Inflation targeting has been the goal of central banks for decades now, either implicitly or explicitly. Of course, they say that they have multiple goals, but they give most attention to the price level. That’s probably because it is easy to measure and more directly related to their activities than the unemployment rate and private financial activity. Price level targeting and inflation targeting are not quite the same thing – but I’m not in favor of either. This post describes what happens when the central bank targets the price level and offsets other changes in the economy in order to achieve their goal.
Volatile Capital Prices
If consumer prices are constant in the face of productivity shocks, then capital prices adjust instead. Capital is just goods that create other goods. If capital becomes more productive, then that means being able to produce more at given prices or being able to produce given quantities at lower costs. The demand for capital is ultimately determined by how profitable it is. This includes the costs of maintenance, the price of output, and the capital’s productivity. All else constant, changes in the revenue produced by the capital for the firm affect the equilibrium price of capital.
If NGDP is constant and capital productivity improves, then output rises and consumer prices would fall. With unit price elasticity of output demanded, the total revenue of the firm remains constant and the nominal capital price does too. The 19th century gold standards had plenty of problems. But one feature was that long-run consumer prices fell and long-run capital prices were more stable.
If, instead, the Fed stokes NGDP to prop up consumer prices, then the firm’s revenue rises. Demand for the capital rises and so does its price. The opposite occurs when there is a negative productivity shock. So, capital price volatility is the trade-off for consumer price level stability if productivity changes. We can argue about which price volatility is better in regard to inequality, investment planning, financial stability, etc. But my strong low-hanging fruit point is that consumer price volatility just pushes the equilibrating mechanism to a different set of prices.
Volatile Income
As I said above, if the Fed wants stable consumer prices, then it must offset the impacts of productivity shocks with changes in aggregate demand – its only lever. Negative productivity shocks are offset with aggregate demand contractions.
People act like they have adaptive expectations. Of course, people differ by how forward-looking they are. So, on average, their expectations are formed by what happened during the prior period or the last time that they observed a similar circumstance. Why does this matter?
Have you heard about the abundance movement? It basically says that we should enact a mix of regulatory and supply side reforms in order to produce more for everyone, especially the least economically advantaged. The reforms extend to the housing market and ensuring adequate housing.
There’s an argument that building any housing, even at the high end, can reduce the cost of shelter for everyone – even people who would never live in the newly built housing. The idea is that high income people switch to the newly built housing and leave less attractive housing. Someone else in that high income bracket snatches up the older place, leaving their prior housing vacant. The vacancy shuffles around high priced rentals until, ultimately, the vacant rental price must fall in order to attract a renter, such as someone further down the income distribution. Then the entire process continues, with the game of vacancy musical chairs working its way down the renter income distribution.
The more overlap that there is between housing consumption choices the quicker there is an impact on lower priced housing. If you think that high income people consume higher priced housing, then you might think that there is a substantial difference between housing consumption choices and that it will take a long time for this ‘trickle down’ to get to the people who need it most. If income groups compete more for the same housing, then the effects on price will occur sooner for the lower income people.
How much Rental Overlap is there?
Miami, Florida has some of the highest housing costs in the US. Below is a histogram of annual rental costs in Miami by household income quartile (ACS 2024). I restricted the data to positive incomes and rents and the highest rents are censored down to $98.4k annually. First, we can definitely see that the highest incomes (quartile 4) have the most censored annual rents and that the 1st income quartile (lowest) has the most annual rent payments nearer to zero. So, the histograms make sense in that way. But I was surprised by how much overlap there is. Different income quartiles are consuming many units in the same price range!
Back in May 2024, I wrote about grocery prices in 2024 compared to 1980. Relative to average wage increases, almost everything was more affordable — the one exception was bacon.
Grocery prices have continued to climb since 2024, but so have wages. What does the comparison look like now? Well, I have good news for bacon lovers:
Figure 1
The chart shows the change in relative affordability, as measured by how many minutes of work at the average wage it would take to purchase the item (using consistent product sizes and weights). As I wrote in that 2024 post, these are not items that I cherry picked. These are all 24 grocery items where BLS has price data in both 1980 and 2026 (out of about 150 items total). Perhaps there is some survivorship or selection bias as to which items are available in both years, but looking at the list this seems like a pretty reasonable shopping cart for a typical consumer (well, maybe there aren’t buying all the meats every week, but probably every month). The prices are updated through July 2026, with the CPI data just released this morning.
But consumers probably aren’t thinking about prices relative to 1980. Their time horizon is likely shorter. What if we made the same comparison to 2026 using the 2019 prices, which is right before the pandemic and within most shopper’s recent memory:
Figure 2
Relative to 2019, things do look quite as rosy. Some items are “on sale” in terms of affordability, but a lot of items aren’t, especially a lot of proteins. And no doubt many consumers will focus on the items that are less affordable, rather than those that are more affordable (and even for these items, the nominal price is higher, so consumers might still be frustrated).
It is important to note that this basket of 24 items isn’t a perfect representation of all the items consumers purchase. Compared to the CPI “food at home” index, average wages have actually increased more since the beginning of 2019. But consumers are right to feel that beef and few other items are much less affordable than 2019, even if over the long run they are more affordable.
Leopold Aschenbrenner is a very bright guy. Born in Germany to physician parents, he skipped enough grades to graduate from high school at age 15, allowing him to enroll at Columbia University in 2017 at that same age. He went on to graduate from Columbia at age 19 as valedictorian with a degree in economics and mathematics-statistics. An econ prof at the time said his “record of scholarship exceeds that of any student in the department in the previous 20 years.” The Mercatus Center, a think tank at George Mason University, recognized his potential, awarding him an Emergent Ventures grant.
Leopold Aschenbrenner, Columbia Class of 2021 Valedictorian
After graduation, Aschenbrenner moved into the effective-altruism research and grantmaking world, making notable contributions in various ways. In 2023, he joined the newly created “Superalignment” team at OpenAI, that was charged with making conceptual and engineering progress on aligning systems “much smarter than humans” before such systems were built. The next year he was discharged by OpenAI; the company said it was because of a security leak, but his version (which I find more credible) is that he was ousted as retaliation for circulating an internal memo arguing that the company’s protections against model-weight theft and algorithmic exfiltration were inadequate for AGI-relevant work.
Two months after his departure from OpenAI, he self-published Situational Awareness: The Decade Ahead, which you can download here. This monograph synthesized scaling-law extrapolations, geopolitical analysis, and AI-lab security commentary into a single forecast: that the largest AI labs, on current trends, will plausibly reach AGI around 2027 and that an intelligence explosion to superintelligence could occur in the subsequent few years. This work went viral on Wall Street, and before you can say “monetization”, he was leading a hedge fund named, appropriately enough, Situational Awareness. There he put into practice his convictions that the demand for compute would be voracious, and would be limited by physical constraints such as electricity and chip fabs.
Thus, in his fund he went long companies like SanDisk (memory fab), Bloom Energy (makes solid oxide fuel cells), and Nebius (builds whole data centers). Very long, in fact, with leverage reportedly as high as 400%. He tried to hedge this long book by shorting software companies which are viewed as vulnerable to disruption by AI. This strategy worked fabulously for a while. His assets under management (AUM) climbed to $45 billion, with returns in the first 6-7 months of 2026 approaching 400%. Not bad for a 25-year-old.
But then, genius failed (yet again)- -in a stunning reversal, the market rebelled in late July against big AI capex spends, dumping memory fabs and infrastructure, and bought into the maligned software (SaaS) sector. So, BOTH his long and short legs went against him, followed in due course by the dreaded margin calls. Aschenbrenner’s $45 billion shrank to a measly $10 billion in a matter of days, as he was forced to sell off his public equities at a discount to those friendly capitalist sharks at Citadel. Wise old heads wagged, saying, yup, this sort of Black Swan event always happens sooner or later, and you then get carried out on a stretcher if you run a highly leveraged bet that is not truly hedged.
Down, but not out – – The word on the Street is that folks with money to invest are already lining up to entrust more bazillions to our altruistic wizard. We have not heard the last of Leopold Aschenbrenner.
The Congressional Budget Office released a series of posts explaining questions they have about major federal budget policies that past research does not adequately answer. For any economist looking for paper ideas or for a way to influence policy, CBO’s posts are a great place to look:
At AEAs this year Heidi Williams emphasized how huge bills like permitting reform are being discussed by Congress without much research to inform key aspects of the bills, so CBO & some Congresspeople would genuinely like to see your work on these questions if it is well done
This also your regular reminded that I maintain a page of economics paper ideas. Until now all the ideas there have been my own, but I will be adding links to pages where others share their own paper ideas, starting with CBO’s.
Figure 1 is one way to look at the problem. It shows the cumulative 7-year inflation rate for groceries, going back almost 100 years. 7 years is an arbitrary time period, but I think it makes sense: it can reasonably be described as “recent memory”; right now it encapsulates the period going back about 6 months before the pandemic; and it has a few time periods of around 100% grocery inflation and a few with close to 0%.
Figure 1
First things first: grocery price deflation over a 7-year time horizon is highly unusual. The only time it happened was the 1930s, a time when you had general price deflation and groceries followed that pattern. It was also a pretty bad time for the economy and society. I’m not saying you can’t have general food price deflation with a major depression, but it doesn’t show up in the historical record going back over 100 years.
Now to the present: the most recent 7 years look pretty bad. In absolute terms, 33 percent grocery inflation is above the long-run average of 25 percent, and definitely above the average of the last 40 years of 21 percent (for most adults, the past 40 years is as far back as their memory goes in terms of being acutely aware of grocery prices). Yes, there have been a few time periods with higher grocery inflation, notably the two World Wars and the 1970s.
But the really important context is the 7 years prior to the pandemic, when grocery inflation was so low (4-5 percent every 7 years) that it probably felt like 0% to most people. That was the recent experience people had become accustomed to before the pandemic. The only other time since the Great Depression it was that low was the late 1950s through the 1960s — though that 15-year window is bookended by two periods of around 100% grocery inflation!
Under Jerome Powell, a typical FOMC meeting had become almost a market event in itself. Traders didn’t just care about the rate decision. They dissected every word of the statement, every sentence of the press conference, and especially the “dot plot,” looking for clues about where rates might be six months or a year from now. The Fed wasn’t simply setting monetary policy—it was guiding expectations. Markets often moved as much on hints about future decisions as on the decision itself.
The first two FOMC meetings under Kevin Warsh have felt very different. The dot plots are gone. Forward guidance has largely disappeared. Instead of trying to signal the likely path of policy, Warsh has repeatedly stressed that the Fed will respond to incoming data when it arrives, not commit itself to forecasts that could prove wrong. At his latest press conference, he described avoiding forward guidance as “prudent” given current uncertainty, while reminding reporters that “There is no soft or alternative inflation target—only 2%.”
This is a huge change in communication style, which is having real world consequences.
Warsh long argued that forward guidance can box policymakers into decisions based on yesterday’s forecasts instead of tomorrow’s realities. That is, once their tentative plans had been put out in public, there was a psychological bias among Fed members to lock in on those projections, which would inhibit their ability to rationally interact with the most recent data and situation. So now, rather than telling markets what the Fed expects to do, he wants investors to make decisions based on fundamental economic conditions, knowing that the central bank will react only after the facts justify it. At the latest FOMC meeting he said, “Market participants are learning to play the ball, not the referee—and market prices will continue to respond in the direction and magnitude they see fit. This is, in my view, a change for the better—and we are just getting started.”
That approach chips away at what investors have come to call the “Fed Put”—the belief, built up since the 2008 financial crisis, that the central bank will fairly quickly and forcefully step in to support markets whenever things get rough.
If that belief fades, financiers may think twice before taking excessive risks. Leverage becomes more dangerous if there is less confidence that easier monetary policy will quickly arrive to cushion losses. Risk premiums may better reflect actual economic uncertainty rather than expectations of future Fed support. That is the possible good side of Warsh’s more hands-off approach. Ideally, business people will exercise more prudence on their own, lessening the odds of financial catastrophes that would require Fed intervention.
On the other hand, markets hate uncertainty, and less guidance means more volatility around Fed meetings. I think Powell tried to use sheer talking (jaw-boning) as a tool to influence market rates, lessening the need for the Fed to actually employ its blunt instruments there. Warsh seems to have taken that tool off the table.
Also, I think some (not all) the causation for the rise in 30-year Treasury bonds to twenty-year highs, and of home mortgage rates to one-year highs accrues to Warsh. First, by eliminating dot plots and forward guidance, he has increased uncertainty about the future path of policy. Investors can no longer confidently assume the Fed will ease at the first sign of economic weakness. That uncertainty can raise the term premium, pushing long-term yields higher.
Second, if markets believe the “Fed Put” is weaker, they may demand higher yields to hold long-term bonds because they perceive less protection from adverse economic or financial shocks. In other words, investors require more compensation for risk.
Whether today’s higher long-term rates are a healthy reflection of economic realities, or an unhealth drag on growth, is a matter of debate.
Abstract: Language models increasingly settle real resource tradeoffs on behalf of principals yet their economic preferences remain unobserved. We demonstrate their generation rule is isomorphic to the random utility model of discrete choice. This allows internal logit scores to structurally identify preferences. Estimating risk attitudes across twelve models in a portfolio task reveals universal but heterogeneous risk aversion. Although models reject strictly dominated options, their elicited preferences fail invariance tests and violate the independence of irrelevant alternatives across varying experimental prompts. Finally, fine tuning establishes that a principal can explicitly engineer a target risk attitude.
I hope you will refer to the manuscript for details, but I will share one picture here. This is panel (a) of Figure 3: Empirical indifference curves for open-weight models mapped over the portfolio space.
In simple language, what the red/blue picture shows is that the Qwen language model is picking the portfolios that offer more money (in expectation, with a distaste for excessive risk). That’s basically what a rational actor should do. We find that the language models make fairly consistent choices and rarely violate the monotonicity requirement for a well-behaved utility function.
How we describe this figure in the paper: “Starting from a base bundle with expected return µ = 10 and risk σ = 30, we sweep over the dense grid of alternative portfolios from our experimental protocol and record the position-corrected logit gap between each grid portfolio and the base. The yellow dashed line overlays the indifference curve implied by the mean-variance structural estimates, and the heatmap colors encode the sign and magnitude of the logit difference, with blue regions preferred to the base and red regions dispreferred. Several patterns emerge from these plots. All six models produce upward-sloping indifference curves, confirming that higher risk must be compensated by higher expected return.”
We think this basic research on behavior is important, for alignment research and for business applications with delegating work to AI agents. The first question to ask, before testing whether we can impose our preferences on AI agents, is whether those agents have preferences at all in a consistent sense.
Suggested citation: Buchanan, J., & Foster, J. (2026). The innate economic preferences of language models [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2607.26288
GDP growth data for the second quarter of 2026 comes out tomorrow. As I have been doing for the past several quarters, here is an update on two model forecasts (Atlanta and NY Feds), betting market implied estimates (Kalshi), and an average from a survey of economists (WSJ). Yellow shading indicates which forecast was closest to correct in each quarter (green is if two forecasts were about the same).
In the past two quarters, the WSJ survey has been the best predictor. The Atlanta Fed GDPNow model used to be my favorite, but it has performed pretty poorly in the past 3 quarters. As I have discussed before, an average of the Atlanta Fed and Kalshi was better than any single predictor. I continue to include the NY Fed estimate, even though it seems to be a very terrible predictor, because some people like to talk about it.
The Atlanta Fed, Kalshi, and the WSJ survey are all showing very similar estimates for Q2. If I was a betting man, I would bet on 1.8% for the BEA advance estimate.
Zachary Bartsch: “A skill can be just plain text written conversationally, it can be a list of rules, mathematical expressions, or even the foundational code that you want your AI to readily modify and apply. Essentially, saying ‘skill’ is the same as saying ‘pre-prompting’ with various degrees of specificity. Rather than writing a prompt each time, you can recycle a set of prompts that you’ve stored in a file. That’s all that a skill is.”
Plus, Zachary provided some useful history of “Explainer text files”
“Not a single one of these foods is an ‘incomplete protein’. Yes, the mass that you’d need to eat differs, but there is not much that is exciting about legumes and grains as a combination.”
Anyone who has gone grocery shopping in 2026 knows this is the year of protein.
Out of curiosity, I checked the price. Within a week of this post, the price of silver had actually gone up. But after a final peak in late January, the price has declined. As of today, it is down from any of the prices posted in January of 2026.
“Gleeful researchers, competitors, and hackers promptly downloaded zillions of copies. Anthropic issued broad copyright takedown requests, but the damage was done.”
He talks about the referee process, since that is where the main decisions happen, as much as the “writing process.” No one has all the answers, but Mike is doing us all a favor by getting some of this real talk out in the open. Please comment if you have more ideas on where to go from here.
I’ve seen chatter about this topic on Twitter/X, but I’d love to see some more blog posts from tenured folk because it helps with the hidden curriculum problem.
“Would you have guessed that in the “good old days” of the 1950s and 1960s, the average US family was spending 30-40% of their income on food and clothing, something that today we spend barely over 10% on? To understand the challenges we face today, it’s important to have the context of how bad the past was.”
Jeremy has been telling this story for years. Interestingly, world cup tourist discourse seemed to push a few more people over the fence (why hadn’t they just read our blog?). Most Americans are rich.
Since the ChatGPT launch, I have heard conflicting stories on the impact of AI on white collar jobs such as software engineering. There have been layoffs and, for example, ex-Meta employees who struggle rematch in at their old salary. I have also heard claims that the demand for software engineers is actually increasing, perhaps because AI makes them more productive.
The problem echos Makowsky’s post, which ultimately rests on readers and the referee process. I like to say “readers are that which is scarce,” meaning that it’s not difficult to produce writing.
Blogs are not niche anymore. More people than ever, including many researchers at top schools, have decided to start a Substack. Of course, peer-reviewed and prestige-published research still has a primary place in the discourse. Many of the blog posts are ABOUT the primary objects of research.
I saw something called InTheWeights in 2026 that made me think folks at research schools might be strategic in starting to blog now. ChatGPT reads our blog. One reason I think that to be true is that some of our reader traffic comes from ChatGPT.com and Claude. I think our work is getting repackaged as LLM answers to millions of people, some small percentage of those answers provide attribution to us, and then a small sliver of those answers results in users clicking over to us as the primary source for an answer.
It will be a long time before tenure decisions are based on where you are In the Weights. But our crew would do well on that metric. Our work is legible to AI because we have been blogging ungated here for years.
Economist Writing Every Day 