You Cannot Inflate the Debt Away
There is a popular meme that the government has an incentive to inflate, because the same debt is worth less in real terms at a higher price level. If inflation is high enough, then the government (it is said) can make the debt go away. Inflation is, after all, a tax; doesn’t it then follow that if the government increases that tax quite a bit then it can get itself out of hock in real terms?
It turns out that this is difficult to do, at least over a range of ‘normal’ inflation rates. The reason is that it is hard to get the debt to go away fast enough when the market adjusts interest rates to reflect the level of inflation. I want to try and illustrate the general idea here. I’ve looked at this before, so I know the answer, but I’ve never put this in a blog post before.
First, let’s start with the current distribution of debt for the US. The chart below is from Bloomberg, showing how many trillions of Treasury bill/note/bond maturities will happen every year. For a host of reasons, prime among them being incompetence, the US did not take advantage of the artificially ultra-low interest rate environment to extend maturities. That makes inflating the debt away even more difficult.
The average interest rate on US debt is 3.21%, about double what it was in early 2022 and the highest it has been since 2009 (and there is a lot more debt outstanding now than there was then, thanks to a decade and a half of massive deficits designed to save the world). Having a lot of debt mature every few years is a great way to increase the sensitivity of your interest expense, and therefore deficit, to short-term inflation and interest rate fluctuations. That’s not feature, it’s a bug!
Starting from the distribution above, let’s model how the debt would grow over time using a couple of simple assumptions and then tweaking those assumptions. We will start with this: assume that the maturing debt, plus the annual Treasury deficit, is rolled into new debt distributed 1/30th to each year. We will assume a 4% inflation rate, and an interest rate of 1% over the inflation rate. Because we are spreading out the maturities, we are explicitly extending the maturity of the debt. The more rapidly the average interest rate reacts to the current interest rate, the harder it is to inflate the debt away. If we started with that 3.22% interest rate and rolled the debt as I have just described, here is how the average nominal interest rate evolves over the next 25 years.
I should note that I am also assuming that the deficit is initially $1 Trillion, and grows by the rate of inflation every year. So, in year 0 it is $1T; in year 1 it is 1.04T, which is the same amount in real terms. Note that this implies that discretionary spending is decreasing rapidly. In year 0, the Treasury would pay interest of about $876 billion ($27.2 trillion in debt * 3.22% interest rate). In year 1, the average interest rate rises to 3.80% and the deficit is $1T larger, so the interest paid rises to $1072 billion. Since we assumed that the overall deficit increased by only $40bln, the implication is that discretionary spending fell $156bln ($1072 – $876 – $40). So, the notion that the deficit grows only equally with inflation is not realistic. But even using such magical thinking, the total debt rises from $27.2T to $72.5T over 25 years. The real debt (remember, we’re inflating it away!) stays basically steady at $27.2T in today’s dollars.
So constant inflation actually doesn’t do it. It’s even worse than the picture suggests, as I said. If we instead assume the nominal debt increases by the change in the interest outlays – which means discretionary spending still decreases, but only in real terms – then that chart looks distinctly uglier. The nominal debt in 25 years with these assumptions gets to $294T, and the real debt steadily climbs over $100T.
We can see that this is just not going to work unless we decrease discretionary spending. Again, we’re just trying to figure out if there is some way to inflate ourselves out of this mess. So let’s go back to the assumption that the deficit only rises pari passu with inflation, but now let’s get rocking on inflation and increase it 1% every year, from 4% to 30% over the next quarter-century. Assuming that investors still demand a positive 1% real rate on all new debt, we get this picture.
We finally make some headway on the debt! Because so much of it is short-term, though, it takes a good 6-7 years before we see a lot of progress on the real debt because we’re just rolling it over at higher rates too quickly.
By the way, I neglected to mention that Medicare is an off-balance-sheet ‘debt’ that we can’t outgrow in this fashion. And it’s a big item. It is impossible to inflate away Medicare.
Now, most people who suggest that we can inflate our way out of debt do not imagine a steadily-accelerating inflation rate forever. That’s not what we would call a ‘re-electable outcome’. So let’s instead assume that we spike inflation for two years to 10%, and then drop it back to 4%. You can look at the chart below and see that once again this doesn’t work because the debt is turning over too quickly. There is a quick, small benefit to the real debt, but then it levels off again.
Unless you can spike inflation to, say, 100% for a year or two and then put it right back down to the prior level, you really need accelerating inflation because you need the principal amount of the outstanding debt to fall in real terms more than the interest payments are accumulating. If you can have inflation a bit above the average interest rate on the debt, then you’ll make headway but since the interest rate responds you have to keep doing this for a while. And remember, we are assuming that the government outside of interest payments is steadily shrinking in real terms.
What about if we balanced the budget? Well, then obviously the real debt will tend to decline over time, although more and more of the budget becomes interest payments.
So far, we can find no realistic way to inflate our way out of debt, other than moving toward hyperinflation (and the faster, the better). And even with these simulations, I am making a very unrealistic assumption about how the deficit evolves when the interest costs of the debt blow up. Hyperinflation, with a balanced budget, is what you need. Good luck with that. The only way that happens is if the dollar collapses and no one will lend us any more money anyway. And there is no one in government today, I feel confident in asserting, who thinks that outcome is a decent tradeoff in order to get out from under the mountainous debt. At least…let’s hope not.
Have a better idea? Let’s hear it!
Bounce in Money Growth is Good News and Bad News
The monthly money supply numbers are out. I have bad news and good news.
The bad news is that the contraction in the money supply appears to be over. That’s not bad news per se (see below), but it’s bad in that the anti-inflationary work that was happening is coming to an end before it’s quite finished. Although I would be reluctant to annualize any one month’s change in M2, the $92bln increase in M2 in March was the largest increase since 2021. It only annualizes to 5.5%, so it isn’t exactly running away from us – but it’s positive. The 3-month and 6-month changes are also positive, and the highest since early 2022 in each case. Again, we’re only 0.72% above the ding-dong lows of last October, but the sign is now positive.
With the money supply figures now in, and with the advance Q1 GDP report due this week, we can revisit our chart of “how much more inflation ‘potential energy’ remains.” (see “Where Inflation Stands in the Cycle,” November 2023). As that article (and this chart) illustrates, if M2 doesn’t go down then this gets more difficult. M2 in Q1 rose at a 1.24% annualized rate over Q4. GDP is expected to rise 2.5% annualized. So M/Q…barely moves, as the chart shows.
We will eventually get back to the line, unless velocity is permanently impaired. Despite all of the crazy people who told you it was, there’s no evidence of that. M2 velocity will rise about 1% (not annualized), if the GDP forecasts are on point. That will be the smallest q/q change in several years, and velocity will be getting very close to the 2020Q1 dropping-off point. But there frankly is no reason for velocity to stop there; higher interest rates imply higher money velocity. However, we are getting close.
(Incidentally, if you’re curious how we can be almost back to the dropping-off point of velocity and yet still be 5% below the line in the first chart above, it’s because I’m using core inflation. With food and energy, we’re a little closer to the line and have used up more of the ‘potential energy.’ But food and energy are of course volatile and so while a good spike in energy prices would look like we’ve used up all of the potential energy, that could just be a one-off effect.)
Either way, we aren’t too far away from getting back to home base and that’s good news. Yes, prices by the time we are done will have risen 25% since the end of 2019, and that can’t really be characterized as a ‘win.’ Let’s go Brandon. But we are getting closer.
The good news about the new rise in M2 is that it’s timely. Markets and the economy were starting to show signs of money getting a little tight; losing a little lubrication in the machinery. An economy does need money to run, and while the only way we can get back to the old price level is to have money supply continue to decrease, that’s also a painful process. In the long run, we would have price stability if the change in M was approximately equal to the change in GDP. If we want 2% inflation, then we need M to grow about 2% faster than GDP. Vacillating velocity means that it isn’t purely mechanical like that – the steady decline in velocity since 1997 is the only reason that inflation stayed tame despite too-fast money growth over that period – but the long downtrend in velocity is likely finished since the long decline in rates is finished. Thus, if we get money supply growth back to the neighborhood of 4%, we can get our 2-2.5% growth with restrained inflation over time.
I am not super optimistic that all of that will work out so nice and cleanly like we draw it up on the chalkboard, but I am more optimistic about it than I was two years ago. We still have some sticky inflation ahead of us, but if the Fed keeps reducing its balance sheet then eventually we will get inflation below the sticky zone and back towards ‘target’ (even though there isn’t a target per se any more).
Re-Blog: Volatility and Position Size
This is one of my favorites, and every few years I re-blog some portion of this article. The original, I wrote in 2010. The basic question is, what is the correct way to respond as an investor to increasing uncertainty? In the original blog and in various re-posting edits, I’ve applied a basic idea called the “Kelly Criterion” to explain why responding to market selloffs by trimming a position, rather than adding to it, is often the right strategy (in the sense of it being mathematically optimal, not in the sense of it always producing the best returns). The idea also applies to the question of what to do when the general level of uncertainty and volatility rises (or falls) in markets. With developing uncertainty in the Middle East and the US spiraling towards what looks to be a summer of crazy politics, it is rational – even optimal – to ‘take some chips off the table.’ Read on for why.
(“Kicking Tails” originally appeared February 12, 2018)
Like many people, I find that poker strategy is a good analogy for risk-taking in investing. Poker strategy isn’t as much about what cards you are dealt as it is about how you play the cards you are dealt. As it is with markets, you can’t control the flop – but you can still correctly play the cards that are out there.[1] Now, in poker we sometimes discover that someone at the table has amassed a large pile of chips by just being lucky and not because they actually understand poker strategy. Those are good people to play against, because luck is fickle. The people who started trading stocks in the last nine years, and have amassed a pile of chips by simply buying every dip, are these people.
All of this is prologue to the observation I have made from time to time about the optimal sizing of investment ‘bets’ under conditions of uncertainty. I wrote a column about this back in 2010 (here I link to the abbreviated re-blog of that column) called “Tales of Tails,” which talks about the Kelly Criterion and the sizing of optimal bets given the current “edge” and “odds” faced by the bettor. I like the column and look back at it myself with some regularity, but here is the two-sentence summary: lower prices imply putting more chips on the table, while higher volatility implies taking chips off of the table. In most cases, the lower edge implied by higher volatility outweighs the better odds from lower prices, which means that it isn’t cowardly to scale back bets on a pullback but correct to do so.
When you hear about trading desks having to cut back bets because the risk control officers are taking into account the higher VAR, they are doing half of this. They’re not really taking into account the better odds associated with lower prices, but they do understand that higher volatility implies that bets should be smaller.
In the current circumstance, the question merely boils down to this. How much have your odds improved with the recent 10% decline in equity prices? Probably, only a little bit. In the chart below, which is a copy of the chart in the article linked to above, you are moving in the direction from brown-to-purple-to-blue, but not very far. But the probability of winning is moving left.
Note that in this picture, a Kelly bet that is less than zero implies taking the other side of the bet, or eschewing a bet if that isn’t possible. If you think the chance that the market will go up (edge) is less than 50-50 you need better payoffs on a rally than on a selloff (odds). If not, then you’ll want to be short. (In the context of recent sports bets: prior to the game, the Patriots were given a better chance of winning so to take the Eagles at a negative edge, you needed solid odds in your favor).
Now if, on the other hand, you think the market selloff has taken us to “good support levels” so that there is little downside risk – and you think you can get out if the market breaks those support levels – and much more upside risk, then you are getting good odds and a positive edge and probably want to bet aggressively. But that is to some extent ignoring the message of higher implied volatility, which says that a much wider range of outcomes is possible (and higher implied volatility moves the delta of an in-the-money option closer to 0.5).
This is why sizing bets well in the first place, and adjusting position sizes quickly with changes in market conditions, is very important. Prior to the selloff, the market’s level suggested quite poor odds such that even the low volatility permitted limited bets – probably a lot more limited than many investors had in place, after many years of seeing bad bets pay off.
[1] I suspect that Bridge might be as good an analogy, or even better, but I don’t know how to play Bridge. Someday I should learn.
Inflation Guy’s CPI Summary (Mar 2024)
After a week when the NY/NJ area saw an earthquake, an eclipse, and a gorgeous 75-degree spring day, it is time to get back to work.
Today’s CPI report was not expected to be particularly great. In fact, one of the biggest conundrums of market behavior recently has been the question of why investors seemed to remain very confident that the Fed will cut rates several times this year, even as forecasts for the path of inflation have backed off of what they were last year (when most forecasters had core CPI returning placidly and obediently to the neighborhood of 2% this year). The a priori consensus forecasts for today’s CPI figure were +0.28% m/m on core and +0.33% m/m on headline. The Kalshi market was in line with that, although CPI swaps were a touch lower on headline at +0.29% (seasonally adjusted, but CPI swaps trade NSA CPI). That’s not wonderful: 0.28% on core would annualize to 3.4% y/y.
The assumption has been that even if in March we are annualizing to 3.4%, the coming deceleration in rents will push everything back down to where it needs to be. The problem with this has always been (a) the strongly-held belief that rents would slip into deflation this year were never based on good analysis, and more importantly (b) this assumed that nothing unforeseen would happen in the other direction. It is characteristic of inflationary periods, of course, that bad things happen on the upside. So this was always sort of assuming a can opener,[1] but at least forecasts for the current data were reflecting that these things had not happened yet. To be fair, the consensus on core has been low relative to the actual print for four months in a row, but at least folks are forecasting mid-3s, rather than 2.0.
Now, let’s review one other thing before we look at some charts. The recent story boils down to this: sticky rents, sticky wages. While core goods has been pulling down core inflation, that game was running out of room. The next part of core deceleration relies on un-sticking the sticky rents, and sticky wages.
So here we are. Today’s figure +0.36% on core CPI, +0.38% on headline (seasonally adjusted on both). This makes the last 3 core CPIs 0.39%, 0.36%, and 0.36%. The chart below of the m/m core CPI figures does not really give the impression of a decelerating trend.
We always look these days first at rents, because that is so important to the disinflation story. Rent of Primary Residence was +0.41% m/m, down from 0.46% last month. Owners’ Equivalent Rent was steady, at +0.44%. Remember that there had been some alarm two months ago, when OER for January jumped to 0.56%, that this was due to a new survey method or coverage and it was going to be repeated going forward. That was always pretty unlikely, but now we have had two months basically back at the old level and the January figure appears to be an outlier. 0.41% on Primary and 0.44% on OER is not hot, just sticky. It isn’t going up; it’s just not going down very fast.
Rents will continue to decline. But the failure of rents to slip into deflation is a source…maybe the source…of the big forecast error made by economists about 2024 CPI. Our cost-based model for primary rents, which never got even vaguely close to deflation, has now definitively hooked higher with the low coming in November. Rents haven’t been decelerating as fast as our model had them, but if anything that’s a source for concern on the high side.
Outside of rents, core inflation ex-housing rose to 2.38% y/y. That sounds like “most of the economy is on target,” but that’s not how inflation works. There’s a distribution, and if the ‘rents’ part of the distribution is going to be higher than the target then everything else needs to average something below the target. We aren’t there. And, as I noted above, we’ve squeezed out just about everything we can from core goods. Actually, y/y core goods dropped to -0.7% thanks partly to continued declines in Used Cars (-1.1% m/m) and some decline this month in New Cars (-0.2%). I think the latter might partially reflect discounts on the EV part of the fleet, where cars for sale have been piling up as manufacturers under political pressure have been producing far more of them than people want.
Note that core services, even with the decline in y/y rents, moved higher this month to 5.4% from 5.2% y/y. Some of that was medical care, which was +0.49% m/m driven by another jump (+0.98% m/m) in Hospital Services. The y/y rise in Hospital Services is now up to 7.55% – the highest since October 2010.
Partly driven by hospital services, the ‘super core’ (core services ex-rents) continues to re-accelerate.
Again, this is not what the Fed wanted to see; and it’s driven partly by the stickiness in wages. The Atlanta Fed’s Wage Growth Tracker has decelerated but is still at 5.0% y/y. That’s not the stuff that 2% core inflation is made of.
Let’s take one moment to look at a piece of good news from the report. My estimate of median CPI, which is my forecast variable because it is not subject to outliers like Core CPI, is +0.32% for this month. Because I have to estimate seasonals for the regional housing numbers, actual Median might be a teensy bit higher or lower but in any event the chart of Median CPI is much less alarming than the chart of Core CPI.
I should observe that the news there is not completely good, since a signature of inflationary environments is that tails are to the upside – that is, core is persistently above median. That was true during the upswing, but during the moderation core has gone back below median. But this bears watching, and if core starts to routinely print above median it will be a negative sign. For now, though, the Median CPI is good news. Relatively.
So let’s talk policy.
The Administration always seems to be confused about why, despite strong jobs numbers, consumers consistently report dissatisfaction with the economic situation. There really shouldn’t be any confusion. Consumers, especially those not in the upper classes, hate taxes. And in addition to a high direct take from the government in explicit taxes, consumers are also facing persistent inflation that the Administration says isn’t there. Inflation is a tax, and it sucks worse than direct taxation because you can’t rearrange your consumption very well to avoid it. (You can rearrange your investment portfolio, but a strikingly small number of people seem to have actually done that even three years into this inflation episode. If you’re curious about how, you really should visit Enduring Investments and ask.)
On the other question of policy, and that’s the Fed: I can’t see any rational argument for cutting rates in June. Actually, on the data we have in hand I can’t see an argument for cutting rates in 2024, except for the one the Fed doesn’t consider and that’s that interest rates don’t affect inflation. To cut the overnight rate, the Fed would have to rely on forecasts that inflation is going to get better. And to do that now, when forecasts have been persistently wrong (and not by just a little bit but about the whole trajectory) since 2020, would be incredibly cavalier. The FOMC still consists of human beings, so never say never. And the inflation data should improve as the year goes along and rents moderate. I just don’t see any sign that it’s going to moderate enough to say ‘we’ve reached price stability.’ Sticky in the high-3s, low-4s is still where I think we’re coming out of this.
[1] A physicist, an engineer, and an economist are stranded on the desert island with nothing but a crate of canned food. “How are we going to get the food that is inside of these cans?” asked one. The physicist says “well, we could heat the cans, carefully, in a crucible we make from ocean clays. Eventually the heat will cause the can to burst and we can get the food inside.” The engineer says “that will take too long. What we need to do is take some of these coconuts, raise them up to a great height with a series of ropes I will design, and allow them to smash down onto the cans, breaking them open.” The economist says “I have a solution that is far easier than what you fellows are doing. Here is how we do this. First, assume a can opener….”
Changing the Fed’s Target – FAIT non-accompli?
As the steadier measures of inflation (core, median, or sticky depending on your preferences) have started to overshoot expectations slightly – the y/y measures continue to decline, but slower than expected as the m/m numbers have surprised on the high side – the markets have continued to price Fed policy becoming increasingly easier over the course of 2024 and into 2025. While Fed officials continue to push back gently on this assumption, it seems that most of the FOMC is comfortable with the idea that there will be at least some decrease in overnight rates later in the year and the only question is how much.
While inflation has not been settling gently back to target, there have developed two big holes in the narrative that the Fed was depending on. First, there is no reason to think that rent of shelter is going to cross over into deflation, either in 2024 or any time in the future. The belief that the CPI for rents would follow the high-frequency data into deflation was never well-founded, despite some fancy-looking papers that claimed you could get three pounds of fertilizer out of a one-pound bag if you just squeezed it the right way (I discussed “Disentangling Rent Index Differences: Data, Methods, and Scope”, and why it wasn’t going to tell us anything we didn’t already know, in my podcast last July entitled “Inflation Folk Remedies”), and while rents are declining they are not plunging, and home prices themselves have turned back higher and are growing faster than inflation again.
Second, core-services-ex-rents (so-called ‘supercore’) inflation needed to see wages decelerate a lot in order for that piece to get back towards target. They haven’t, and it hasn’t.
This isn’t to say that these things may not eventually happen, but so far the expectation that we would get back to target sustainably by the middle of 2024 looks quite unlikely. Why, then, are people talking about when the first eases will happen? The only way that it makes sense to do so is if the goal to get inflation back to 2% sustainably is no longer driving policy.
This has led to some observers pointing out that the Fed doesn’t actually have a 2% target any longer. In 2019, the Fed moved to Flexible Average Inflation Targeting, or FAIT. Under this rubric, the Fed doesn’t need to regard 2% (or about 2.25% on CPI) as a target that they need to hit at a moment in time but only as an average over some period of time. This obviates the need for overly-aggressive monetary policy in either direction, such as the instantaneous adjustment linked directly to the inflation-miss that is required by the Taylor Rule.
Unfortunately, under that rule the Fed has little if any chance of meeting its mandate. It would have a better chance of hitting 2% in…um…let’s say a ‘transitory’ way, as rental inflation swings lower and we pass close to the target briefly before inflation goes back up to its new equilibrium level. Back in August 2021 I noted that the Fed was already above the FAIT projected from the announcement of that policy, and in fact had used up all of the post-GFC slack. Obviously, it has gotten worse since then. Below, I update the two charts from that article. The first chart shows the CPI from August 2019, along with the average-inflation-targeting line and the forwards suggested by the CPI swap market (showing where inflation futures would be trading, if they were trading).
The second chart shows the CPI back to January 2013. We’ve made up all of the inflation from the post-GFC deflation scare, and then some.
Note that the inflation swap market is not indicating any expectation that prices will return back to the trendline. The market is acting as if the Fed is still operating under the old rules, where the goal was to get inflation to be stable at 2% from here, wherever “here” is. This means one of four things will have to happen, or it implies a fifth thing.
- The Fed needs to re-base its FAIT to start from the current price level. In that case, the red CPI-plus-2.25% line will shift abruptly upward but then will parallel the inflation implied by the inflation market; or
- The Fed can keep the original base, but concede that the actual target now is 3% (about 3.25% on CPI), which means that if the inflation market is right then it should be back on target by late 2029 (see chart); or
- The Fed can dedicate itself to fighting inflation for much longer, and publicly disavow the notion of reducing interest rates in the next few years. If CPI went completely flat then the Fed would be back on the line by sometime in 2028.
- The Fed can abandon FAIT, because it has become inconvenient, and validate the inflation market’s assessment that the Committee would be happy with 2% from here, not on average.
If none of these things happens, and the Fed then implies that the inflation market is going to permanently imply something different from what the Fed claims to be its modus operandi. In that case, it would be very hard to argue that the central bank had not lost credibility, wouldn’t it?
Four Quick Thoughts on Fed Day
Four fairly quick observations on this Federal Reserve meeting day, not all of which have anything to do with the Fed:
1. The FOMC today announced unchanged policy for now on the overnight interest rate, on the pace of QT runoff, and on the collective expectation of the Committee for the number of rate-cuts in 2024 (three, 25bp cuts). But it beats noting that while three cuts is the median expectation, the mean expectation dropped substantially. Only one official sees four rate cuts in 2024, compared to five who saw that many or more, as of the December survey. Those four folks moved to ‘three’, and one of the ‘three’ folks moved to ‘just one.’ Nine of the nineteen dots are for fewer than three cuts this year, so we should say this is a closer call than the market seems to think.
2. The longer dot plots also show some increase in Committee members’ expectations for the neutral short-term interest rate (the so-called ‘r-star’ originally popularized, I think, by Greenspan). The significance of this for investors and traders is that the overnight rate is unlikely to go back to zero unless we get another enormous calamity; the significance for the economy is essentially nil since it is money, and not interest rates, that matter. I’ve written before about why there are good reasons to think of something like 2-2.25% as the neutral long-run real rate, and so if CPI inflation is expected to be 2.25%-2.5% then something around 4.5% is neutral long-run nominal rate. We are mighty close to that now, so there is no compelling reason to think that interest rates should decline markedly from here. At the short end of the curve, we should eventually be lower – but we need to also keep in mind the growing imbalance in the supply and demand for Treasury paper, which (in the absence of recession) will tend to keep rates on government paper higher than they otherwise would be in equilibrium – and as one consequence, by the way, credit spreads will tend to be lower than they otherwise would be for a given level of creditworthiness.
3. The Fed clearly believes that the situation in Commercial Real Estate (CRE) and its effect on the banking sector is manageable. If they didn’t think so, then they would be hastening to lower rates to ease the refinancing problems that are hitting that sector. I have been reading alarmist analyses saying that the $1 trillion in CRE mortgage maturities due this year will lead to ‘hundreds’ of bank failures. This falls into the Big Number is Bad and Scary school of analysis. One trillion is a lot of mortgages and that will cripple banking! Except…
Let’s suppose that 20% of those mortgages go into default – a number more massive than we’ve ever seen before – and that recovery is 80%. For reference, in the 2008-09 crisis CRE values fell by about 36% according to the Greenstreet Commercial Property Price Index (chart below), and that was against a backdrop of 1%ish inflation. The nominal price decline should be less in an environment where underlying inflation is 4% per year, naturally. Since the CRE peak, real values have fallen 31% but nominal values only about 21% on the basis of that index. But the drop from the peak isn’t the relevant part. Even the shorter loans now coming due were struck 3-5 years ago, and the drop from that level is only about 9%. Plus, the initial loan-to-value levels were not 100%. So (and all of this is just to cuff a rough estimate) a 20% loss when selling out the collateral on a defaulted mortgage seems conservative.
Those numbers mean the $1T in mortgage maturities could produce a loss of $40bln (1,000 * 0.2 * 0.2). That’s still a big number, but remember that it is spread over a lot of banks. Suppose that it is spread over only 2,000 banks, and that the losses have nothing to do with bank size. Then you are looking at losses per bank of $20mm. That’s bad for a small bank, but the losses at a small bank will of course be smaller because they have smaller books. Will that sink ‘hundreds of banks’? Only if they are small, fairly insignificant banks.
Will some banks fail because they lent too much against commercial real estate which has fallen in value, at too-high loan-to-value ratios, and end up owning property that they can’t sell? Almost certainly. But after negotiations and forbearances and the eventual foreclosures – in an environment where the price level is rising 4% per year – I just don’t think this is something we should worry about. To be fair, the fact that the Fed is not worried about it is something that makes me worry about it.
4. I have been befuddled recently because airfare prices in the CPI have been higher than would be anticipated given the movement in jet fuel prices. Belatedly, I think I know what is going on. The issues with Boeing planes has meant that (and I didn’t know this) Boeing has greatly reduced its deliveries to airline companies as they sort out the problems with their Max jets. I became aware of this only recently when a Bloomberg story highlighted how Southwest Airlines is cutting capacity and freezing hiring because they aren’t getting the planes they need. Steady demand and constraints on supply means higher airfares, as I also discovered this week when I was booking a flight to Chicago. Yikes! With jet fuel prices also rising again, this is something to factor into CPI forecasts going forward. It’s surely ‘transitory,’ but it takes a long time to build a plane and in the near-term this is more likely to be solved on the demand side if we have a recession, than on the supply side with a sudden influx of planes.
Inflation Guy’s CPI Summary (Feb 2024)
I must say that I didn’t see this one coming. Credit where credit is due, though: while Street economists were just a little low (consensus was +0.40% headline, +0.30% core), the CPI swap market at least got headline right (there being no market for core inflation CPI swaps) by pricing in +0.47%, seasonally adjusted. The actual print was +0.44% on headline CPI, and a lusty +0.36% on core. I was lower, even though I got the big pieces right. I had some tails going the wrong way. Let’s get into it.
The things which threw me were airfares and used cars. Based on declines in jet fuel, I had anticipated that airfares would be roughly -6% m/m, and I merely got the sign wrong as they were +6.6%. Jet fuel was tighter on the east coast, and I suspect regional differences there is what caused this wide divergence. If I’m right, then airfares will underperform jet fuel over the next few months. If, instead, it’s a cost-of-labor or cost-of-equipment thing, or if it’s increased pricing power from airlines because of capacity constraints, then airfares won’t drop back and that would be a bad sign.
Similarly, Used Cars continues to outperform the Black Book survey. I had penciled in -1%, and Kalshi markets were around -1.5%, but Used Car CPI came in +0.5%. This is a volatile series, and this miss is only interesting because Used Cars keeps missing a little high compared to the Black Book survey. That could be an issue of sample mix, but I’m not sure. New Cars were -0.10% m/m. Car and Truck Rental was +3.83% after -0.74% last month, so that’s another upper tail. Overall, core goods was steady at -0.3% y/y.
I said I got the big pieces right. I refer to rents. Remember that last month we had a large deviation between Owners’ Equivalent Rent (OER) and Rent of Primary Residence. Normally, these two track pretty closely, but occasionally they deviate and last month OER was 0.2% higher than Primary Rents. That contributed to the very high median CPI in January, and there was a ton of discussion about whether the BLS had done something weird with the survey – they had, in January 2023, refined the OER weighting method and there was concern that this was a ‘mix’ problem that was going to continue to push OER higher than Primary Rents for a while. The BLS contributed to this sense of confusion by sending out a blast email which seemed to suggest it was so; they had to walk that back and to their credit did a very nice webinar and has spent a lot of time this month explaining in excruciating detail how the OER survey is conducted. Bottom line: there’s nothing to see here; sometimes the two series diverge slightly. Moreover, as I’ve pointed out previously, when natural gas prices are declining it tends to mean that the cost of imputed utilities is declining which, since they’re deducted from the rental survey used for OER means OER should be slightly higher than Primary Rents over time. Not 0.2% per month, though, and I expected this aberration would mostly close this month.
It did, with OER +0.44% m/m (was +0.56% last month) and Primary Rents +0.46% m/m (was +0.36% last month). Year over year, they’re about the same but OER has moved slightly above Primary.
So the surprising part to me was that Primary came up some to help close that gap, not that the gap closed. I continue to expect rents to decelerate, along with everyone else – only, as I will keep saying until I am blue in the face, we are not going to go into rent deflation as so many people have been forecasting (folks seem to be backing off that now!) but rather we should drop into the 2%-3% range y/y before rebounding later this year.
There seems to be evidence of that in the independent rent measures. Below is a chart from a recent Redfin news release. It bears noting, of course, that these rent measures also all went into deflation and misled all of those economists who lean on these high-frequency-but-low-quality data. (Having said that, Redfin does seem to be better than some others, but it’s still measuring something different than what the CPI is measuring).
Now, the story starts to become a little clearer, albeit concerning. Core services rose to 5.4% y/y from 5.2% y/y, while core goods was unchanged as I noted above. Rents are coming down, but outside of rents we are seeing some stabilization at higher-than-pre-COVID levels. The chart below shows Shelter CPI, and Core CPI ex-Shelter, which has been roughly stable for three months around 2.25%. That sounds great, since 2.25% on CPI is roughly equivalent to 2% on the Fed’s PCE target…except that 2.25% is higher than it was pre-COVID. The theme, and we’re seeing it in several places, is inflation being sticky at higher levels than it was pre-crisis.
There were some good parts to the report – notably Food, which was tame m/m for both Food at Home (-0.03% m/m versus +0.37% last month) and Food Away from Home (+0.10%, was +0.47%), although the latter is probably not sustainable given rapidly-rising wages. Still, it’s positive. Unless you’re buying baby food, which is +9.2% y/y!
Actually, babies got a lot more expensive this month. The largest increase in the categories used for Median CPI was Infant/Toddler Apparel. In general, apparel categories were right-tail items this month. But there were not enough of them to explain the high core CPI. Median was +0.39% (my estimate); since that’s right in line with core it says the tails weren’t what moved this number. It’s just that this month, inflation rose at something like a 4.25%-4.75% annualized pace.
With this, and with Core Services ex-shelter (“Supercore”) at +0.47% m/m – which means supercore accelerated to +4.3% y/y – it is inconceivable that the Fed will yet consider cutting rates. It is possible that they may later in the year, but there is far too much exuberance in the bond market about that prospect.
Indeed, there’s far too much exuberance generally. Stocks rose on an inflation report showing that inflation was higher than expected. I’m not saying that equities should crash on this data, but that’s the sort of reaction that you tend to see in bubbles. The market will be semi reserved going into an economic report, but then rallies afterwards regardless of the data. I have seen that sort of environment, where such a thing happened regularly, a couple of times in my career and they never ended well. To jump on this data, as if it was in any way positive, says that people were just waiting until after the number to buy, and they were going to buy no matter what. That’s not a healthy market, especially when that happens at high prices rather than low prices.
I continue to expect median inflation to decline to the high-3s, low-4s, maybe dipping a little lower than that in Q3 if rents bottom then as I expect. The bottom line is that we’re near levels where I have been expecting inflation to get sticky, and it seems to be happening. I didn’t see this particular month being sticky, but the general tenor of the data makes sense to me.
Understanding Biden’s Poll Numbers Despite a ‘Strong Economy’
The Biden team keeps talking about how they can’t believe how underwater the President’s poll numbers are, when the economy is so frickin’ good. “As soon as people figure out how frickin’ good it is, they’ll come running to vote for him.”
At some level, one can be sympathetic with that view. Inflation is down to only 3.1%, the Unemployment Rate is still sub 4% even with the most-recent rise, well below the levels when he took office; Average Earnings are up and gasoline prices are down around $3 after being above $5. What’s not to like? Moreover, put this record next to Trump’s record! When Trump came into office, Unemployment was 4.7% and when he left it was 6.7%!
The problem that the Biden team has – and, frankly, the one it has always had – is that they have no idea how actual people experience the economy, and no idea how actual people think.
Americans, on average, tend to be fair. When people think about the Trump years, they recognize that it isn’t quite fair to saddle him with COVID. While they don’t think this explicitly, their memories about the 2016-2020 period fall into “pre-COVID” and “post-COVID” zones. In other words, if in mid-March 2020 a particular consumer was positively disposed towards the Trump economy, then that’s what their memory is. When COVID hit, it started a new time period in their memory. So to the normal person, they remember Trump coming in with a 4.7% Unemployment Rate and watching as it fell to 3.5% in February 2020. “Then COVID hit.” This works against Trump in little ways too; no one gives him credit for the disinflation that happened between March 2020 and the end of his term.
So this is the way that normal people see Trump’s record:
Now, the best part of Biden’s record is that Unemployment fell from 6.7% when he took office to 3.7% as of January. Other than that, though, his record in the minds of Americans looks unimpressive. (Of note is – and folks, don’t shoot the messenger; I’m just showing the data – that the Biden team persistently claims that real earnings have risen during his Administration, while it isn’t so.)
And so now, let’s put them side by side. Inflation is higher under Biden, gasoline prices have risen under Biden, real earnings are down under Biden, and food costs are up (a lot) under Biden. The unemployment rate has fallen more, but is now higher than it was pre-COVID under Trump!
If you realize that Americans are not going to blame Trump for COVID, then it gets very easy to understand why Trump polls better on the economy.
Recession For Me But Not For Thee?
In late 2022, I often said that while I didn’t think it would be severe I figured we would have a recession in 2023 because we had never seen an energy spike at the same time that the Fed was aggressively tightening and not had a recession. Indeed, it would really be weird if those things could happen and not result in a recession. Then, what causes a recession?!?
And as we all know by now, 2023 was not a recession. So, like any good trader who makes a bad call I want to look and figure out why that happened. Funny thing is…I wasn’t completely wrong on that.
We need to continue to remember that the volatility of 2020-2022 is something that doesn’t just vanish; the oscillations echo and repeat with slowly decreasing amplitude. The story of those years was this:
The economy was mostly shuttered in mid-2020, and the federal government and Federal Reserve showered money on consumers and businesses. Because service-providers were basically closed, the money was poured into goods. This surge in demand led to long port delays and lead times, higher prices for goods, and a boom time for manufacturing. In the chart below, the orange line is the ISM Manufacturing New Orders index where 50 represents a dividing line between expansion and contraction (it’s a survey, so it’s not absolute levels but rather the change that is noted by respondents).
We can also see the boom in industrial production, next chart. The y/y increase in production was at 4% or 5% until late 2022.
In March 2022, the Fed hiked 25bps. They did 50bps in May, 75bps in June, another 75bps in July, and of course they kept going. Crude crested at $123/bbl in June. So by late summer of 2022, goods production started to see this in declining orders (first chart) and weakening production growth (second chart). My sources in industry started to see a buildup in client inventories, leading to lower orders – the celebrated ‘bullwhip’ effect. By late 2022 and for much of 2023, manufacturing was absolutely in a recession. The Conference Board’s Index of Leading Indicators had gone negative m/m in March 2022 and actually is still negative today.
Normally, that set of events would have produced a recession. But they didn’t. Why? Because the reopening of service industries was gathering steam over 2022. Remember that lots of restaurants were not allowed to operate at full capacity until the second half of 2021. You can see the purple line in the chart above jumped higher late in 2021 and from the middle of 2022 remained above the goods-producing orange line. (“New Orders” for services industries is a more complicated concept, but you get the point). Remember how difficult it was for many service providers to find employees willing to work, and the spike in wages that was necessary to lure them? Well, you don’t have to remember, because here’s a chart showing wage growth for services employees (Atlanta Fed Wage Growth Tracker for services, in dark blue) against ‘Supercore’ core-ex-shelter CPI.
The later blooming of services, and the difficulty of service providers to build up capacity, stretched out the services expansion so that while manufacturing was in a recession, services were in an expansion. And, since services are a much larger part of the US economy, this meant that we never recorded an actual recession on overall growth. Moreover, the decline in goods prices helped flatter the increase in pressure on services prices, so that inflation measures turned lower while inventories were being right-sized.
Now, we are starting to see manufacturing start to turn higher – that orange line in the first chart popped above 50 in January and retreated to just below it in February. This is consistent with what I am hearing from my contacts, who are being more discerning about responding to this increase in demand by greatly expanding capacity (and then possibly being burned again). It means that goods prices are no longer falling, and in many cases are rising again. I do think that there are some signs of consumer stress, such as auto loan delinquencies, and the purple “New Orders (Services)” line in the first chart looks to be slowly decelerating. So I think it’s possible that this year we actually do get a recession, but I think it will be mild because manufacturing is oscillating in the upward direction now. That oscillation means that growth will not be as soft as it would be if services and goods were synchronized, which is one reason I believe this will be a mild or ‘garden variety’ recession the likes of which we haven’t seen in a while.
Eventually the two parts of the economy will re-synchronize, but the way the reopening happened is I think why the macro call has been so difficult.
AI: Even a Big Deal is Smaller Than You Think
So, we are back to the argument about whether we have reached a new era of permanently higher growth and earnings, and because of productivity also a permanent state of steady disinflationary pressures.
Live long enough, and you’ll see this argument come around a couple of times. In the late 60s with the “Nifty Fifty” stocks, in the 1990s with the Internet, and now with AI. As a first pass, it’s worth noting as an equity investor that the first two of those eras were followed by long periods of flat to negative real returns in equities. But my purpose here is simply to revisit the important fact that productivity is always improving, so something which improves productivity is normal and not exciting. The question which arises periodically when we see some really golly-gee-whiz innovation is whether that innovation can meaningfully accelerate the rate of productivity growth over time.
Total real growth over time is simply the growth in the labor force, plus the growth in output per hour (productivity). Assuming that the labor force grows at roughly the same rate as the overall population,[1] real GDP per capita should grow at roughly the rate of productivity. The chart below extends a chart which first appeared in an article by Brad Cornell and Rob Arnott in 2008 (“The ‘Basic Speed Law’ for Capital Markets Returns“), updated to the end of 2023Q3. Note that real earnings and real GDP grow at almost the same rate over time – the log regression slope is 2.09% for real per capita GDP and 2.17% for real earnings.
(By the way, although it isn’t part of my discussion here note that the middle line, real stock prices, isn’t parallel. It was, back when this chart first appeared in 2008; the fact that it isn’t any more is obviously attributable to increases in valuation multiples over a long period of time. Discuss.)
A permanent (or at least long-lived) increase in the long-run rate of productivity growth, then, would be massively important. It would mean that GDP per capita – standard of living, in other words – would rise at a permanently faster pace. This is the crux of the question, as I said above and as NY Fed President John Williams said in an interview with Axios a few days ago (ht Alex Manzara):
“One way to think of it is AI is – and this is my own, but based on what I heard from others – is AI is just that new thing that’s going to get us that 1% to 1.5% productivity growth that we’ve been getting for decades or even a century.
“It’s the thing that gets us that, just like computers did or other changes in technology and how we produce things in the economy. So it’s just the thing that gets us that 1% to 1.5% productivity growth.
“The other view, which I think has some support, is AI is more of a general purpose technology. …So there is a possibility that we could get a decade or more faster productivity growth if this really is its general purpose and revolution. You can’t exclude that.”
What Williams said, about AI being a “general purpose technology” that spurs faster productivity growth for a decade or more, is something that we honestly have a pretty good history of. The explosion of the internet into general use in the late 1990s triggered an equity market bubble that eventually popped. Greenspan mused, in late 1996, that it’s hard to tell when stock prices reflect “irrational exuberance” and in February 1997 he said “history counsels caution” because “…regrettably, history is strewn with visions of such ‘new eras’ that in the end have proven to be a mirage.”
Was it a mirage? There is no question, a quarter-century later, that the internet has completely changed almost everything about the way that we live and work. If there was ever a ‘general purpose’ technology that led to a sustained long-term increase in productivity, the Internet is it.
My next chart only goes back to 1979. It shows US Nonfarm Business Productivity, calculated quarterly by the BLS as part of the GDP report. Obviously, the quarterly numbers are incredibly volatile – so much so, in fact, that I’ve truncated a large portion of the tails. It’s devilishly hard to measure productivity. More on that in a moment. The red line is the 20-quarter (5-year) moving average. The average over the whole period is…surprise!…1.92%, very close to the average increase in real earnings and real GDP per capita. As I said before, that’s what we expected to find.
But there is certainly a bulge in the chart. Noticeably, it doesn’t happen until long after the internet hype had crested, but it is definitely there. The average on this chart from 1979-1998 is 1.78%, and the average since 2005 is 1.59%. But the average from 1999-2005 inclusive is a whopping 3.11%. An acceleration of productivity growth of 1.4% or so, for 7 years, means that our standard of living moved permanently higher by about 10% during that period, over and above what it would have done anyway.
That’s meaningful. I would also argue that it’s probably the upper limit of what we should expect from the AI revolution. Starting in few years, if this is a “general purpose technology” advancement, we could conceivably see growth accelerate by 1.5% per year for some part of a decade. Let’s all hope that happens, because that 10% total growth is the real growth – it is extra growth without any extra inflation. A free lunch, as it were. I say that’s probably the rough upper limit because I can’t imagine how the AI revolution could possibly be more impactful than the internet revolution was, or any of the other major technology revolutions we have seen over the past century.
That’s the good news. If this is real, it would be a wonderful thing and there’s some historical evidence that when the market gets excited like this, it might not be entirely a mirage. Now the bad news. If this is an internet-style leap forward, the aggregate incremental increase in real earnings we should expect compared with the normal trend is…10%. Not a doubling, or tripling, but 10%. Naturally, those gains will accrue to a smaller subset of companies at first, but the other lesson of the internet boom is that those gains eventually percolate around because that’s the whole point of a “general purpose technology.”
Have we gotten our 10% yet? Seems like maybe we have.
[1] This assumption is clearly false, but it’s false in transparent ways. Right now, the population is growing faster than the labor force due to immigration. As Baby Boomers retire, the labor force will grow more slowly than the population. Etc. The assumption here is not meant to be uniformly and universally true, but approximately true on average so as to make the general point which follows. To the extent that this assumption is transparently incorrect, we know how to adjust the general point which follows, for the specific conditions.
