Welcome to Economics Applied from the Federal Reserve Bank of San Francisco. I'm your host, Steven Davis, senior fellow at the Hoover Institution. With me is a special guest, Mary Daly. She's president of the Federal Reserve Bank of San Francisco. Welcome, Mary.
Glad to see you again. And thank you for coming here to San Francisco Fed. Yeah, maybe I should welcome myself to the Fed, but in any case, I'm happy to be here. You know, we last spoke in this format at the Jackson Hole Economic Policy Symposium in August 2025.
We talked about central bank communications. I'll come back to that in a second. The setting was a little bit different. Now we're in your studio. Great lighting, great audio. Last time audio was a little more challenging, but we had the Grand Teton Mountain.
Exactly. Exactly. And we had to discipline ourselves not to get distracted by that view.
That's right. That's right. So hopefully we can... Well, I think we did okay last time. We'll try to focus even more this time. So, you know, last time we did have a pretty meaty discussion about central bank communications, and I'm not going to try to review all that, but two things came out of that discussion very clearly, I think. First, the Fed has changed its communication regime dramatically over the decades, even over the course of your career and mine, as we discussed. Second, there are a lot of subtleties that aren't apparent on the surface in how the central bank should communicate with the broader public. It gets into issues of transparency, accountability, credibility, and trust and all that. So for those who really want to get into that kind of broad sweep historical perspective, they can watch our earlier conversation. But what I wanted to ask you about today is something, you know, pretty timely to the moment, which is how, if at all, has the communications regime at the Fed shifted under the chairmanship of Kevin Warsh or is shifting relative to what the regime was under Jerome Powell.
Sure. And you know, you're right. I think in big historical sweeps because that's been the dominant, a dominant development in my career is we went from communicating almost not at all to communicating quite a bit. And what I like to do is opposed to putting it to chairman and chairs is put it to time periods. So in time periods where we have a lot of clarity or we're in the zero lower bound or we have a financial crisis or we have a pandemic, again the zero lower bound, well we communicate a lot and we learn a lot about what communication can do for us. The place where I think we're at now is to step back from all of those learnings over the last 30 years and ask when do we use which tool and are tools just because you have them always good to use or do they move across purposes to your principles of accountability, transparency, credibility and trust. Importantly, I'll add to that list which I absolutely believe in, the part about clarity. You know, we want a high signal to noise ratio. And when we're communicating a lot, but the signal to noise ratio is low, then we have to step back and ask ourselves, is this really doing what we want? Is it building trust? Is it leading to more credibility? Do we have transparency and ultimately accountability? So, right now, I'd say that the, this, the committee as a whole is stepping back. The chairman wants to step back. We want to get external perspectives from the task force that the chairman named and just actually do a rigorous review of not only how should we communicate but what is the right communication at the right point in time. You know, notably other central banks are doing the same thing, right? You know, Madame Lagarde did this, Christine Lagarde did this for the ECB. Others have announced similar ways of thinking and there's a big discussion about should we give forward guidance, should we not? I'll tell you a little bit about where I stand without having all the input from the task force and our discussions is I don't, I think that tools are easy to think you always use them. You know, there's even a saying about that. If I have a hammer, I'm always looking for a nail. But central banks really have to be careful about the types of things we're willing to say with definitiveness because people make plans based on our definitiveness. And if we're not sure that can be at cross purposes to our goals, right?
Okay. So the toolkit analogy is a good one. Just because you have a toolbox doesn't mean everything in the toolbox is suitable for every moment, every situation. Absolutely. So I take that point. You nicely deflected my question about the change in the chairmanship. So I'll just say the follow which you can disagree with as well if you'd like. Let's say that the change in the chairmanship was a catalyst for thinking about it. It may have been an opportune time to rethink what was already underway under some rethinking already. But I also want to pick up on two themes of what you just discussed or my characterization of two themes. One is very clearly goes back to the toolbox analogy. Circumstances change over time that calls for different tools at different times. But I think another subtheme in what you said is especially in the wake of the global financial crisis, the GFC, the Fed engaged in more aggressive efforts to manage expectations through its communications protocols than it ever had before, at least to my knowledge of Fed history. And that may have been entirely appropriate in the initial aftermath of the GFC, but you learned some things and probably in retrospect there were some things you would have done differently. So I see it as both times have changed, conditions have changed, but also there's been a learning process underway about some of the downsides of this very heavy emphasis on communications.
That's an excellent point and I will completely agree with you that Chairman Warsh has been a catalyst for rethinking a variety of things and I think that's a good thing for the evolution of the Fed. You know, the Fed should always be evolving and whether it's a time frame that makes the change a shock to the economy like a pandemic or a GFC or a change in chair, these are good times to say well what do we know and what don't we know and what should we rethink. You know, when I became president even though I love the legacy of the previous presidents in San Francisco I did that very stepping back and saying well what do we want to keep, what do we want to give up and importantly have we refined our responses well enough on a whole host of things. So the chairman's doing that for sure. We're doing it with him. On the point you make about learning, that's one of the characteristics of a big institution that's long-standing that I think it's almost inertial in this way. You have a toolbox that ever grows and you keep using all of them. And it's a really important question to ask when do we use the tools? Why do we use the tools? And should some of them just be discarded. The academic literature also evolves, right? We knew a lot of how to do things in the GFC. Many many papers, many many dissertations have been written that try, many conferences I've attended that really asked the question, are all of those tools ones we would use again? And are some of those tools ones we would never use unless we had a crisis that looked like the GFC? So, I think that's a really, we're ripe for asking those questions and it's appropriate we do. The chairman comes in, he wants to ask those questions. And the only caution I would give all of us is to not attribute too much to whatever a particular person thought before they took the role and contribute more to the catalyst. Maybe everyone's thinking and importantly time period and is this really the tool set we want to go into the next 10 years with. Our circumstances have changed dramatically and we should evaluate tools that fit those times.
Well, so on the toolkit point again, are there particular tools that you are already firmly convinced and you want to share with us here that you want to put back in the toolbox or you want to wait to hear from your colleagues first and let this process.
I do want to wait for details, but I'll give you some topics that I'm thinking about and that will be a nice balance, right? Okay. The topics I'm thinking about are really how do we, are we doing as well for the public as we would like by giving remarks and being... I'm asking myself this, not talking about my colleagues, giving remarks about what I think we'll do next time. I don't tend to do that much but it is a good question about is the next meeting really the discipline that we want to share. I think it's made us more likely to be misinterpreted as data point dependent and more likely to be misinterpreted as short thinkers rather than long thinkers.
Can you give us an example of that?
Yeah. So if you talk about the October meeting which I will tell you that the media would like to ask us about constantly and the public wants to know do I need to do anything before October, right before the October meeting. If you go down that path you miss the part that's really important about policy which is policy acts with a lag. You can debate what you think the lag is, 12 months, six months, 18 months. But there's nobody who thinks we make the interest rate change and everything changes and so we have immediate impact. Right? The reason that's important to remind people of is that that means we have to think forward. Our policy response should be on that forecast. And if we're less certain about the forecast, we should be less certain about the path and we should be less certain about our policy. We'll come back to that when we talk about AI and what the AI... it's exactly on point for that. So this comes back to forward guidance. I really do think that, you know, giving talks about what you're going to do next is almost a form of, it is a form of forward guidance. So what is, when is a good time to do forward guidance? To me, a good time to do forward guidance is when you know where you're going. And a bad time to do forward guidance is when you're in the process of trying to balance the risks, understand the evolution of those risks and importantly then that means we communicate more on how do we think about this and less on what do we think we'll do. So that's a nuance that I think is important, but it also reflects forward guidance in that. The other thing that I will throw out there which is going to put two things together is the balance sheet task force that we've got and the communications task force. I do wonder whether we really need to couple the behavior of the balance sheet with what we do with the funds rate. And it's a question I'm asking myself and I think that's a ripe question to ask for us because we have, the balance sheet is another type of tool to get to the interest rate. But the one that's the speedboat in our...
Balance sheet you mean both the size of the overall portfolio but also its composition.
Its composition and whether we are buying or not buying assets or they're rolling them off. All of that got tangled up because we had one situation, the GFC, and now we've just come forward with that. I think that's ripe for discussion and then, you know, we can get into the details of any of the aspects of how we communicate but do we really want to give point estimate forecasts? Are we better off with scenarios? You know, so these are all questions that I think are ripe for us to consider, address. The only thing I want to share with your listeners and viewers is there are no single right answers. It's really about thinking what is appropriate for now. What could we get better? And then having that humble enough mindset that you're saying this isn't perfect. It won't last another 30 years. It's an evolution and we really are trying to respond to the needs of the public and the conditions of the economy.
Yeah. Let me just make one other observation that builds on that. The public is really many publics and they differ in many respects but in part they differ one way in which they differ is how closely they pay attention to the Fed's pronouncements and how often they pay attention at all.
That's absolutely true. If you are a bond trader, then you're hanging on Mary's every word and every other word.
The sentence structure to ask what I mean.
But if you're just the homeowner who's thinking, you know, maybe once every 10 or 20 years, they're in the housing market, right? Well, they may be, they're going to think a lot about mortgage rates at that point in time, but normally it's something in the back of their minds. So, and there's a whole range in between. So the last point, the last detail on that is in the wake of certain events or certain developments there's a much larger set of the public that is paying attention to the Fed when inflation surges. Obviously if you're sitting in 2018, 2019 we'd been through a long period of let's call it quiescence. Most people weren't paying that much attention to monetary policy. So that's another layer of complexity in your communications challenge as I see it.
No, it absolutely is. And it's one of the reasons that I spend a lot more time answering questions about when inflation is going to come down now than I ever have in my career because we lived through the decades of low inflation, low and stable prices. Now, we've had high inflation above target for almost six years. That's, believe me, there are a lot of people stopping me in the street, in the airport, wherever I am, saying, 'When are we gonna get inflation down?'
Okay, that may not be an entirely good thing for a central banker that people want to stop you in the airport because I think when things are going well, they just walk on by.
They just walk by. No, we went from walk on by to 'That's Mary Daly. Let's ask her when inflation's coming down.'
Okay, very good. So let's transition a bit. We're going to get into AI and how to think about that. But before we do that, I think it would be helpful to lay out some framework for how you think about not just AI but various shocks. And I'm going to put it in a particular way that seems relevant to the present moment. You can expand on my presentation of the issue, but I want to contrast or draw a potential contrast between how the Fed responds to demand overheating, let's call it that, aggregate demand shocks, upward aggregate demand shocks that push inflation uncomfortably high versus an aggregate supply shock that has upward pressure on inflation, an oil price shock, stock, tariff hikes, maybe the AI generated increase in demand for certain kinds of products and infrastructure and so on. So at least as I think about it those you might want to respond differently to those but let me get your sense. So how do you think about demand overheating sources of inflationary pressures versus aggregate supply sources and what kinds of responses they call for from the Fed?
That's a great question and it's an important question. So, let me give you both conventional wisdom, the kind you and I grew up with as economists, and then also just the policy maker perspective from what our tools are effective at and what they're not. So from the conventional wisdom, the standard rule of thumb is the Federal Reserve should look through a supply shock because it's an exogenous shock. It's outside of us, oil prices, tariffs, and its duration is expected to last for a shorter time period than the duration of our policy lags. So if we respond to it, we can't really offset it because we're not controlling the reason it was created. It could be a shortage. It could be just, you know, it could be disruptions in it. It could just be a cost increase. But those roll on and they roll off and they roll through. They also only inflict inflation once, right? Prices rise in tariffs, but they don't keep rising because of tariffs. And so that rolls off and it's not inflationary over a persistent period of time. So it doesn't last as long as it would take for us to do the policy action and we can't really offset it that well.
And you mean the inflationary impulse? Inflation doesn't last as long. There might be a one-time but persistent effect on the price level.
So the price of oil could go up and stay up to $90 a barrel, but it only went up once and that's the inflationary impulse and over time it's now just that's the price of oil. So we don't offset that. A demand shock on the other hand is an economy that is outpacing, spending is outpacing what the economy can produce and that will spur inflation across a broad array of goods and it won't go down until the economy, you either have to increase supply which usually takes a long time or you have to bridle demand and our interest rate is a very good way to do that bridling of demand. We raise the interest rate, spending comes down, people don't borrow as much, firms don't hire as much, spending comes back down, and it gets in balance with existing supply. And the art of monetary policy oftentimes is figuring out what's a supply shock and how does that inflationary impulse, how long will it last and what's a demand shock. The challenge we're in today is that we have a demand shock in AI that might create a positive supply response later. And then we have tariffs. That's the hope, right? And then we have tariffs and oil that have had a one-time effect, but in the case of tariffs, now we're talking about them again because that can be another one. And then we have oil prices that have been going up. They come back down, but then they go up again. And they've been just high and rising. If you look at a longer swath of time for a while and you have to ask the question, at what point does something that was a one-off start bleeding into inflation expectations just because inflation's been so high? And if firms didn't have to pass on the energy price increase originally, maybe they do now. And then of course, I'm going to introduce a phrase. You're probably going to want me to do it, but there's second round effects on oil prices, right? So we have oil prices now, diesel prices are very high and if those continue that and hopefully people will see this, oil prices, it affects the price at the pump etc and then it goes away, you're fine. But if it gets into diesel then that means all the trucking that's bringing goods to families in the US, businesses in the US, that's getting higher so now that's bleeding into the cost of goods and often services and then we have a broader inflation problem. So the thing we're grappling with is do the conventional pieces of wisdom where our tool doesn't really work for supply shocks and they dissipate on their own. Does that really apply when the shocks are overlapping and they're long-lived and now we've got it coming at the same time we have a pretty robust demand shock for AI and is that going to spread into other goods and services or just be narrowly changing the relative price of IT goods?
So does that, so should I take from that that you're less inclined to look through the current set of supply shocks than would typically be the case under the conventional wisdom. Is that correct?
That is and I will say absolutely that's how I feel today. But if you had come, we talked 6 months ago I was very much on the look through part. Okay. And but now just the preponderance of evidence about their duration and where I'm seeing producers a little more willing to pass along prices. We're seeing them pass along things to other businesses. We're hearing from surveys that businesses are more willing to raise prices above where they were. They see they have that opportunity. So that's a little more inflationary in a persistent way than I had hoped. I still hold out some... well, let me put it this way. I still have a very open mind that we could wake up tomorrow and that the war could be over. Oil prices could come down. We could get tariff settlements and we'd have that one-off. And so you don't want to overreact in my opinion and say this is definitely going to be long-lived. But I think we have to be a little more cautious.
So you have adjusted your views in the last six months as to the likely duration of the inflationary pressures coming from these three sources you've mentioned, but you remain open to the possibility that things will in just a few weeks time look more favorable on that front. That's how I'm hearing you.
That's a really nice way to summarize it which means it has implications for policy response. Yes, it does.
So, yeah. And I want to come to that, but I want to ask you about two more things in terms of your framework. One, you haven't explicitly, I think it's implicit in some of what you said, but you didn't explicitly yet mention the potential impact of inflationary supply shocks on expectations of future inflation rates. And maybe that's because you're thinking we're living in a world in which they're pretty well anchored.
I do not. But you didn't talk about... Okay, you did. Okay, let me talk through that because that is absolutely a big part of my thinking. So I do not take that for granted. I think that would be a mistake if we simply said, 'Well, they're anchored, so they'll always be anchored.' That's how you get into dangerous territory. So I agree with that. Absolutely. So one of the things that I do all the time is look at inflation expectations, short run, medium run, long run, surveys, consumers, businesses, market pricing of inflation. And something that caught my attention in the time periods leading up to the September FOMC meeting was that inflation compensation, the compensation market investors are asking for because they're worried about inflation, had risen. It had risen since the July meeting. That to me is a worrisome first sign that you might be seeing more of that, not de-anchoring. De-anchor is when you lose the anchor, but just like water's blowing around, moving around the anchor and you're like this is too much. We got to get something done here. So, I think there's that part. We have to continuously watch that. And importantly, you know, if you look at the inflation risks, we have to be very clear with the public that the risks are all to the upside on inflation right now. You can hope that inflation will come down as these shocks dissipate and you can watch for that but you have to recognize that the pressures are drifting upward and that we might, we have to be in position and ready and willing to do more if more is required.
That makes sense. And well I'm glad you got your eye on the ball with respect to how well...
Oh, absolutely. I have a dashboard of inflation expectations.
Okay. Okay, good. Does it like flash red if things go wrong?
Oh, okay. Good. Okay. The other thing I want to ask you, I'll prompt you on this. It strikes me as an important subtlety, but I don't actually hear central bankers talk a lot about it these days, which is I'm going to put out there that there's evidence that nominal wages are sticky downward more than they are upwards. Okay? And there's I think over many decades...
I've read two papers on that very thing. So I completely... so I've worked in that area too a bit.
So and there's lots of... right. So let's start with that as kind of an assertion of how the world works. But then the following question arises in my mind. If you have an inflationary supply shock there's a sense in which we're poorer. Certainly if you think of tariff hikes or it's a little less clear now because the US is now a net oil exporter. But for a country that's a net oil importer, oil price shocks mean our real incomes are lower.
There's a sense that as a consequence, real wages have to adjust downward if you want to equilibrate labor markets. So you need some downward adjustment of real wages in response to at least some inflationary supply side shocks. On the other hand, for demand overheating shocks, depending on the genesis of those shocks, the history by which you got there, there may or may not be a need to adjust the overall level of real wages. So, that's a different reason than we've talked about so far for potentially different responses by the central bank to demand overheating and inflationary supply shocks. And I just wonder how you think about that.
You know, I think it's certainly part of what we see happen and you're seeing it now. The supply shocks have moved the rate of inflation above the rate of wage growth and real wages are not rising for many people and that is workers feeling like they're falling behind. Now in some way that is the supply shock moving its way through and it means that if we did more we would simply further that issue and so if it's already happening us doing it just makes that issue worse and you can end up breaking the economy when the dynamic that you already want to, that needs to occur is already occurring. Right. So in demand shocks we just want to restrain the overall...
rate of growth to match the overall supply in the economy so that we don't get that constant demand-supply mismatch. What happens to wages is really up to the market equilibrium to determine. So if you don't need labor share to be constrained in that component because you end up with an overall demand shock, we simply slow the economy. How the economy shakes out is up to the economy, essentially. But on the supply side, we could easily make the issue worse if we restrain the economy because we couldn't get inflation down, but we would definitely restrain the labor market and real wages likely fall more.
Right. But there is some tension because you need real wages to fall to re-equilibrate the labor market. And it's very painful to have slow or negative real wage growth, but it's probably more painful to have high unemployment. And in some circumstances, that's kind of the conundrum.
So certainly if you would see that, right, in people's responses to their world, people would rather stay employed and get a little less than lose their jobs and have nothing. We see that again and again.
I think that is the weight of the evidence. Yes. But it makes it very challenging for a policy maker because nobody likes slow real wage growth or a real wage cut. It's almost like the instinct is the opposite thing then. You would say, 'Well, the labor market looks weak. Look at wage growth.' I don't look at it that way. I look at it as there's no inflationary impulse coming from wages.
No, that point's well taken. You've talked about three shocks multiple times so far. There's the tariff shocks, the oil price shocks, the AI. I think with respect, I'm just going to make an assertion. I want to focus on the AI because I think that's the most subtle one and maybe the most...
I agree with you, and it's not something in conversation people often treat with subtlety.
They have strong opinions about how it will shape up, but it's highly subtle.
I think with respect to tariff shocks, oil price shocks, we can agree there's a lot of uncertainty. That uncertainty is outside the hands of the Fed completely. You should factor that uncertainty into account, especially when you think about your point earlier about kind of long variable lags. You know, you don't know what oil prices are going to be like in six months. So you want to be prepared for many contingencies.
But the AI thing is trickier. And you've already alluded to part of the reason why, which is the upside. The promise of AI is that it's a massively favorable supply shock coming down the road. But the current manifestation is happening on the demand side through the demand for chips, through the buildout of the AI infrastructure and data centers and so on, through the investments that many firms, not all firms, are making and how to integrate these AI technologies into their business operations. So that's where the rubber is meeting the road now.
Okay. But you got this potential for really a generational, and some people think more than a generational, positive supply shock. So how do you roll all of that into the framework and the uncertainties that you've been talking to us about?
Absolutely. And I think a lot about this. I work on AI a lot and how it affects the economy. And I think a starting point is to remember that the window for monetary policy is one to three years.
If it changes the growth rate ten years from now, then that doesn't really matter for the experiences that we have to manage full employment and price stability today. So that's the first thing. And so I can be pretty bullish on this being a transformational technology, a general purpose technology that looks like the kinds of effects we saw with electrification.
Right. I could be very bullish about that, but it doesn't affect how I'm deliberating in my mind about the one to three-year period over which monetary policy really matters, and that's where really our decision.
That's a key point. That's a key point. So then what's going to happen in the next one to three years? And there I would say we're dominated by uncertainty. We don't know. Let me tell you what I mean because I think you set it up well. Right now, what do we know? We know that the hyperscalers are building and they're not stopping. We know that there's a huge demand for chips, you know, GPUs and memory chips and then all the other ones. We're hearing again and again from our contacts that if they just need ordinary chips, as they're starting to call them, that are not the ones that Nvidia would like or that the LLMs need, then it's not that that's still competing against them because if you're a maker of chips and you have any capacity to make the more profitable ones...
You would move to that capacity.
So it's clogging up a little bit the supply chain there. So then I ask, how much is it clogging up the supply chain? Because that's definitely inflationary. It starts getting into cars and coffee makers and, you know, all the things, you know, self-driving cars all the way to your espresso machine that now has a chip in it to read the coffee pod. That's a huge array of consumer goods, not to mention computers, iPhones, etc. So we asked them and they said, 'Well, we're already starting to re-engineer our products to use the chips that we have for a longer period of time. So we were on a cycle where you're depreciating them quickly, but what if we don't have to?' That's the economy responding to things on the demand side. So that's an important subtlety. So there's substitution responses on the demand side that could ease these pressures.
And so as a central bank, you know, we don't want, I don't want to go in and say, 'Well, let us do it,' if the private sector, the marketplace is going to do it. So if firms are going to say, 'I can see the writing on the wall. My chips aren't going to be as available. So I've got to figure out how to use them with greater efficiency. I've got to start remodeling my product designs, which you can do more easily now because you have AI,' they say. And so they tell us. And so I think that's a place where we don't want to necessarily run to, we have to offset this demand impulse because we don't know how much will be done by demand restraint or changes on the firms themselves. On the other side, I think, you know, if you had asked me a year ago, I would tell you that my outlook for when firms might start to see productivity gains that are more sustainable than just the 'I used an off-the-shelf software package that I turned on in my suite of already vendor operations and a few people used it. We got a little more efficient.' We're beyond that now. We have been spending the last, you know, six weeks, well, we do this every time, but in between each of the FOMC meetings we do a lot of meetings with contacts, boards, councils, etc. It's changing rapidly. It's changing from three months ago. Firms are now starting to say, 'Oh yes, I brought this to the front of house. I'm fronting my, you know, my consumer-facing products, my business-facing products. I'm now using AI for a whole set of things that I didn't think was possible last year and I think it's even more possible when you use inference.' And it's interesting, you know, I think people thought agents were going to be the magic weapon. It's really inference, using them to make inferences to do things. You can do things faster, better, less expensive.
Give me a concrete example.
A concrete example is I need to read a document and I need to imagine what I would write in response if I'm in a vendor relationship. Well, then they've got to look at all your documents, infer what you would have said given your history, and also you can tell it, look at the forecasts of things, what should I be positioning for? And now firms are just in a better place to bring those negotiations because both sides are doing it right, bring those negotiations to a settlement or decision faster. Well, that saves loads of people and then those people can go and do more creative, accretive to the value stream things. And so our firms are now telling us, 'I see revenue generation opportunities, not just cost savings.'
Yes. So what I hear you now is saying...
