 Next paper is the economics of the Fed put, and Annette Vissing Jorgensen presents it. Thanks very much. Hi. Is my microphone on? Can you guys see me? Pretty good? All right. The guys will manage it. I thought if I ran around a bit, then I could keep everyone awake. So this is a, I don't work well and it's just like who said Duke. This is on the Fed put by which I mean strong monetary policy accommodation following stock market declines. So we are, sorry, it's eating half of my header there. It's all right. We're interested in figuring out how much monetary policy reacts to the stock market. We're going to focus on the Federal Reserve. And we're interested in both how much the Fed accommodates stock market slums, but also how much it might lean against the win in stock market booms. And of course, how much policy should react to the stock market. These are longstanding questions given the size of the stock market. Of course, this is potentially a big deal in terms of the stock market drawing the economy, driving policy. There are some identification challenges, which is that not only does the Fed potentially react to the stock market. It may also affect the stock market. So you can think of sort of a demand supply system where if you can get variation in demand, you can identify supply and the other way around. So on top of that, then you have omitted variables concerns that both the stock market and interest rates may be reacting to underlying macroeconomic news. So you have sort of the worst of both worlds. You have omitted variables as well as reverse causality. So the way that people have addressed this in the literature is there's a sort of a very clever identification by Hilo-skidisticity approach that Vigabond and Sack have promoted, which is essentially, if you think again about a demand and supply diagram, if you can get one of the curves to move around a lot, then you can identify the slope of the other one. So there, essentially, I say whenever the stock market is very volatile, then that changes the covariance between the stock market and interest rates in a way that will allow you to identify how much the central bank reacts to the stock market. So this works well as the policy shocks are homoskedastic so that it's only the stock market volatility that changes over time and parameters are stable across regimes. Another approach promoted by Bernanke and Girdler has been to look into whether the stock market is significant in a tail rule above and beyond controls for central bank expectations for growth and employment and inflation. There you have the advantage that you could to some extent overcome the omitted variable problem by including the expectations, the green book variables in the sense that those presumably are sufficient statistics for how the macroeconomy is doing and so nothing sort of should be significant above and beyond those controls. Now there are some concerns with these approaches. So Montlain has a paper where he tries to figure out if central banks have specifically the Fed has an asymmetric response to the stock market using this identification to hit a scarcity and finds that the results are very sensitive to sort of specific identifying assumptions. In terms of the stock returns in the tail rules, that approach is sort of helpful for figuring out whether the central bank might be overreacting to the stock market. It's not as helpful for figuring out whether it's reacting to the stock market in the sense that you could have that the stock market is sort of a really strong driver of the Fed's growth forecast which then in turn drives policy. You can't see that from just running regression where you add the stock market into a tail rule. Furthermore, some questions that these approaches are not designed to answer is how important is the stock market relative to other potential drivers of policy? And also what is the economics behind any impact of the stock market on policy? What is the Fed thinking as it potentially is reacting to the stock market? Is it about investment consumption? What are the underlying economics? So we're going to focus on these four research questions based on what I laid out with some of the shortcomings of existing approaches. First, I'm going to just do statistics. We're going to run just predictive regression and see does it look like the Fed accommodates following low stock market returns or following high numbers for initial jobless claims? Just to get a sense of, okay, does it seem like at least statistically that this is the first sort of thing to think about that the Fed might be reacting to the stock market? Then we're going to argue causality and mechanism by textual analysis. We're going to, with the help of lots of Python code, read a bunch of FOMC minutes and transcripts. The basic idea will be that if the Fed is reacting causally to the stock market, it better be the case that they talk a lot about the stock market after it's gone down. You could think of this as sort of a necessary condition for causality. If they never talked about the stock market, then presumably it could not be causing their thinking. In terms of mechanism, we're going to look for what are they talking about in the same context that they talk about the stock market? So we'll look at sort of paragraphs where they talk about the stock market, what sort of stuff are they talking about in this paragraph sort of disproportionately relative to what they generally are talking about. We'll look at, they're talking more about consumption, investment, and so forth in the context of their stock market discussions. And then we'll have various approaches to discuss whether the Fed, if by then I've convinced you that they're reacting to the stock market, whether they're overreacting or whether they're reacting appropriately. So you guys are more of a monetary policy crowd. For the asset prizes, let me just show you a couple of slides to convince you that really understanding whether and why the Fed reacts to the stock market is really also a first order importance for asset pricing. So this will also give me a chance to show you a slide or two from my earlier work, which will show you why we got into thinking about this Fed put. So Anna and I have an earlier paper at Air Morris where we argue that not only as we'll focus on in the current context, this will seem like the Fed is reacting to the stock market, but also it's driving the stock market. So it turns out that, let me show you the picture here, that if you do an event study relative to the day of the FOMC meeting and you keep track of the five-day return on the stock market, then there's this very weird pattern that in even weeks in what we call FOMC cycle time, the stock market has done very well. So we argue in that earlier work that this is actually the Fed driving the stock market. We have various approaches to document that that's a causal impact of the Fed on the stock market. It doesn't line up with any other calendars to the extent that there's inter-meeting target changes. They tend to be in these even weeks. Fed fund futures on average go down in these even weeks. And then also we link the even weeks with high returns to some actual decision making or discussions inside the Federal Service, specifically the Board of Governors Board meetings. Now, what's also interesting is that in that earlier work, we find what you might call a Fed put in stock returns. So if you sort on the axis, look at the left picture, which is for even week days. If you sort days in even weeks based on how well has the stock market done over the past five days? So has it been a good week or a bad week? Let's just do five quintiles based on that. You can see really high one day returns in these even weeks. That's the point up to the left top is when last week was a really bad week. And today is one of those even weeks where we argue that the Fed is doing its thinking. And so that suggests that that shows that strong monetary policy accommodation follows poor stock market performance. But it doesn't tell us whether the Fed is reacting to the stock market directly or whether the stock market perhaps just goes down when growth news or employment news comes out negatively. So it's helpful for understanding document that the Fed is driving the stock market, but it's not helpful for understanding whether the Fed is directly reacting to the stock market. So that's what we'll dig into here. That's the title, the economics of the Fed put. Now, in terms of the main findings, I'm going to try to convince you that negative stock returns are a very powerful particular statistically of changes in the Fed funds target. It's as powerful as most macroeconomic indicators out there. And then I'm going to, as I said, do this causality and mechanism, my textual analysis. What will come out of this is that the stock market seems to cause Fed policy and that the key mechanism behind this is that the Fed has a strong belief in a consumption wealth effect. Okay, so when the stock market goes down, households are losing wealth and that causality drives their consumption. In turn, it drives the Fed's growth forecast and through that in a standard table framework, the Fed's policy made. And there's not going to be a lot of evidence, actually, that the Fed is overreacting to the stock market. We'll have some details there. Okay, so let's take a deep breath and let's just take a look at what does this Fed put look like in the Fed funds target? Okay, so before I showed you a Fed put in stock returns, here's the sort of standard Fed put in the Fed funds rate target. So to the left, I'm using data since 1994 up to the Seolow bound. To the right, data from before 1994. The Fed put shows up in the, from some time in the mid-90s. I'm using the 94 cut-off just because that's what we had done in earlier work. But again, I'm sorting on the x-axis on how well has the stock market done in the recent past here. I'm sorting on how well it's done since the last FOMC meeting. And then on the y-axis, I'm keeping track of how much does the target change, either over one cycle, one FOMC cycle, a period of two cycles or so forth. The most useful dot is the orange dot in the bottom left. That's a change over a one-year period. The interpretation is that if the stock market in the intermediate period has been in the lowest quintile, that's about a loss of 7%. Then over the next year, the Fed statistically, in a predictive way, has lowered the target by more than 100 basis points. That's there in the post-94 period, but not before. That's split in terms of the timing will show up in our textual analysis. I'm going to show you that. They start talking a lot about the stock market in the middle of the 1990s. Now, if you run this as a regression, just in terms of r squared, again, so far it's a statistics. Let's run the change in the Fed funds rate on two lakhs. That's the first column. You get an r squared of 35%. Then in column three, let's add in the stock market return over the intermediate period as well as one lakh of that. I'm putting a minus here to indicate that I'm interested in, I'll call this my Fed put variable. It just means that in the regression, I'm only using a variation in the right-hand side variable over the negative range, arguing that they're particularly sensitive to negative news. You can see now the r squared goes up here to 51%. So a substantial increase in r squared. We'll look at that incremental r squared next for macroeconomic variables and we'll be able to compare how much incremental r squared do we get out of the stock market versus how much we got out of various macro indicators such as ISM or non-farm payroll and so forth. The statistical relation documented here is something that the Fed has come under criticism for. Former Governor Kevin Warsh, I put aside here, says they look to me as a price dependent more than they look economically dependent. When the stock market falls like it did in the beginning of this year, they say, oh, we better not do anything. Now this suggests that this reaction, this statistical predictability of policy is a problem. We'll have to dig into whether that is a problem. We're going to be, I think when all is said and done, much more on the Fed side of this given a strong consumption wealth effect, then it is actually rational to react to the stock market. But let's take a look at how much incremental predictive power we get out of the competitors to the stock market, namely all these macroeconomic indicators. So we took data from, for all the main macro releases in Bloomberg. There's about 38 of those. And we ran the same two regressions that I just talked you through. First, just the change in the Fed funds target on two lags. And then second, the same thing, add in some, in this case, macro data. There's some missing data. It's just disregard those. Then in the column that says incremental are squared, you can see how much extra predictive power we get out of the stock market. That's the number that I just talked you through. The sample here, the number is slightly different from the two numbers I showed you before because the sample only starts in 96 as opposed to 94. The main point here is if you take, look at the other data, which is all macro releases, then you get smaller incremental are squared. The point here is not that the stock market is slightly better than the macro release, but that the stock market incremental are squared seems to be pretty large relative to most macroeconomic news. You can even take, in the spirit of combining macro data, you could put in the, here at the bottom, the Chicago Fed National Activity Index. That's the first principle component of about a hundred, as far as I remember, macro series. Even then, you only get to about 12.9% incremental are squared. It seems like this from a statistical perspective, the stock market is a good predictor of Fed policy, so hopefully that will make you more interesting in what I'm going to say next. There's two ways that this could work out. Either the stock market causes policy by which I mean that it drives the economy, for example, through the consumption wealth effect. Or because it's just a great sufficient statistic for predicting the way things are going. Okay, both of those I'm going to classify in the causal category where rational Fed should pay attention to the stock market. I want to first distinguish that causal bullet point from the coincidental one, which is that the Fed really doesn't care at all about the stock market. It just happens that the stock market does poorly when, say, unemployment goes up and the Fed cares about unemployment. And I already talk through what is the main, there's really just one idea in the paper which is, you know, you have a decision maker that writes down how it thinks. Let's look at what they're saying. And first, we're going to do this to figure out if they talk a lot about the stock market after it's gone down. We're going to do that first by reading it manually and then by Python code. Then we're going to do text analysis one more time, reading manually what's the mechanism and also coding with Python code what they're talking about. So here's the text that you get to work with. So the FOMC meetings are highly scripted. First, the staff talks. Then the policy makers, which are called the participants in Fed terminology, they talk and then there is the decision. Now, we're going to study the meeting, the minutes. There's a text analysis choice to be done here. One is you could say you could use the statement, but that's kind of too short in order for there to be much discussion of how they reach their decision so it's not the best possible text. You could say, well, what about the transcripts? Well, there you get almost too much text. You get 200 to 300 pages per meeting. It's not structured in the sense that one guy might ramble along about the stock market for 20 pages. So the minutes we think is probably the best text you get seven to 10 pages per meeting and also the minutes are available up to the end of the, you know, up to currently. So you don't have this lack that you get to the transcripts where there's a five-year delay. Nonetheless, I'm going to also look at the transcripts for robustness as we get into the numbers here. So we have all the text in the minutes and then we look for the word stock market and it turns out that there are many different ways that they say stock market. But just so you can get a sense of how many counts we have here. So in the period since 1994 and the second line from the bottom, we have eight minutes per year. So we have 184 minutes in total and we get seven to 10 pages long. We get about 1,000 mentions of the stock market. Now what we want to, here's a time series, notice I don't know if you can see the dot here, it actually does work suddenly now. Here's the mid-90s. There's not, with the exception of the 87 stock market crash, there's not a lot of talk about the stock market before the mid-90s. And then there's sort of consistent attention being paid to the stock market after that. Notice that that's actually pretty consistent with that fat put only showing up in the P94 period in my graph before and not in the post-94 period and not in the P94 period. Now we're going to, that's the transcripts you get basically the same. Okay, so all these 1,000 mentions of the stock market, half of them are kind of uninteresting. At the FMC meeting, staff summarized the financial situation and it's someone's job to say how well the stock market is doing so that's not as interesting. More interestingly, there's 272 cases where the participants, and again, those are the decision makers, the chair, the vice, chair, the governors, the regional fed presidents where they talk about the stock market. So we're going to go ahead and we're going to read these about 1,000 paragraphs to figure out whether they're talking about the market going up or down and then here's the time series of them talking about the market going down. Now the little red triangles that indicates that the intermediate stock return was in the lowest quintile and so what you want to see is a lot of red triangles at the top of the spikes. That means that they are expressing more concern about the stock market after it has actually gone down and you can see that they're sort of the well-known cases, the Lehman and so forth. The positive mentions are not as interesting. The relation between the intermediate stock return and then the counts we did here, again, a bucketing by the variable on the x-axis, you can see the relation here a little bit better than you could in the time series picture when the stock market does poorly. They talk more about the stock market having done poorly and conversely on the upside. Most interestingly, you can actually predict the changes in the target based on how much concern they express about the stock market at the meeting. So let's focus on the first column is all stock market mentions. The more interesting one is probably column three where it's the stock market mentions by the decision makers. The pink, let me just give you here the terminology here. When it says number of stocks and minus that means the number of stock market mentions that are concerned about the market going down. In case of the minus refers to them being concerned about going down whereas plus refers to them talking about the market going up. Now if you look at column three, you can see that the stars are all on the minus variables. So that says when they talk about the stock market going down, then they follow through and lower the target. When they talk about the stock market going up, that's all talk. It doesn't have any impact on the target. So in that sense, you can see here what you could see, you could call the fat put in the tax zone analysis which lines up quite well with what we saw first in the stock market and in the target and now in the text. I put the economic significance down here. One sigma increase in negative stock market mentions means they lower the target by about 34 basis points cumulatively over the next few meetings. Now we want to be, we want to check that this is robust using the transcript. So we teach Python to think like an economist in the sense that we want to tell the code, okay, look for the word stock market or the phrase stock market and then look for whether it says going up or going down, you know, near the phrase stock market. That's a little harder than it sounds. There's a lot of different ways you can say stock market. There's many different ways you can say down and many different ways you can say up. But once you have trained the code on the minutes where you have already vetted manually and you know what the paragraph is saying, you can then run the same stuff in the transcript and you get similar relations. All right. So that was the conclusion on question two on causality. We believe that the fact is reacting costly to the stock market because they talk more about, they do talk a lot about the stock market at the meetings, especially when the market has gone down and you can predict changes in the target with how much they talk about the market going down. So moving on to our third question about the mechanism. We first take those thousand paragraphs. Now we read them one more time to figure out what they're talking about other than the market going up or down. So here's a list of ways in which they talk about the stock market. So the first kind is not that interesting. This is an example of a purely descriptive mention where they say brought you as equity price indices were highly correlated with foreign indices over the intermediate period and posted net declines. Fine. That was some staffer's job to describe how things are going. The more interesting ones are the ones that give some specific economic context. So here's one saying consumer spending had held up recently well in recent months despite a variety of adverse developments including the negative wealth effect of stock market declines. An example of an investment mention would be here many businesses also were inhibited in their investment activities by less accommodated financial conditions associated with weaker equity markets and tighter credit terms. Investment was being cancelled. Then somewhere they talk about demand without specifying whether it's consumption or investment. Somewhere they talk about the stock market as being one of several financial conditions. Generally when they talk about financial conditions they talk about interest rates both long and short. They talk about the dollar. They talk about credit conditions and they talk about the stock market. I'll show you textual analysis count for the other financial conditions. Later on the reason we focus on the stock market here is because that's where there's something new going on from the mid 90s. They start focusing on the stock market whereas the other financial conditions that I mentioned they have been focusing on sort of throughout the period since the early 80s where we have data for. The one that I had expected to see a lot of was that the stock market is just a good predict of the economy. And we tried really, really hard to look to find words saying that. I found one here saying it brought the clients in stock prices straight over the intermediate period was seen as mostly reflecting the incoming data pointing to a weaker outlook for growth. I was expecting to find hundreds of these and we can barely find any. Let me save you the weeks of reading paragraphs. Look at the column with the red numbers for our reading of the 272 paragraphs where some of the decision makers, the participants talk about the stock market. What are they talking about? Consumption. So out of 272, 150 are cases where they talk about stock market in the context of the consumption wealth effect. 29 is about an investment, 40 about financial conditions, but almost none. 12 are the cases where the stock market is being mentioned simply because it's a good predictor of the economy. So they think the stock market is driving the economy as opposed to just predicting it and that's due to a belief in a solid consumption wealth effect. There's a couple of quotes just so you don't think I made this up. Bill Doddley, the former head of the president of New York Fed, saying that a rise in equity prices can boost household wealth, which is one factor that underpins consumer spending. Richard Fisher, former Dallas Fed president, saying that they front loaded at the Fed an enormous rally in order to accomplish a wealth effect. So you might be a little skeptical that once I had read 100 of those paragraphs, I was just looking for the word consumption. So we did some Python code here to try to more systematically look at what else is being mentioned in the context of consumption. And let me just show you here what happens for the participants. So we look for all economic concepts. We make our own dictionary of economic concepts and we see what's disproportionately being mentioned in the context of a stock market mention. And it's resorted then by the economic concept that disproportionately are being mentioned in that context and the stuff that shows up is really all consumption related. So this was sort of the Python code check on the reading of the paragraphs. All right. So finally, the fourth question remember was whether the Fed is overdoing it and we're going to have three approaches. We're going to compare the updating of the Fed's screen book or teal book forecast for growth and inflation to see if the Fed updates their expectations more aggressively than the private sector. We're also going to look at how much actual predictive power there is in the stock market for growth and inflation. The second approach will be the tail rule approach. And the third one will be to look at whether consumers worry about the stock market when the Fed does. And by all accounts the Fed is going to look actually pretty rational. So here's the updating of the green book forecast in response to the stock market. So this is a constant calendar quarter. So think about Q zero is the current quarter. You had some expectation about growth in that quarter before this FMC meeting. So at the last FMC meeting, then since the last meeting we get some incoming data in the stock market, we update our thinking about the same quarter. We update our thinking about Q zero, Q one, Q two and Q three. I'm going to add them up. So this is an updating in growth over the next year. There's some gradual reaction to the stock market. If you sum up those coefficients you get basically 10. So focus here on the pink lines. The interpretation of this is that if the market drops by 10 percent, then they lower the growth expectation by about 1 percentage point. For unemployment they lower in column three the unemployment expectation one year out by about a half percentage point. So just remember the numbers 10 and 5. The update growth expectation 10 on employment 5. Here inflation there's basically nothing going on. Here's the private sector updating. Look at the red line 9 and 5. This is from the SPF. It's pretty similar. If you do the blue chip you get 6 and 5 you get 6 and 4 a little lower but not much. Notice also that if you let's go back to the fact if you look at the numbers from in columns 2 and 4 from the pre-94 period you don't see much updating. So you get a consistent pattern that before 1994 they were just not paying attention to the stock market. Not in their minutes, not in people talking, not in the target, not in the growth updating. And after 1994 the whole thing changes. Notice again also that the put shape showing up there's no statistical impact of the in the positive range of the stock market. So when the market goes down they downgrade the expectation when the market goes up they have an asymmetric response. In terms of the actual predictability you get look at the top left you get basically 10. For unemployment column 4 you get 6 for the fact we had 10 and 5 it's all looking pretty consistent. So based on this first approach the fact is looking pretty fashionable. The second approach was the Taylor rule. So column 1 is just the change in the Fed funds target on some lags. You've seen that one before. Column 2 we use an information criteria to pick out which variable wants to enter the Taylor rule growth, inflation and employment. And then column 3 we do the horse race and you can see that in the pink and darker pink growths that once you put in the growth and inflation data in particular the growth is what matters here. The coefficient on the stock market becomes almost insignificant and also in world 1 the auto correlation is a lot lower actually. So if you calculate out how much economic pay-to-powers that the stock market once you put in the green book the green book expectations there's not much. So a 10% drop in the stock market as we saw in the very first graph on the target moves the target by about 100 basis points cumulatively. Without the green book controls once you put those in you're down to 21. It turns out actually that consistent with this asymmetric thinking in column 4 asymmetrically to growth so more it turns out it reacts more to low growth and to high growth then the stock market is no longer significant. So in that sense the fat is looking pretty rational. So let me show you just my final there's various reasons it could be rational to react to the stock market above and beyond growth but since I don't that's not as interesting let me just show you how much the this is how much the Michigan care about the this is how many of those Michigan respondents say we have heard about some concerning economic news and it's about the stock market. So this is the households concerned about the stock market let me just overlay here the fat's concern about the stock market let me do this again just because it's fun the correlation is 0.68 so again the household overall the fat is looking pretty good in the interest of time I'm going to skip the other counts and look forward to many of comments. So thank you the second one yeah so thanks a lot for giving me the opportunity to discuss this really interesting paper it had been some time that I wanted to read it so this was a good commitment device to finally sit down I'm Central Banker so there's a disclaimer on the slide so this is a really interesting paper that studies not just one but a set of important questions the paper contains as you could see contains lots of results many of which I won't be able to do justice today in my discussion so I will really just focus on primarily the first question that Annetta and Anna raised which is whether the stock market predicts decisions relative to other news about the macro economy among other things the paper uses textual analysis on FOMC minutes and transcripts so there's really a new innovative methodology aspect to the paper which I liked a lot the main conclusion of the paper is that the stock market causes Fed policy and if confirmed I think this is a really important result the outline of my discussion will be I will first very briefly summarize the paper and then I have essentially three comments or questions the first question is I want to challenge a little bit the office about the systematicness of the Fed's response to the stock market then I want to ask whether it is really about stock returns in particular or financial or economic conditions more broadly what the paper cares about and the stock market being one indicator of those of course and then I have a comment on the econ metrics so let me jump right in so the paper studies the relation between the stock market and the stock market outcomes and Fed policy the key findings are and that did a really nice job summarizing the results I'll just go through the four bullet points that sort of is my read of the main results the Fed's rate cuts are predicted by very low stock returns in proceeding FOMC cycles this is interpreted as the Fed cushioning equities in bull markets and so Annette and Anna call this the Fed put the stock market mentions in FOMC meetings by staff but also by FOMC participants predict target rate changes and the reason for this is that FOMC believes in something like the wealth effect okay let me jump into my comments right away so the first question is is this really a systematic reaction that the Fed shows to stock market downturns when you look at the sample period that Anna and Annette consider you see that there's essentially here I'm plotting the Fed funds rate onwards up until 2016 the main part of the analysis focus on the sample 1994 to 2008 what you see is that there is essentially only two easing cycles over that sample and so over this period Annette and Anna identify a Fed reaction to the stock market and the two easing cycles in the sample were 2001 which is essentially just after the the dot-com bubble burst and 2007-2008 so essentially the financial the financial crisis and the ensuing global the great recession so not surprisingly both of these years 2001 and 2008 are highly inferential for the document relation between negative stock returns and future Fed funds rate changes so Annette was so kind as to send me the data so I and here I'm just showing basically the first set of regressions on an annual basis on the left we have the Fed funds rate changes on average over a given year on the right we have the the Fed put variable the negative stock return accrued over that same year or the previous years FOMC cycles I should say and what you see is that basically this negative relationship this positive relationship is really seems to be driven to some extent by the 2008 and 2001 observations and so when you when you exclude those two years and I'm not saying you should exclude those years but just as an observation if you exclude those two years then the relationship seems to be gone so these two years which have seen important financial crisis if you will the dot burst of the dot com bubble an important financial market event and the financial crisis is an important financial market event when you exclude those two years there's doesn't seem to be much of a systematic relationship between stock returns and the Fed policy now looking at individual meetings we see that there's particular meetings that seem to be very influential for this documented relationship and those are the meetings on March 20th October 29th 2008 so so what was going on on those days let's let's ask the FOMC so just reading from the FOMC minutes on for March 20, 2001 we see that the members saw clear downside risks and the outlook for consumer investment spending in the context of the market decline that had occurred in equity prices and consumer confidence and unexpected business profitability and they were concerned that weaker exports might also hold down the expansion of economic activity the minutes from October 29th, 2008 we all know this is a couple of weeks after Lehman came down over here it's a nice summary of what had happened actually over this period a number of adverse financial developments influenced economic and financial market conditions over the inter-meeting period Lehman brothers holdings that filed for bankruptcy the day before because of losses in Lehman that the net asset value of a major money market fund fell below $1 per share spurring a substantial outflow for money market mutual funds and straining their liquidity the rapid deterioration of AIG and Vakovia along with the closing of Washington Mutual led to intensified market concerns about the condition of financial institutions and now importantly in this environment investors pulled back from risk taking funding markets sharply and equity prices registered steeped declines so does the FOMC only care about the stock market when deciding on policy rates in times of crisis according to these minutes clearly not there's at least only in these two meetings minutes there's a number of other variables that I mentioned consumer confidence expected business profitability exports funding markets for terms beyond overnight credit risk spreads and so on but above other economic financial conditions one needs to show that the stock market put variable survives when controlling for all these other indicators and the authors are aware of this issue so they do control for economic macroeconomic news unfortunately they only study the predictive content of the put variable for federal funds rate changes in comparison to and not together with macro news and in these regressions and then at the results individual macro news but we know that the stock market is a good information aggregator so I'm not quite sure what we learn from this regression just comparing our squares of regressions of federal funds rate changes on past stock market returns in comparison and not jointly with macroeconomic news so let's just and that gets me to my second point whether the relationship that and that on a document is really about stock returns in particular or financial and economic conditions more broadly so let's jointly consider economic and financial conditions in the federal funds rate prediction for example let's use consumer confidence and credit spreads as additional regresses and those were two of the variables that were mentioned in the minutes of these important events and so specifically I replicate one of the regressions from the paper of changes in the federal funds rate two lags of that same variable and then the stock market put variable so the negative of this accrued stock returns in the previous in the meeting period and it's lag and then I will add to this regression consumer confidence news and a variable that essentially measures the increase in credit spreads over the intermediate period and I will just to be able to compare this I will also include their lags as regresses and then I will just compare the slope coefficients for the stock market put variable and it's lags before and after adding these additional variables so importantly macro news are only available from 1996 November onwards so just to confirm that the put variable alone remains the put variable on the left and then it's lag they survive over that sample so that's important now adding consumer confidence and the BAA spread the put variable largely uses significance so this tells us that essentially when you add other measures of economic and financial conditions that probably the Fed cares about the put variable the stock return the negative stock returns accrued over the intermediate period again the 10 percent level for the for the lag but it really becomes much weaker when you add other measures of financial economic conditions for example the Philly Fed business outlook which also is a strong predictor of Fed funds rate news there's essentially no explanatory power left so this is just to basically say that I think stock market downturns are one of many indicators of an important role for Fed decisions the stock market there's two little quibbles also that I just wanted to highlight much of the explanatory power of the put variable actually doesn't come from the accrued negative stock returns in the intermediate period but this lag so this means that the the negative stock returns between the second to last and the last FOMC meeting predict target rate changes today and that is thoughts on this more over this lag put variable interacts strongly with the second lag of the Fed funds rate change so the dependent variable that also enters the regressions and I couldn't think of a rationale for having sort of two lags of the Fed funds rate change in the in the regression so this is just a minor thing final comment on the econ metrics the put target return and zero so this variable is essentially a truncated censored variable so more than half of the observations in the 94 to 2008 sample are zero of this variable and now we know from Rigobon and Stoker they have a number of papers on this that a linear regression with censored explanatory variable variables leads to bias and efficiency loss and Rigobon Stoker a particularly insidious practical problem namely estimates that are too large the phenomenon which return expansion bias can give a spurious impression of the importance of a regressor and so quite ironically they actually label this expansion bias in the context of this so a nice way to actually document or illustrate the issue that can arise if you have a censored right inside variable is from this basically have a lot of observations that you truncate at a given value in their example it's 80 but in Annette and Anna's paper at zero then you can basically find a larger regression slope then might actually be in the continuously registered variable so that's a potential econometric issue that I think Annette and Anna might want to look into turns out the dependent variable the fed funds rate change also has many zeros even though it's not zero censored it's basically oftentimes the fed doesn't change policy rates a question that I had and couldn't answer immediately was whether this amplifies the expansion bias that Riegelman Stoker mentioned or whether this potentially also characterized the efficiency loss and linear regression with censored regressors and proposed a maximum likelihood estimator for normal mixed censoring models so that's maybe something the authors could consider as a robustness check they also characterized the bias and linear regression with censored regressors in their follow up paper and it seems to be just taking a quick look at the results and so potentially there is a risk of a bias here so the authors may want to make sure that the censoring doesn't unduly affect their results so with this let me summarize and conclude this is a really interesting paper studies an important question or studies important questions the findings suggest that there isn't a symmetry in the fed's response to stock price fluctuations in my view the jury is still out as to it's important in and of itself or simply as a proxy for broader financial and economic conditions and I encourage everyone to read the paper to form their own views thank you thank you Manuel Annette you want to add thank you very much for those comments I agree that we only have two crashes in the stock market between 94 and 2008 one of the reasons we started looking at not only the fed funds target but how much the fed drives the stock market was precise in our first paper was precisely to get more high frequency evidence so in the picture that I showed you from the stock market I showed you weekly stock returns and then I showed you today's stock market return in even weeks and that graph we are able to run all the way up to you know currently that gives us more data both in the sense it's more high frequency but also because we have data during the zero lower bound period so it turns out the fed put as measured by whether the stock market mean reverts in even weeks is if anything stronger in the zero lower bound period before so that gives us a little bit more data if you think about the recent history one of the cases where the fed really got busted in the financial press for acting like it reacted to the stock market was in 2015 when they delayed liftoff because the stock market you may remember that event so I agree that we only have as much data as we have but by looking not only at the target but also at the stock market we can do a little bit better another thing about that's useful is that in a text or an analysis that you can follow from week to week or from meeting to meeting how much to talk about the stock market that also is having things that's going to be too regardless of what expenditure variable you have how it lines up in 2001 and 2008 is going to matter in terms of other financial conditions I didn't have a chance to show you but we have graphs for that in the paper and I completely agree that they likely care about other financial conditions as I said the main ones they care about are short and long rates credit spreads they say that's the one where there's something new going on they start talking about that in the mid 90s which is when we see the stock market put showing up both in the target in the stock market in the text or analysis so in fact the reason we started doing text or analysis was because we were running horse races we thought okay are they caring about credit spreads the dollar the stock market you know with the one time sales you have for the target it's the statistical nightmare to sort this out so then we said well at least these guys will tell you what they're thinking about so let's figure out okay are they what are they talking about when and that gives you essentially more insights than just one linear regression I don't think that just one regression is going to sort this out but the text analysis hopefully will tell us that they care about a whole bunch of financial conditions including the credit spreads and so forth but what they newly care about in the period since 94 is the stock market and in terms of the big one and SAC we should definitely read that thank you for the reference that we have done as non-parametric relations as we could that's why I showed you this bucket I showed you graphs where I just had here's quintile so the stock market and here is how the target changed there was because we've also you know concerned about what is the right functional form and you saw that it was basically flat except in which again tells you that you know results are going to be sensitive to whether you if you don't have any stock market crashes in the sample period you know you cannot estimate the relations so thank you very much for the comments and hopefully you guys have some more questions so there is growing empirical evidence that financial conditions affect the downside risk to economic growth but not really the upside of the distribution and so if you take a risk management approach to monetary policy whereby it's easier to react to booms than to busts this could provide an additional justification why they are so overreacting to stock prices they are reacting to stress to financial conditions that will transmit to create a downside risk to growth and to the economy in general so you went very fast I want if possible for you to talk a little bit more about what happens when you introduce growth into the picture because I was wondering what is your interpretation of the finding if it's that mainly the Fed cares about the stock market as a predictor of future economic conditions that are not captured by the standard market economic variable current so about or if something it's beyond that something about the consumer sector reacting in a rational way to the stock market or things like that they are beyond economic even future economic conditions let me just reply to those two before I forgot is that okay I agree about the downside risk I had one regression where I split up in the tale of role growth into the positive range and the negative range and definitely they react more to the negative consistent with the approach Veronica in terms of how they think about it basically they think about so here I'm going by Narayana Koshilakota's discussion and John Williams comments so basically they think about the stock market as a demand shock through the consumption wealth effect and I think if you ask by name he would also say perhaps also Gertler that it would go through investments as that's probably some of the investment consumption investment both work their way into the growth forecast and that's why you saw in my tale of roles once you put in the Fed's growth forecast then the stock market was not doing much above and beyond that Fed thinks in terms of the wealth effect when they judge movements in the stock market and whether they should do anything but I also think that they don't believe that very high frequency movements in the stock market are going to have impact on consumption it's sort of large persistent effects so what that suggests is what Emanuel was talking about the two periods where your story makes the most sense are the 2001 drop and the 2008 drop so here's a possibly half baked thought compare your simple model over those two periods where you're going to pick up the drop and the rates associated to the stock market with Emanuel's broader model this is at the stock market and the question is how does Emanuel's model do around that period does he miss the drop in interest rates so if that's the case what that would suggest is maybe screening out small variation in the stock market and just kind of running a regression where you just look at big drops as the market understood that the Fed reacts to the stock market is a good driver of the expectations about Fed decisions and also quickly what is the impact on the equity risk premium of the fact that the Fed reacts to the stock market there's another one the US has a very high stock market participation so the wealth effect is really pronounced as known not to be as important and see whether you pick something similar which would then be indicative of whether you're picking up other non-genuine stock market effects or not let me just thank you very much for the suggestion you had asked if the market understands the Fed put in I would say you know not full yet I mean if you track the number of media cases that if the stock market goes down and today if the stock market went down over the past week and today is one of those even week days where we think the Fed does its decision it's still the case that you see high stock returns which you shouldn't see if everyone had caught up so in that sense it seems like the market is sort of catching up but still the Fed has come out because everyone becomes aware that there is this insurance then you know bad things could start happening but it doesn't seem like the market is fully caught up at least not yet and in terms of the equity risk premium in our first paper we we show that VIX moves exactly opposite to the stock market and so if you use Ian Martin's equity premium estimate from the what seems to happen is that the way that the Fed increases the stock market following stock market declines is through a sharp reduction in the equity risk premium and we suspect that there are some element of whatever it takes in that that the Fed comes out and says look we in the Fed's terminology we stand ready to act as needed as opposed to do whatever it takes but refit in terms of the comments on other countries that's a great that's a great comment we had also checked the housing mentions since if the Fed believes in wealth effects driving consumption since the housing wealth effect is generally thought to be larger than the stock market wealth effect they should talk about that as well and they do in the housing crisis but not you know that's also one of those more in wealth effects in recent years it would be interesting to see whether the ECB reacts strongly to the stock market given that we saw that there was little relation between the stock market and inflation you know a central bank focusing on an inflation mandate we want to react less strongly to the stock market than one without so I suspect someone will check that going forward I think that was it another question there Bernard so this is a very interesting paper one question is what happened in 1994 is it the Fed's reaction function so Greenspan giving up on warning against irrational exuberance of markets and becoming a cheerleader for the new economy and then the credit and housing markets or is it something structural like stock market participation or the wealth effect becoming more important so maybe a bit of a storytelling about 1994 your message is the Fed is pretty rational now which Fed the early Fed the late Fed some doubts on the house on the wealth effects the work by John Mulebauer and John Duker shows what really matters is this aggregation so the debt and then housing and financial wealth separately and then one hypothesis which goes a bit in the direction of Keynes and Shin and others the beauty contest so what matters is not the wealth effect it's just that some people believe there is some wealth effect that's enough which is a confidence effect really not a wealth effect and that brings me to the final thing a manual co-author to be as Adrian has propagated this term structure of risk and the GDP at risk which actually means that you should be most worried about financial conditions when they are actually buoyant because that's a good predictor for a crisis in the medium term a few years out so how do you relate that you should start to be worried about prices going up and not going down so maybe the early Greenspan is right not the late Greenspan so I close my remarks here so in terms of why 1994 so in the actual data the stock market is a perfectly good predictor of growth and unemployment even all the way going back to 1947 so in that sense it's the late Fed that's correct not the early Fed my colleague Martin Lettow gave me a little bit of the history of what exactly happened inside the Fed as to when they started focusing on this and you may remember the dates of the famous Lettow-Ludwigson papers about the consumption wealth effect and the consumption wealth ratio predicting the stock market there was some arguments between the New York Fed of Lettow-Ludwigson and then the Fed in Washington about the size of the consumption wealth effect and you can follow that sort of in the various articles and just one final comment in Narayana Kochalkota's discussion of his paper it was interesting he said the Fed actually hasn't gotten it right yet because if you see that the stock market is tanking still predicts low growth then we're not doing enough so in his view that you keep going until the predictive power stock market goes away but as a spectator there will be some disagreement with that on the committee where Thank you thanks a lot to both speakers there is a coffee break now please come back quarter quarter to five