Thomas Drechsel on the Effects of Political Pressure and Identifying Monetary Policy Shocks

Thomas Drechsel is an assistant professor of economics at the University of Maryland. He joins David on Macro Musings to talk about the political pressure on the Fed and the new ways to measure monetary policy shocks. Thomas and David also discuss fiscal and monetary dominance, the impact of political pressure on inflation, why we should care about central bank independence, and more.

Check out our new AI chatbot: the Macro Musebot!

Read the full episode transcript:

Note: While transcripts are lightly edited, they are not rigorously proofed for accuracy. If you notice an error, please reach out to [email protected].

David Beckworth: Thomas, welcome to the show.

Thomas Drechsel: Thanks, David. It's great to be here.

Beckworth: It's great to have you on. You've got some really interesting work. Your research has dealt a lot with measuring shocks, innovation, identification issues using this narrative approach. So, there's always something new being developed in the academy, and it's also very relevant, because there's a lot of discussion about Trump and the Fed and what's going to happen after the election. So, you've got some really neat papers that we're going to get into in terms of political pressure and monetary policy shocks. But before we do that, tell us about your journey into your career and how you got here.

Drechsel: Sure. So, I grew up in Germany. My accent gives that away quickly. In high school, I was interested in a lot of things— In math, sciences, history, languages. I was sure that I wanted to go to college, but I didn't quite know what to study, and it was difficult to decide. One subject that was high on the list was medicine. So, in Germany, the system is a little bit different in that you decide pretty early what you study. So, in the case of medicine, you'd study that from the get-go. And so, that's probably what I would have studied, but then it also turned out that I was one of the last birth cohorts in Germany that had to do either military service or civilian service as a male citizen. So, you either have to join the military for, I believe it was nine months, or do work in social services, and that's what I did.

Drechsel: So, I worked at a hospital for quite a while after high school, and that was a very formative experience because after that, I knew that I didn't want to study medicine. And the reason was that— I have a lot of respect for nurses and doctors and the work that they do. I think that it's great, and I'm grateful that there are these people doing this work, but I didn't find it intellectually stimulating in a way that I wanted from my job. I had the feeling that I would give up on my interest in history and politics and maths, and then I had this idea that [in] doing economics, you can still do all of those things in a mixed way. That's how I saw it, how I expected it, and I think that it turned out to be the case. And so, I think that I made the right choice there.

Beckworth: Absolutely. How did you pick macroeconomics over micro?

Drechsel: Right, so, I then went for my undergrad to the University of Frankfurt, to Goethe University, and that was a great location and a great time to study economics. It's where the ECB is, where the Bundesbank is, and it was during the financial crisis, actually. So, those were topics that we got interested in. Then, also, my intermediate macro professor was Thomas Laubach, who spent most of his career at the Fed, but he had a short stint at Goethe University. So, he was my intermediate macro professor. I really enjoyed his course and got interested in monetary policy and in macroeconomic issues.

Beckworth: So, this is the Laubach of the Laubach-Williams R-star measure. I believe he headed the Monetary Affairs Division at the Fed for a while [at the] Board of Governors; great accomplishment, great researcher, so, wow, small world.

Drechsel: And he sadly passed away a few years ago at a relatively young age, but he was great.

Beckworth: So, that brought you into macro, and now you’ve got your graduate degree. Now, you're at the University of Maryland, you're doing research. And you're doing research, again, on a very interesting topic of political pressure on the Fed. And we talk about this a lot on the show, and I'm sure listeners think about it a lot as well, but you're taking a very unique and, I think, novel approach, using the narrative approach to deal with it. But maybe before we get into your paper, and we'll spend a lot of time there, but before we do, maybe just lay out the basics. Why should we care about central bank independence in the first place?

Why Should We Care About Central Bank Independence?

Drechsel: Great question. The motivation for doing this work is a growing concern that the Federal Reserve gets politicized. We know that Donald Trump, in his presidency, pretty openly criticized the Fed for keeping policy too tight. We know that inflation was high, so, perhaps, naturally, the Fed gets more scrutiny. Debt levels are high, so the temptation for politicians is strong to influence central banks. Appointments at the Fed are politicized a lot, so there's a lot going on in this space, and in my view, it's important to study that. It's actually pretty challenging to study it, because political pressure is a somewhat fluffy concept. We need to define it properly. We need to measure it properly.

Drechsel: That's difficult, and those are things that I do in the paper. And in terms of what we knew about the effects of political pressure, it basically comes from two separate literatures. I'm talking about empirical insights now. So, one literature is older work that uses cross-country comparisons and cross-country regressions. So, Alesina and Summers is a key reference that listeners might know. So, the idea is that you assign a score of central bank independence to different countries, based on expert surveys or something like that, and then you correlate it with inflation. And the typical finding is that countries with more independent central banks enjoy lower inflation. That's one approach. That's cross-country. That's useful.

Drechsel: The other one is coming from the literature on fiscal-monetary interactions. So, here, there is a notion of political pressure in the sense that there are regime switches between a fiscally dominant and monetary dominant regime. And there’s political pressure in the sense that if we have a fiscal regime, somehow the fiscal authority entices the monetary authority to accommodate public spending, to accommodate deficits, and that literature has a theoretical branch that's very insightful. It also has an empirical branch where these models of interactions are estimated on macro data, and these regime changes are basically backed out from the model and the data.

Drechsel: So, we have cross-country work, and we have work that takes models to the data. What I do in this paper is a bit different. I try to look at the US economy through time and then identify changes in political pressure. The way that we would identify shocks to monetary policy or fiscal policy, I try to find a way to identify shocks to that and then construct impulse response functions to those shocks, the way that we would do it for monetary or fiscal shocks.

Fiscal and Monetary Dominance

Beckworth: I love the fact that you're still using VARs or Bayesian VARs or versions of that. You also use ridge regressions to, I guess, get at the shocks, in another paper, but great work. Let me just park for a minute in this discussion of fiscal pressures or political pressures. So, you mentioned the model approach where you take it to the data. I have a colleague here, Eric Leeper, who's real big on fiscal dominance, monetary dominance, and these regime changes that you mentioned. Is there a way to know when we're slipping from a monetary dominance regime to a fiscal [dominance regime]? I would think, for example, that a clear-cut case would be World War II, in the US at least, where the Fed was doing everything it could to support the war effort, so it pegged interest rates.

Beckworth: The yield curve was pegged. It was buying up lots of debt. That, to me— okay, it's pretty obvious when we're in it, but more recently, people talked about [how] we're getting closer and closer, and maybe you could even argue in 2021, maybe 2021, when the Fed had rates really low, we're still buying up lots of assets, that that could be a form of fiscal dominance. Again, fighting a war effort, public health war effort, so maybe it's justified, but are there any ways to know when we're getting close, or is it just happening, and whoops, here we are?

Drechsel: It's difficult to do that in real time. You can look at what's going on in the world, and you can apply judgment, or you can use a formal model and macro data to back out such a regime change, or you can, in some way, try to find measures that tell us something about that directly. That's a little bit the motivation for my paper, although I don't really frame it in terms of these monetary-fiscal regimes anymore. It's sort of where I came from with this work. I read a lot of the work by Eric Leeper and others in grad school, and I always had this idea that maybe we can do more there in the data. And that's sort of where this story comes from.

Beckworth: So, we're talking about an idea that we haven't named explicitly, but the fiscal theory of the price level, and I think maybe, more generally, the consolidated government budget constraint, where we see that monetary policy and fiscal policy are linked. You can't really separate the two, and then how does it get manifested and stuff? So, this is all very interesting, and, as you alluded to, very relevant, because we have these huge debt levels. It looks like we're going to continue to run big, persistent primary deficits in the US. One last question on the current context before we go back to your paper. What is it like in Europe? We talk about all of these pressures in the US. Are there any concerns in Europe and the ECB for fiscal dominance pressures?

Drechsel: I think that there's a discussion around that pretty much everywhere. I think that one interesting aspect in Europe that is something I would really like to study further is that, in Europe, there's a little bit of, also, pressure going the other way in the sense that the Fed doesn't comment so much on fiscal policy. But in Europe, there has been more of that since the Euro crisis that the ECB says. They always frame it in terms of the monetary transmission mechanism and things like that, but they tell governments, “You need to get your house in order.” So, if I were to study Europe, [then] that's a place where I would start. That's a little bit different from what we're going to discuss here, but that's just something that I can highlight there.

Beckworth: So, maybe Europe is a case of strong monetary dominance versus fiscal dominance. One last thing on this. It's a need to frame this. Monetary dominance is a world that most people think that we're in. Most central bankers would say, “Oh, we're in a monetary dominant world, because that means that we're making a difference. We control the price level, and we're active in the fiscal dominance world.” The Treasury or the fiscal authority is the active one and leading, and they're setting the price level, and the Fed would be keeping the government solvent. And so, the roles flip, and there are two corner solutions. So, I'm sure that there's some gray area between the two where we may be getting close to. But let's go to your paper, and you've kind of touched on it already, but you look at a sample from 1933 to 2016. So, tell us about this, because you're taking the narrative approach, and I'll say out of the gate, this looks like a lot of work that you put into this, to collect all of this data, all of those statements. Tell us the story.

Estimating the Effects of Political Pressure on the Fed

Drechsel: In the paper, I study what happens when US presidents exert political pressure on the Fed, what happens to the economy. And I do that with a new data set that I created from archival records, a historical data set, and I combine that data set with a narrative identification approach. So, let me tell you a bit about [the] data set. The idea is that I want to find some way to measure when politicians pressure the central bank. Now, the idea is pretty simple. I looked at how often do they actually personally interact. You can do that for former US presidents by accessing their daily calendars. US presidents, after the presidency ends, they get a presidential library. Say, the JFK Library is in Boston.

Drechsel: Those libraries store documents from the presidency. They also act as museums and are pretty interesting to visit. Part of the documents for each presidency is the daily calendar of the president. That's a detailed account of what the president did each hour or even each 10 minutes, in some cases. So, I, and with the help of some dedicated research assistants, went through all of those schedules, starting with FDR in 1933— That's when the schedules start. If you go back earlier, you don't have them— going all the way up to Obama. And then we went through all of those schedules every day and found meetings that were with anybody from the Federal Reserve. For some presidents, that's pretty easy to do.

Drechsel: Some of the libraries have a digital representation of these schedules. [They even] come as a database and you can search them. So, that's easy. For some other presidents, in particular for George H.W. Bush, the situation was different, and my research assistant had to go to College Station, Texas, go physically to the library, and go through all of those schedules manually. Then, what we have is basically every meeting that we find in those schedules between presidents and Fed officials. Now, those meetings don't necessarily reflect political pressure. They can meet for various reasons, and that's why I combine it with the identification strategy. But let me defer that a little bit. I'll talk about identification more.

Drechsel: Let me just tell you a little bit [of] what we learned just from the raw data. What is interesting is that they meet, all in all, pretty frequently, [and] more often than I would have expected. So, we found that from 1933 to 2016, there were more than 800 meetings between presidents and Fed personnel. Most meetings are with the Chair. More than 90% are with the chair, but they occasionally meet with the Vice Chair, or a governor, or the New York Fed president, for example. And there is really enormous variation across time. There are really extreme cases. Bill Clinton, for example, was president for eight years. He met with Fed officials six times in eight years. Those were mostly official swearing-in ceremonies and things like that. If you look at Richard Nixon, he met 160 times with people from the Fed, and that was in five and a half years that he was president. And in the very intense period that I'm going to talk about more, actually, he met with Fed Chair Arthur Burns, on average, more than once a week. They met very frequently.

Beckworth: Every week, huh?

Drechsel: Yes. Now, this is interesting. This is just an interesting account of history. If you plot that time series of meetings, you see that it looks a little bit like different eras, different regimes. So, in the '60s and '70s, there were more meetings. In the '90s and 2000s, [there were] very few meetings. There was a little bit more action before the Treasury-Fed Accord in 1952, for example. Then, there's a small downward blip after that. So, it's an interesting account of history, but it, at this stage, has nothing to do with political pressure. They could be talking about baseball, or they could be talking about what's going on in the economy.

Drechsel: And indeed, that's actually an identification problem. I just told you, they met a lot in the '60s and '70s. It could be that they just met because inflation was high. Then, the president asked the Fed Chair, “Hey, can you tell me what's going on?” But what I want to investigate is, what happens when the president pressures the Fed to change policy? So, the data alone doesn't allow me to do that, while it is pretty interesting data to begin with. By the way, you can also see— I also collected if they had dinner, or if they had breakfast, if they met in a social context, or just in the Oval Office. It's fun to look at that.

Beckworth: I looked at the paper closely, and you had some really fascinating tables. This one figure that I think you're alluding to [is] the time series chart, it spikes in the '70s. So, you can see that it's very different than the other periods.

Drechsel: That's right, yes. So, when it comes to identification, I think about it the following way. This time series of meetings, in many instances, probably doesn't reflect political pressure, but in some instances, it can reflect political pressure. What I do is the following. I take that time series of personal interactions between presidents and Fed officials and stick it in a VAR with traditional macroeconomic data— GDP, inflation, some fiscal variables, and so on. Then, I apply something that is called narrative sign restrictions. That is a pretty new technique developed in a paper by Antolin-Diaz and Rubio-Ramirez, and that is a technique in which you can combine a narrative account with direct restrictions on a VAR.

Drechsel: The idea is as follows. You say that the meeting variable, the personal interaction variable, is in the VAR with other macro data. So, everything can depend on everything, in principle, but in certain periods, I tell the VAR that this movement, a particular movement in this meeting variable, is exogenously happening because of a political pressure shock. So, I pick a particular period and tell the VAR that, in that period, this variable moves because of one structural shock. A bit more expressed, a bit more technically, it means that I'm putting a restriction on the historical variance decomposition of the VAR. Now, the periods that I pick, the periods that I exploit are periods in which we know from other accounts, from historians and so on, that there were presidents that actually exerted political pressure on the Fed.

Drechsel: Those presidents for which we know that, for which historians have found that to be the case, are Richard Nixon— I already mentioned him— and Lyndon B. Johnson. The Nixon episode is a little bit clearer than the Lyndon B. Johnson episode. So, in one version that I use for the VAR, I only impose that as the shifter in political pressure. In another version, I use both episodes. Then, you put that restriction on the VAR, and you put some additional sign restrictions, traditional sign restrictions, to define what that shock is that you're imposing on the VAR that happens in that period.

Drechsel: So, in the case of political pressure, the way that I think about it is that a political pressure shock is something that moves personal interactions up, moves interest rates down, and leads to an initial increase in inflation. That's how I define it. The idea is that, in some way, through personal meetings, the president entices the Fed to ease policy. Important to note is that there could be political pressure to tighten. That wouldn't be in my shock.

Drechsel: That's not how I define it. So, I provide those signs. That's how I define the shock. Then, I tell the VAR that that shock, defined by those signs, was the main driver of meetings during certain quarters in the Nixon administration and during certain quarters in the LBJ administration. That's the idea of the identification approach. Then, in the paper, I provide a lot of narrative evidence of those two episodes. So, I can tell you a little bit about each one of them.

Beckworth: Yes, please do.

Drechsel: Nixon appointed Arthur Burns as Fed Chair shortly after Nixon was president. Nixon and Burns went back a few years. They knew each other. Burns was a member of the Republican Party. In fact, when Nixon ran against Kennedy earlier, that election that he lost, Burns was already an advisor to him and coached him on that famous TV debate. So, they go way back. Nixon becomes president. He appoints Arthur Burns as Fed Chair. Then, 1972 comes around, and Nixon wants to be reelected. We know that he pulled a lot of strings to be reelected. He also then, in 1971, leading up to the election when the campaign started— he exerts enormous pressure on Arthur Burns.

Drechsel: He calls him up, has frequent meetings with him, and very explicitly tells him, "I want you to ease monetary conditions, because I want you to stimulate the economy, because I want to be reelected." He very clearly says that. We don't know that from the meeting data that I collected for this paper. I only saw how often they met. But nicely for researchers, Nixon actually recorded conversations in the White House.

Beckworth: The Nixon tapes.

Drechsel: The Nixon tapes. There's some very nice work that summarizes the conversations about the Fed that are found in the Nixon tapes, and you see that he very clearly says that. He has this very clear understanding or this very clear view that we should ease monetary conditions to make voters happy, because it will stimulate the economy in the short run. There's also additional evidence in the diary of Arthur Burns. He wrote a personal diary that was not publicly accessible for many years. I think it came out sometime in the 2000s. So, I read through that entire diary, and it's overwhelmingly clear that Nixon exerted that pressure. The diary, by the way, for listeners— A big recommendation.

Drechsel: You can buy that on Amazon. It's a fun read, The Arthur Burns Diary. So, that's very clear, that there was pressure. It's a little bit more contested whether Arthur Burns gave in to that pressure. Here, there are different views. My view, and also the premise of my strategy, is that he caved. He gave in to the pressure. And I provide some evidence, also, for that. For example, if you look at estimates of monetary policy shocks during that period— If you take the estimates of Romer and Romer, which we might speak about later on, they find enormous easing shocks prior to the Nixon election.

Drechsel: So, that means that, based on economic conditions, you cannot really explain the monetary easing that happened in that period. That's one piece of evidence. The other one is that if you look at FOMC voting behavior, then you see that, prior to the Nixon re-election, more FOMC members dissented from Burns' view by wanting tighter policy, and that flips after the re-election. So, after the re-election, the other FOMC members want easier policy than Burns, and the interpretation is that Burns realized his mistake and wanted to correct it by then being overly tight in the inflation surge that followed this period.

Beckworth: So, there's a lot of circumstantial evidence that really supports this claim that he did cave to President Nixon.

Drechsel: And I accept that it's not crystal clear and we don't really know. I think that there is circumstantial evidence, and I use that in my identification approach.

Beckworth: Very interesting. So, you take this approach. You call it the narrative approach, and, again, [it was] fascinating to read. I loved your summary statistics of the measure. You mentioned 800 times, so many on the weekend, so many social interactions, just really fascinating by itself, but then you add structure to the data. So, you identify the shock, which is always the big thing in macro. How do we know for sure that we're identifying an exogenous shock versus just being caused by the environment, the economy? In fact, we'll talk about this [with] your next paper. How important is it to identify shocks versus systematic monetary policy? But you have some findings, so walk us through your findings. What are the results of your research here?

Breaking Down the Research Results

Drechsel: If the political pressure shock hits, then what you get is an easing in policy. After an increase in meetings by 10 meetings, you get around a 100-basis-point easing in policy. I'll circle back to how we best interpret that from a quantitative standpoint, but let me first talk about the qualitative results. So, you get an increase in meetings. You get a reduction in interest rates. Then, you get a pretty slow but gradual and strong increase in inflation. In the course of the following 10 years, you see that gradual increase. And for that 10-meeting shock, you get an increase of about 5% in the price level.

Drechsel: I should say that it's a gradual increase in the price level. So, there's inflation, a very gradual and very persistent increase in the price level of around 5%. Interestingly, little to nothing happens with real activity variables and with fiscal variables. We talked about earlier that one motivation to pressure the Fed, for politicians, might be to accommodate a fiscal expansion, say. So, you would think that that's related. That's not what I find. And in the case of Nixon, you might think, well, he wanted real activity to be stimulated. That also didn't happen. That's interesting. So, you get an increase in inflation, and you don't get much else. How do we interpret that? Now, I also show, in the paper, that you get a strong increase in inflation expectations that happens with a small delay, but it is also pretty strong, similar to the inflation increase. So, my interpretation is that private agents, firms, and households, to some degree, understand that there's political pressure and they adjust their expectations and, therefore, their behavior. I show, in the paper, that it is the case that these meetings got reported on a lot. 

Drechsel: If you think about it, if those meetings were entirely secret, [and] if the president goes to the Fed chair in complete secrecy and compels the Fed chair to make an interest rate decision that is not warranted by economic data, [then] that would be a monetary policy shock. That would be an exogenous shift in policy. No one understands it. It's a surprise. That would be a monetary policy shock. So, I should see if it was in complete secrecy. I should see that the responses are similar to a monetary policy shock where a monetary policy easing stimulates activity. Here, it's different. It can be different to the degree that agents observe the meetings and draw conclusions from them.

Drechsel: And if you look systematically at newspapers, the media did report on that, in general, across presidencies, but also, in particular, in the case of Nixon. So, the Wall Street Journal had articles where they talk about the clash between Burns and Nixon and will Burns cave and what does Nixon want and so on and so forth. So, it was public knowledge. There's an adjustment in expectations around inflation, around perhaps future accommodation of other shocks, and that leads to inflation. Now, quantitatively, one nice thing about what I do is that I can provide quantitative estimates. I talked earlier about, say, the literature that does cross-country comparisons.

Drechsel: Of course, you can draw quantitative conclusions from that too, but it's more directional, the way I read that evidence. But I can tell you, if pressure goes up by one, inflation goes up by X. Now, the way I do that here is that I can give you the inflationary increase per increase in pressure-induced personal interactions. Now, what that really means is not so clear. So, the way that I give the quantitative estimates is that I normalize it in terms of the behavior of different presidents. The number I provide is, if there is a political pressure shock that is half as strong as the political pressure shock during the Nixon period, and it lasts for six months, [then] that increases the price level in the following 10 years by 8%.

Drechsel: That's the quantification, so, half as bad as Nixon. Most presidents weren't even half as bad as Nixon, but it's half as bad as Nixon, [and] for a period of six months, [there is a] 8% higher price level. That's an economically meaningful effect. But interestingly, I think that it's, in some sense, a conservative estimate as well, because if you just look at raw data— say, you read that paper and you say, “I don't believe all of that VAR stuff, but I'm convinced by these meetings that there was something going on. Let me look at the Nixon period, before and after.” After that episode, the price level went up by 100% in the following 10 years. Of course, there was other stuff going on. Later on, there were oil shocks [and] other things. It was the end of the Bretton Woods exchange rate arrangement. There are other things going on. If you just do a case study, [then] it's not clear how you take that out. In the VAR, I try to take that out by identifying this shock, and then I get this 8% estimate, which is big.

Beckworth: It's very big, yes. And that's what struck me, is just the size and that illustration you gave, that if you did half as much as Nixon over six months, the price level would be 8% higher. So, it's huge. Now, I want to go back to this point that you made about how these shocks are different than monetary policy shocks, even though they're being implemented by the same institutions. So, the Fed does a monetary policy shock. It also does a political pressure shock. Both cases are adjusting the interest rates, the target rates, and you do a nice exercise in the paper where I think you use Romer and Romer shocks side by side and you show that if it's a true monetary policy shock, [then] you actually get more traction in real GDP.

Beckworth: With the political shock, you get basically none, but inflation is more pronounced with the political shock than with the monetary policy shock. And I think that you nailed this inflation expectation story. In terms of a New Keynesian Phillips curve, it's that expected inflation term there is that— I guess, the parameter is getting big in front of it, and it's important. People are getting concerned. I guess I want to just park here for a minute. The more that they are aware of it, I think I heard you say, the more of an effect we should expect to see. Is that what I heard correctly from you?

Drechsel: The fact that they can be aware of it makes it different from a monetary policy shock. It's unclear whether I can map that into an elasticity of one with respect to the other. I would need to think more about it. Intuitively, the fact that they can observe it makes it different from a monetary policy shock.

Beckworth: I mean, intuitively, if you believe that the Fed's becoming more politicized and it's in your face, it's [on] the front page of the paper, you see it in the news, [then] you're going to begin to worry. I guess, what would be interesting to see— and I know your sample stops in 2016, but what happened during the Trump period, prior to the pandemic? I know it was a low inflation period, but maybe inflation was higher than it would otherwise have been, or inflation expectations were higher than otherwise would have been. I guess what would be interesting more recently, though, is that Trump was doing a lot of it via Twitter.

Beckworth: So, it wasn't a meeting per se, but man, it was very apparent. I'll give a story that I've given on here before. The one time I was invited to the Jackson Hole meeting, just one time, it was 2019, it was August, we're sitting there, and Powell's giving this talk, and honestly, most people were listening to Powell, but they're on their phones, because they knew Trump was just— and as soon as Trump saw the speech, he was disappointed. He was mad, and he compared Powell to the leader of China. Who's the worst enemy, Xi Jinping or Jay Powell? And that rhetoric, that pressure, that was very in your face.

Beckworth: So, I wonder, [if] that commentary would be destabilizing. I know there's been recent, again, discussion about Trump. It was reported in the Wall Street Journal, and then more recently, he even, I believe, said this in an interview, that he thinks he should have some influence over the FOMC. In fact, I think he talked about [how] he has a gut feeling. I jokingly call it the nominal gut feeling target. But in any event, because things are so apparent [and] open today, these types of pressures, do you think it would have a bearing on that expected inflation term?

The Impact of Political Pressure on Inflation

Drechsel: Let me give a general answer to that, and then an answer more with respect to Trump in particular. Political pressure on central banks can happen through many channels.

Beckworth: Okay.

Drechsel: It can be through meetings between the president and the Fed Chair, meetings of other people. It can be through public pressure. There might even be other channels. It's really hard to pin it down. What I do is, I look at these meetings, and that's one particular channel that we can capture with the data consistently through time, and it's important for that channel that there is some public observability. A separate channel is direct public pressure. Now, that [also] happened a little bit during the Nixon period, actually. But for me, it is that particular channel and it's public observability. Maybe I am missing some other channels in what I do in this particular paper.

Drechsel: Now, the direct public pressure— we saw that, as you rightly pointed out, during Trump's presidency. I don't have Trump in my sample. I also don't have Biden in my sample. I unfortunately can't study those periods. But there is some research that I can point you to and point listeners to by Francesco Bianchi. It's a JME paper where they do a high-frequency approach looking at Trump's tweets. What they find is that the tweets did, to some degree, shift the market's expectations of future policy. So, Trump says, “The Fed is a disaster. They better cut rates.” Then, markets, to some degree, price in a future rate cut. I thought that that's a very interesting finding. I didn't personally study Trump. Maybe I will in the future. So, I also don't know whether Trump, in addition, met a lot with Powell. I can't really say that. I don't have those schedules yet for Trump, but the public pressure was there.

Beckworth: It was intense. At least it appeared intense, [and] by all measures, probably was. And I do remember reading a story in Nick Timiraos' book where Trump called in Powell and, I believe, Vice Chair Rich Clarida came with him. He just let him have it. He ripped into them and stuff. Of course, at the time, he was starting his tariffs, his trade policies, and that was causing problems, and eventually, they were tightening, and they did turn around in 2019. They began to cut rates. The economy looked like it was slowing down, but, nonetheless, very interesting. We'll provide a link to that research you just mentioned. I do recall that high-frequency data. So, there's some— Again, in the grand scheme of things, here's the question. Was it that big? Because we still had low inflation. Inflation expectations were low. But the argument would be that it was higher than it otherwise would have been had Trump not gone on Twitter.

Drechsel: Perhaps I should also— Maybe one more thing to add here. In my paper, the way I define that political pressure shock is a shock where, due to the pressure, the Fed does give in. It's not clear to what degree Powell reacted to the pressure. My view is that, actually— And this is not coming from a position where I systematically study that. That's more my observation. My sense is that he didn't react. I think that Powell stayed firm, and I hope he would in the future. So, that's also a little bit, then, different. Maybe the market prices in some probability that he might, in some state of the world, give in. Then, maybe if, in addition, personal meetings happen, and he gets called into the White House every week and Trump shouts at him, that might change. But my sense is that Powell has managed to steer clear of politics as best as he can.

Beckworth: Well, I think that he definitely did a good job with Trump. So, I think it was, in some ways, very fortuitous for the US that he was the Chair at the time. It's been well covered in the press that he's very politically smart, and you needed to be, in that environment. But your point that you're making, I think, is a really good one, and that is, he held his own, and that may be the reason that the magnitudes might not be that high, in terms of whatever the effect was in these papers. The overall numbers were still relatively low because of the credibility.

Beckworth: I think that also ties into what we see today. We have effectively recovered from the inflation surge of '21, '22. Inflation is pretty close to target. Inflation expectations are definitely back down to pre-pandemic levels. That speaks to credibility, right? It speaks to the trust in the Fed and the institution. So, maybe him fighting against Trump is some of the buildup of credibility that was used to get onto the other side of this pandemic inflation surge. But it will be interesting to see what happens if Trump does get elected and he does pursue a stronger voice on the FOMC as has been reported and as he, himself, said in an interview. So, this will be more data for you.

Drechsel: Indeed, yes.

Beckworth: Maybe not the best world for the rest of us, but you'll have lots of research to do.  So, let's move to your next paper. It's titled, *Identifying Monetary Policy Shocks, A Natural Language Approach.* So, another interesting narrative approach, and you build on the famous Romer and Romer 2004 paper. And this exercise is all about properly identifying a monetary policy shock or getting a better grasp of it, handle of it, because as we talked about earlier, it's hard to identify things in macro. I guess, just to lay the groundwork here, why do we worry about identifying shocks versus systematic policy? Can't they both be important to the long run path of the price level?

Identifying Monetary Policy Shocks: Background, Methodology, and Results

Drechsel: Before I answer that question, I quickly want to say that this is work with my Maryland colleague and friend, Borağan Aruoba, who I really enjoy working with, and this is about identifying monetary policy shocks. Why do we need monetary policy shocks? Again, there's an endogeneity problem here. We want to understand what happens to some macro variable of interest when there is a change in the federal funds rate. Now, the federal funds rate doesn't change in isolation. It is set by the Fed who considers that macro variable of interest in setting the rate. So, there's a reverse causality. Interest rates change for a reason. They only change when macro variables move around. So, what we ideally want is some exogenous shift in policy, something that has nothing to do with the economy. We can talk more about what might trigger that shift, but that's the idea. You want a shift in policy that is not a response to economic conditions.

Beckworth: So, you're trying to truly isolate what the Fed is doing, by itself, that's adding to the changes in the economy. You don't want to get the Fed simply responding to existing changes. So, inflation goes up, the Fed tightens. That's really not the Fed acting independent. We want to truly measure Fed exogeneity, [which] is the technical term. Here's my question, though, is that you can imagine a world where the Fed puts its target rate, let's say, above the neutral rate. So, it's tight. Maybe the first time they did it, it was a surprise, but let's say that it persistently does that. People begin to expect it. That systematic policy might weigh down the economy. Now, maybe you could invoke, in the long run, that the Fed is neutral, so it doesn't really matter, let's go back to shocks. But is there not a role to also think about systematic monetary policy?

Drechsel: Yes, of course. Systematic policy is what we ultimately care about, but to really understand how systematic policy is set and what it means, you need a structural model of sorts to get at that. If we want to use data to look at the effects, [then] we have got to have some exogenous variation to exploit.

Beckworth: So, we need to use these shocks to draw inferences about systematic policies, is your point. Okay, fair enough. Well, let's move to your paper and how you actually calculate the shocks.

Drechsel: Yes, so we begin with the Romer and Romer 2004 paper. It's really a classic reference, a very interesting paper. So, what the Romers suggested is to do the following. You take the target federal funds rate and regress that on variables that summarize the information set of the FOMC. So, you want to regress out what the Fed considers and knows and looks at, and then retrieve a residual, and that residual is that movement, that portion of the movement in the target federal funds rate that is not based on the information set of the Fed. That's an exogenous shift. That's also what a monetary policy shock would look like in a DSGE model. It would be the epsilon T to the Taylor rule, for example.

Drechsel: Now, it's not immediately clear how you can measure the Fed's information set, and what the Romers suggested is to use forecasts from the Fed's Green Book. The Green Book is a document that staff economists at the Fed compile for FOMC meetings where they provide an in-depth analysis of economic conditions, the economic outlook, inflation, housing, financial markets, the labor market, all of those things. And they, as part of that Green Book, provide a table with forecasts of economic variables. The Green Book is closely looked at by FOMC members, and they use it to form their views and then make decisions.

Drechsel: So, the Romers’ idea is that we take data from that Green Book— and in particular, the forecasts of unemployment, inflation, and GDP— and then regress that out from the target federal funds rate, and then obtain monetary policy shocks as the residual. It's a classic reference. Our starting point is the following. We said that the Green Book not only contains forecasts, but the Green Book contains a lot of information that is captured by text, and it's, in fact, carefully crafted by the staff economists. They really think about how they [write] the words. They say, “The housing market— there's some subdued activity,” and then, in the next meeting, they say, “There's elevated activity.” They craft it very carefully.

Drechsel: So, our idea was, can you take the Green Books and somehow throw the entire Green Book on the right-hand side of a regression? Can you take a PDF and include it in a Romer-Romer style regression? The answer is, yes, you can do that using natural language processing and machine learning techniques. So, you need a way to translate text into data. That's where the natural language processing comes in. And then, you need a way to put lots of data in a regression, because out of the text, you can potentially get a lot of information. It's very multidimensional.

Drechsel: So, you need some sort of machine learning technique, some way to discipline how that data features in the regression. So, those are the two elements— natural language processing and machine learning. I do want to say, before continuing about what we find and how that plays out and how we apply those techniques— I do want to talk about one important subtlety here, which is the following. It is not necessarily clear that you want to throw as much information as possible into a Romer-Romer regression. It sounds compelling, you might say. They include these forecasts, but the FOMC members, they look at a lot of things. So, let's just put in a lot of things and get a smaller and smaller residual, and that is going to be cleaner and cleaner.

Drechsel: That is, at first glance, a compelling thought, but that's not exactly true, and that has been pointed out in a comment that John Cochrane wrote on the original Romer-Romer paper. He discussed the paper at the NBER Summer Institute in 2004, I believe, and then wrote his discussion up as a comment. Cochrane said the following. You actually don't need all sorts of variables, because the forecasts that are included in the Green Book are already conditional on a lot of information. So, under certain assumptions, like rationality and things like that, you actually have those forecasts as a sufficient statistic of everything that the Fed knows. So, you only need to include the forecasts. That's what Cochrane said.

Drechsel: So, what we explain in the paper is that things are a little bit more subtle, and there are actually reasons why you'd have to include information to the text, and that's the following. The Cochrane critique or the Cochrane argument is valid if the Green Book forecasts correspond to the true conditional mean expectation of the FOMC of a variable of interest, where the conditioning set is the full information set. It's the conditional mean expectation. Now, Green Book forecasts are actually modal forecasts. They are not mean forecasts. If you look at the Green Book, and if you look at how the staff communicates with the FOMC, they say, “In our modal scenario, unemployment will be 4.5%, but for the following reasons, we think that there is now more risk on the upside.” Then, they explain, in words, how the distribution around the mode has changed. So, if you want to capture the correct conditional mean, you can use a mode forecast and verbal information about the tails around that mode to get at the true mean. This sounds a little bit technical—

Beckworth: It's important.

Drechsel: -but I did want to point that out, because the paper is not about throwing more information into a Romer and Romer regression. It's about the right information, and we have tests and explanations for why that text is the right information.

Beckworth: Very nice. So, walk us through the results. What did you get out of this?

Drechsel: So, the way that we do this, then, is that we take the text, and we construct sentiment indicators from the text. We pick the most frequently appearing economic concepts in the text, say unemployment, GDP, the things that they forecast, and additional things, and we look at how they talk about those things. We look at word windows around when they mention these economic concepts and look at words that, in the English language, are associated with positive or negative feeling. Then, we construct indicators around with what sentiment do the staff economists talk around, say, the housing market, in that meeting.

Drechsel: So then, we get, for many of those economic concepts, time series. We put those in a Romer-Romer regression, and we apply ridge techniques to discipline that regression, because we now have the numerical forecasts that the Romers have used. And in addition, we have a large batch of these sentiment indicators. We use a ridge regression, and then we get a residual that explains a lot less in the variation of the target federal funds rate than the Romer-Romer original residual.

Drechsel: So, our regression has an implied fit of 94%. So, around 6% of the changes in the target federal funds rate is attributed to shocks and 94% to systematic policy. In the Romer-Romer original paper— If I recall the original Romer and romer paper correctly, the fit was less than half. So, a lot of the share is attributed to shocks. We also replicate the Romer-Romer results with updated data. It's around half. Although, again, with the caveat that you might not need to have a very high fit because of this Cochrane comment. But the shocks that we get, with this information that we argue is relevant information, is small. Now, in some way, that is desirable, because that's also how we think about policy, as being mostly systematic.

Beckworth: The Fed's been very systematic.

Drechsel: Yes, that's right. Then, we take those shocks, and we construct impulse response functions to those shocks. What we find is that the responses of output, inflation, credit spreads, stock prices and so on, are much more in line with what theory predicts, which we argue is a sign that these are cleaner shocks. If you replicate the Romer and Romer procedure with updated data, you find some responses that look, in light of theoretical models, slightly strange. And in the paper, we actually link that directly to this mode-mean argument and to different tails being captured by the sentiment indicators.

Beckworth: So, you're basically cleaning out the Romer and Romer shock. You're using better information, or more complete information, in a sophisticated way. And you look at the results, you compare the Romer and Romer impulse response functions to your shock impulse response functions, and all of the results are cleaner, more consistent with standard macro theory, which is great. Now, are these shocks available for others to use, like you've made this nice database available? People can go run and do their own experiments with them?

Drechsel: Yes, absolutely. We make the shocks available on our academic websites. We also make the sentiment indicators available. I should also say, for the paper we talked about earlier— the personal interaction variable for presidents and Fed officials— I also make that available on my website.

Beckworth: Okay, and one last thing about this paper— You updated it with Beige Books because the challenge in this measure, which is great— It's also a challenge for Romer and Romer— is that it relies on data from the Fed that has a five-year lag. When the meeting takes place, you have to wait five years before you get the Green Books. So, you found a way to update that's pretty close to what you'd get with the Greenbook, right?

Drechsel: Yes, that's right. The interesting thing is that, as you point out, the Green Book becomes available to the public and to researchers with a delay, so we can't construct shocks up to the latest meeting. But the Beige Book does come out, in real time, before every FOMC meeting. The Beige Book is composed by the regional reserve banks, by economists there, and they describe economic conditions in each Federal Reserve District and some information about the aggregate US economy.

Drechsel: Now, the Beige Book only contains words, so you actually cannot apply the original Romer and Romer approach to the Beige Book. You can use this natural language processing approach. So, as a sort of imperfect version of what we do, you can go up to the latest meetings and construct shocks. Although, the Beige Book might be missing something that is in the Green Book. So, I should stress that this is not our favorite shock version, but it's something that you can do.

Beckworth: You could use it, though. Two final questions— So, first, could you use the Survey of Professional Forecasters, if that's close enough to the Green Book? Secondly, then, looking at that chart at the end, where you use the Beige Book, you bring it to the present or pretty close to the present— It does look like there was a big negative shock, in terms of monetary policy, during the inflation surge. So, rates were held surprisingly low, unexpectedly low. Is that a fair interpretation?

Drechsel: Yes. On your first question, you can bring in the SPF. You can, in principle, bring in any other publicly available data as additional information to control for. We don't do that. One reason is that we want to point out that it's really this data that is required, based on the arguments that I made earlier. But in principle, you can bring in more data. Then, [the problem is], where do you stop? You can bring in the SPF, and then you can bring in all sorts of other things. On the other point, yes, we applied a method with the Beige Book to more recent meetings. We have to take those results with a grain of salt, because it's not the best version that we have. But you do find that there is some flavor that the Fed was behind the curve. You could spin it in this way.

Beckworth: That's what I saw.

Drechsel: That's not how we write it in the paper, because it comes with big caveats, but you see a little bit of some mild easing shocks before they actually started.

Beckworth: That's fair. You're a researcher. You've got to be serious and measured. I'm a podcaster. I can say, “Wow, look at that. That was a big negative shock, unexpected. That was easy, and the Fed fell behind the curve.” So, I exaggerate, and you rien me in, put me in place. So, I appreciate that. Well, with that, our time is up. Our guest today has been Thomas Drechsel. Thomas, thank you for coming on the program.

Drechsel: Thanks a lot. It was a lot of fun.