In a quarter with very high interest rates, we just had a 3.7% productivity gain, smashing economists' expectations.
I don't think this will be the technological "singularity" that Kurzweil predicted, with exponential acceleration into an incomprehensible future. I don't think it will put all of humanity out of work overnight. But it's going to big. And it might very well put me out of work overnight, because I am a pure knowledge worker. Let's just say I'm not worried that productivity gains will fall short of my expectations. I'm way more afraid that they will exceed them, and that I'll be out of a job. And if that happens, then the only saving grace will be if I own a piece of the companies and technologies that put me out of work.
2. We've probably already beaten inflation
The rest of this post may be a bit of an anticlimax after that diatribe, but let's do it anyway. First of all, in my opinion, we've probably already beaten inflation. On an annualized basis, it's already back at the Fed's 2% target.
Yes, energy prices are rising again. That remains a pretty significant inflation risk. And we could also see food prices go up. But we may be on the verge of reversing some other key inflationary trends.
For one thing, the corollary of productivity gains is falling labor costs. With the higher-than-expected productivity gain last quarter, we had lower-than-expected unit labor costs (+1.6% vs. +2.5% forecast). This will be especially true for knowledge work. The cost of writing, marketing, and software services is going to plunge in the coming years. Service work and manual labor will become proportionally more valuable, so it's going to be a good decade to be a service worker. But with the labor force participation rate back to pre-pandemic levels, we're probably through the labor market crunch.
Next, let's look at shelter. We've had a decade of NIMBY derangement in US housing policy, with hardly any new multifamily housing being built. But the YIMBY movement is starting to change things. Housing supply reform bills have been passed in quite a few US states, including some of the worst offenders like Washington and Colorado.
These reforms are already paying dividends for US housing supply. There are almost a million new units of multifamily housing soon to come on the market, and over 500,000 new starts—the highest number since the 1980s. And with US population growth under 1%, there's not going to be a ton of new demand unless we loosen up immigration limits. (Admittedly, the declining marriage rate means you could see growth in the number of households even if there's no population growth, because you have to house more singles.) The payoff is that we should see real slowing or even reversal in inflation of shelter costs.
And finally, let's also consider supply chains.
3. We're about to have a US manufacturing boom
Another side-effect of AI (combined with dollar weakness and reshoring from China) is that the US is about to experience a manufacturing boom. The charts are genuinely incredible.
Highest factory spending since 1981! How often do you see this kind of capital goods investment after huge interest rate hikes? Mind you, most of this is electronics manufacturing, with US companies betting big on chip and GPU demand from crypto and AI.
4. Leading economic indicators look good
You probably don't want to bet against the US economy when the GDPNow forecast is at 4.1% and the ECRI Weekly Leading Index looks like this:
5. The credit and labor markets still look okay
One reason we've so far avoided a recession despite rising interest rates is that both consumers and S&P 500 firms are in pretty good shape credit-wise. Consumers have relatively little credit card debt relative to their income:
Plus, a whole lot of S&P 500 firms took advantage of low interest rates during the pandemic era to lock in a lot of debt at very low interest, so they're in no great immediate danger of having debt roll over at higher interest rates. Having said all that, my analysis of the credit situation could be wrong. Breakages in the financial system often come on unexpectedly. (I suspect, for instance, that consumer credit would look worse if you used the medians rather the means. I haven't done that analysis yet.)
It's also significant that, for the moment, continuing jobless claims are trending down, so the labor market looks okay. More than any other indicator, this is the one I'm watching as a recession signal. In particular, if Chris Moody's Ultimate Moving Average-Multi-TimeFrame indicator goes green on the monthly chart, I will consider that a strong signal of recession risk. This has correctly flagged the onset of recession for all recessions since the 1960s, with only one false positive. For the moment, though, we're still okay.
How to play it
So, there are a number of good reasons to be bullish! But how do we play it? Chase the Nasdaq and buy a lot of megacap tech, right? Right?
In my opinion, wrong. At 45 price-to-sales ratio for Nvidia, I'd say AI is already priced in for big tech. With all that semiconductor supply being built in the US, there's about to be a lot of competition in the sector anyway, so Nvidia might actually be a short. Besides, with AI, every company can afford to build proprietary software, so it erodes moats for all the big SaaS companies. Falling R&D costs will allow smaller firms to compete more effectively and build propreitary software. Every sector is the tech sector now. Wouldn't it be something if the biggest gains from AI came in sectors like utilities or food and beverages? I think that's the world we're headed towards. I think you buy the whole US market, and maybe the whole global market, and not just tech.
For me, the most difficult thing is to know how much to include fixed income in the mix. Over 4% yield on a 30-year Treasury at a time when I suspect inflation's been beat? That's hard to resist.
There may be a case for some kind of return stacking (i.e., leveraged Treasuries) ETF like NTSX. However, note that the Fed won't have a ton of reason to cut rates if growth is as strong as I think, so rates could very easily stay high. And with the yield curve inverted, leveraged strategies are expensive. Leverage costs are set by the short end of the yield curve, so if you leverage long-end Treasuries during a yield curve inversion then you're paying more for leverage than you're getting in yield. So I'm not sure what the right balance or the right use of leverage is, but I do think the traditional 60/40 allocation makes more sense now than it has in many years. This can help smooth out the volatility if we do see some of the recession warning signs play out.
Comment:
One thing I didn't mention in either of my posts was the risk that Chinese economic weakness might cascade into a global recession. Rumors are flying today that they're on the verge of another Evergrande-style debt collapse. This is where Treasuries can provide some good protection.