Moody's Talks - Inside Economics - Flipped Signs and Forecast Philosophy

Episode Date: July 23, 2021

Marisa DiNatale, Senior Director at Moody's Analytics, joins Mark, Ryan, and Cris and they recall their own favorite and least favorite forecast of all-time. They also discuss different approaches to forecasting, the meaning of being accurate, and who is a hedgehog or fox. Questions or Comments, please email us at [e-mail]. We would love to hear from you.  To stay informed and follow the insights of Moody's Analytics economists, visit Economic View. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Welcome to Inside Economics. I'm Mark Sandy. I'm the chief economist of Moody's Analytics. It's good to be with you. I'm joined by a few of my colleagues. Of course, we've got Ryan Sweet. Ryan's the director of real-time economics. It's good to have you, Ryan.

Good to be here, Mark. Yep. Took you a little bit of a pause there. I don't know. I'm getting over a man cold. So it's been plaguing me all week. So I'm on the mend. What is a man cold?

Yeah. Why is that different from a cold? Oh, so when I get sick in my household, I'm the worst. I'm like, I'm in bed for a while. My wife had the same thing, and she bounced back in the day. And she's like, you must, you know, it must affect men more than women. I'm with you on that. Totally agree.

Yeah. I said it's factually correct. And she's like, no, it isn't. She's like, get Fauci on your podcast and ask him that question. About the man called. That's a good idea. We should get them. on here. Try to get him on. And we got Chris Deereides. Chris is the Deputy Chief Economist. How are you

feeling, Chris? You're in tip-top shape. Feeling great. I think audio's loud and clear this week. Yes, indeed. Finally, you've got it together over there, buddy. Good way to go. And we've got Marissa, Marissa Dina Tallon. Good to have you on the podcast. Marissa. Great to be here. And Marissa, I know what you do, but what's your title? You know, how would you, well, how do we, how do we call you? What should we call you? Well, I'm a senior director. Okay. And head of global forecasting. That's it. I knew it. It was a big, big deal. Head of global forecasting. By the way, listener, it's key to have Mercer on this week because we're, the big topic is the philosophy of forecasting. So a lot to think about

there. And I'm glad to have Marissa on board. Mercia is located in one of the, the best places on the planet, Newport Beach. Do I have that right, Newport Beach, California? I'm south of Newport Beach. I'm in Dana Point. Okay. Oh. On South. Okay. All right. Yeah. Orange County. That's a step down from Newport Beach. No, not really. It is considered quite a step down from Newport Beach in some circles, yes. Is that right? When did you move for, you used to be out here with us. You kind of abandoned us. How long ago was that? I wouldn't say I abandoned you. Yeah. I've always been with you in spirit, if not physically.

That's true. Six years ago. I've been in California for six years. So you must be like a gazillionaire by now. You bought a home? Did you buy a home six years ago? I hope you did. No, I bought a home three years ago.

Still, you're half a gazillionaire. Yeah, it's pretty crazy what the value of home is right now. Yeah. So can I ask what's the value? No, you don't need to tell me. Oh, don't tell me. Wow. I'd love to know, though. I'd love to know. I can, we can discuss offline. Offline. Okay. All right.

Marissa, you, before you joined, how long ago was it when you joined Moody's Analytics? You joined the team? Well, it was 2004, so it was still Economy.com. It was the before Moody's boughteconomy. Got it. So that was, boy, that's, yeah, I'm coming up on my 17-year anniversary. Yeah, very good. And you came to us from Bureau of Labor Statistics, Yes, I did in D.C. Yeah. And what did you do for BLS? I worked on the current population survey, the household survey, and I was in the office that I wrote the employment report. Oh, you did. So are you, like, bummed out when people kind of criticize the household number saying, you know, that don't pay attention to it because it's a small, I think it's 60,000 households,

isn't it? It is, yeah. Pretty small sample. Or, compared to you. to the establishment survey. Did they ever bum you out? No, I think if you know what it is and what it's measuring and what it's drawbacks and strengths are, then I think people understand the differences between the two. I mean, it's true with any survey, right, regardless of size, there's going to be some error, margin of error in there. So certainly when you're data mining, that set for very specific things, you're going to run into huge standard errors. Data mining. Who's doing that?

Are you accusing how to find a data mining? A lot of researchers use the current population survey microdata to look at, you know, very, very fine cuts of demographics or, you know, geographies, things like that that once you start cutting and cutting and, you know, looking at. Hispanic female 15-year-olds who live in Chester County, you're going to have a huge margin of error when you start cutting things that way. Yeah. I mentioned Dana points in Newport Beach area.

You have crows outside what's going on outside your window? You had to beat them off of something with a stick or what's going on? There are these enormous crows or they might be ravens. I think they're crows. And they live here in the trees and on the power lines. And they're the size of a small dog and they're extremely loud. Sounds like a Hitzcock movie. I'm just saying.

Yeah, we were just joking about that. Yeah, they'll land on my roof. And it sounds like there's a person jumping up and down on my roof. That's how big they are. Well, I want you to know, Marissa. Terrifying. Before I came to do this podcast, it was on my back deck. and there were these beautiful little hummingbirds floating around.

I have those too. I have hummingbirds too. Yeah. The crows are probably out there eating them right now. And of course, I'm only in, you know, pedestrian suburban Philly.

So, and I've got my, well, that's where I'm from, Mark. So I love for Chester County as well. Well, I think we should move on.

What do you think, Marissa? We should. Yes. Okay. We've caught up. So, as you know,

and I know you're a careful listener to our podcast that we begin with the statistics before we move into the big topic, which again, today is the philosophy for forecasting. And I'm going to really mix it up. Mercy, you're going first. You are going to go first. Usually I make Ryan go first, but this time, let's just really mix it up here. And you played the game, right? You tell us a statistic. You don't tell us what it is, but we have to, based on our, hey, by the way, I should tell you this, last week we did a podcast. Remember, Ryan, your statistic was Chinese GDP 1.3%. Right. And I got it right. Am I right? Am I right about that? You did get it right. So I've got people throwing

shade at me saying that must have been, the fix must be in. You must have told me before the podcast what this. No, no, we never share. Exactly. Exactly. I want everyone to hear. Everyone, out there to hear that. We don't share. This is, we take this deadly seriously. Okay, Mercer, with that as a backdrop, what's your statistic? Okay, well, first let me preface this by saying that Ryan was trying to get in my head and psych me out all day yesterday. No, yeah, there's games going on. So, okay, the statistic, actually I would like to do two, but let me do the one right now. You see what happens? She comes on and she's trying to change the rules already. I mean, come on. Okay. I have, okay. So the statistic I would like to use is actually from

last week. It came out last week. Okay. Fair. Okay. I didn't talk about it. It didn't come up in previous conversations. So is it still fair game? Yeah, yeah. Yeah, far away. Yep. Yeah. The statistics that came out this week, none of them really spoke to me. And it was a lot of housing, which I figured Chris would use those. Who knows? Who knows? Who knows? Who knows? But probably. We're forecasters and probably. Okay, so my statistic is 39% in June. 39%. This was last week.

By they refers to June. Yes. This month over month, year over year. Is it a growth rate? No. Oh, okay. Oh, so it's like a diffusion index or no. She's shaking.

head no. Give us a hint. Got to give us a hint. It's just it's a percentage. A percentage. 39% of something. I'm drawing a blank. Can you give us can you give us a hint without giving it away? We should know this you're saying. You think we should know this. This is in our strike zone or close to. Yes, but it's a release with a lot of numbers. So this might be a little sweet ask. It might be one of those Should I tell you what the releases and then? Yeah. Yeah. Yeah.

Okay. So it's from the NFIB. Oh, I should have known that. Yes. Which. Well, the percent of firms that said they were raising prices, that was more like, that was higher. That was like 49 percent, as I recall.

I think that was the highest, it might have been the highest on record or you have to go back to the 80s, you know. Yeah, that's true. So 39 percent. You're a labor market. market person. So maybe a percent of firms that say they can't fill an open position. Yeah. Trouble hiring.

That's wrong. Plan on increasing compensation. 39% say they're going to increase compensation. Close. Oh. It is the percent of firms that are actually raising compensation. Okay. There's a second question of plan on.

Yeah, Ryan gets the prize. No, no, he does not. No, no, no, no. So none of you do. Yeah. Can I ask so how high is that? I mean, I'll put some context. It is the highest ever recorded, going back to the inception of the survey.

So 39% of respondents, these are small businesses that report to the National Federation of Independent Business, said that they actually raised comp in the month of June. Yeah. Highest ever. And it was, yes, and it's a little bit higher than where it was, you know, prior to the pandemic start, we had a very tight labor market, obviously. So all these surveys were also showing employers saying jobs are hard to fill and there might be upward wage pressure. It almost, it was, that was kind of like the second highest data point, but this is the highest in a single month. That is a really good one. I'd have to say, kudos. Yeah, good job. That was really, you know, Chris should have known that, but, you know, I'm not working hard. Off. Yeah.

He's off today. He doesn't have a cold. Okay. So you said you had two. You want to do the second one? So the second one has to do with our topic of discussion today. So we could maybe save it for the intro to that. Okay.

Well, let's do that. We'll in fact do that. Okay. All right. Hey, Chris, you're up. All right. This one did come out this week. 0.09.

0.09. Is that a change? Nope, it is an index. Is it one of your weird housing surveys like from Fannie Mae? No, no, this is a indicator. This is an indicator that's covered on Economic View. It is. Shout out to Sarah Crane for covering it.

So ag prices? She covers. Oh, it's from the, is it from the Chicago Fed manufacturing? Oh, I don't. No, no, no, no. I know. I know. I know.

I know. I know. I know. But only because I. Activity index. That is correct. There you go. I got, yeah.

But Marissa, I think I think she was there first. Actually, I want to be. You'll get used to it. We get marked 95% of the way there. And then he, you know, slams it. home. That's sort of true. That is sort of true. I wouldn't have gotten it without Marissa. That's definitely true. So explain that statistic, Chris. What is that statistic? So this is a measure put out by the Federal Reserve of Chicago. It's a composite of 85 economic indicators that are designed to measure

overall activity and assess inflationary pressure. It is, it's down. So it's down from last month. last month it was 0.26. And I should say the measure is centered around zero for trend growth. So a measure of zero is trend growth. If you're positive, it's greater than trend. Negative is below trend. So it's still, it's positive. It's still faster than trend, but it is slowing. So it does suggest that we're losing some momentum in the economy. Of the components, according to Sarah, production, consumption, and the employment factors were all down. what pulled it down. Only sales activity was up. This is the month of June for the month for the month of June. Correct. And just so everyone knows, this is my interpretation. Trend growth means enough growth

to maintain stable unemployment. So if you're above trend, you should be getting enough job creation for unemployment to decline. Right. Yeah. That's that's fair. And of course, that statistic goes up and down and all around month the month. I mean, I think if you take a step back, it's still up pretty strongly, right? It's still indicating growth above trend, growth that's sufficient to bring unemployment down. Yeah, so she reports a three-month average of 0.06. So three-month average, so it's smooth, and it was 0.8, right?

So it's actually been slowing considerably. But on a six-month average, it's been pretty stable around 0.35, right? So that indicates some, yeah, overall the growth is still there, but it's signaling at least some slowdown here or some potential for slowdown here. You know, something is, I'm getting a lot of questions about now. Or maybe I'll wait to my statistic before I bring this up. In fact, the point, the question is around the delta variant,

and what kind of impact they're starting to have. But we'll come back to that. Oh, I hope I didn't give my statistic away. I already know what it is. I do you really. Okay. That's a slip-up. They call that a slip-up, Marissa.

Yeah. But you guys would have needed a hint anyway. So, but I'll let Ryan go next. Ryan, what's your statistic? All right. I'm not going to go out two decimal places like Chris did, but straight 152,000. 152,000. 152,000.

Came out this week? It came out this week. It feels like it comes from the claims report. No? No. Pandemic initial UI. That was like 100,000. That was 150,000.

152,000 is it something buried in the bowels of the existing home sales numbers? No, you're on the right track. It's a housing statistic. Statistic? It is. So, Chris, you should know this. What would that be?

Yeah, 150,000. You know, it feels like something like the number of homes, newly built homes for sale or something, some really low number. You're getting closer. I know you guys are going to yell at me for this one, but this one's really important. It's a really important number that we track.

Okay, give us one more hint, and then we're going to, we'll give. All right. It is, all right, so it's housing. Okay, housing. And it's something that will help cool the housing market, house price growth. Right, so the supply. So supply. permits of some sort.

Getting closer. It's too low. Too low. Yeah. Is it mortgage application? No. All right. I'll throw one thing out there.

I'll just throw one thing out there. Just throwing it against, just throwing, you know, destroying flak out there. Number of manufactured homes to build on an analyzed rate. No. But that's such a good guess, Matt. I'm close, though. I'm close. It is like that.

All right. It is the number. number of homes that are permitted but haven't begun yet. So it's a future sign of housing supply that's coming. Highest since the housing boom in 06. Say that again? What is it? So it's the number of homes that are permitted but not started. So it's kind of like the pipeline of new construction that's coming. And that's the highest since 2006. So it's between a permit and a start. Correct. Oh, I see. So it's possible it doesn't get built, right? Also. Yeah, cancellations, things like that. But for the most part, that that cancellation rate's

pretty low. Well, you know, Brian, it might be different though. I mean, it might be the supply side effects, right? The builders just don't want to, they got the permit. They're just holding it in the permanent inventory, so to speak, because they don't actually want to go out and buy the lumber and the materials at these inflated prices. So maybe that's reflected. Well, lumber prices have come down a lot. They're still high, right? They're still double what they were pre-pandemic. So maybe they're waiting. If I were them, I'd wait another few weeks to see if it comes down even more, maybe, right? Right. I was optimistic. I was like, here comes some more housing supply. No, no, but that stuff that's going to come. It's there. It's reason to be optimistic, but, you know.

Oh, yeah, I agree with the time. We don't know when, but it's going to come. Yeah, it's going to come. All right. I'm going to give you kind of a layup because we're getting really, tough with these statistics. 4.1 trillion. This is a layout. Should be a layup if you're paying attention. It came out this week. Well, it's a statistic.

What does it mean to come out this week? It happened. 4.1 trillion. I'll give you another statistic, another related statistic. 3.5 trillion. Oh, my goodness. Oh, the fiscal policy.

Yes. Because, okay, Chris, What is it? Democrats versus. Okay. Right. So 4.1 trillion is the total sum of the fiscal support being proposed by in Congress. That's the $3.5 trillion reconciliate, so-called budget reconciliation package that the Senate Democrats are working on.

And plus the 600 billion, did my arithmetic right, yeah, 600 billion in infrastructure spending for the bipartisan bill that the Democrats and Republicans are working on. So 4.1 trillion in total spending tax credits to support infrastructure and social investments going forward. That's a lot of money. A big number. And that really will have what happens with that legislation, whether it gets through to law in what form and to what degree, what size goes a long way to determining our outlook, our forecast for the economy. And we're getting pretty close to coming up to the big topic, the philosophy of forecasting. So very important.

Before we get there, though, the statistics we've been following week to week, anything interesting happening with 10-year treasury yields, Ryan, anything we should note? They're still going down, I think, right? What were they? They took back up. Have they? Yeah, last time I checked earlier this morning, like 1.3. Okay.

So last week we were flirting with 1.2. So we're up a little bit. Yeah. And if you go back a few weeks ago, a month ago, it was 1.5, 1.6. So we're down 20, 30 basis points. Correct. So of that, how much of that is more technical related to debt ceiling, bond issuance, that kind of thing?

How much of that is more fundamental related to bond investors growing nervous about the Delta variant? Yeah, it's almost evenly split. Like 50-50, because you have inflation expectations that come down quite a bit. So that's bringing the 10-year-get down by probably 15 basis points. And probably the rest is these technical issues on top of the Delta variant. I would group those two together. By the way, the other statistics I was thinking about doing, which you wouldn't have gotten, but it would have been unfair.

That's why I didn't do it. But it was a good one was the purchasing manager index for the UK. It just came out this morning or this afternoon, 57.7. in June, that's down from 62.2 in May. And if you go look at the commentary, a lot of the slowdown is in transportation and leisure and hospitality, which feels like Delta variant is having an impact. Obviously, the UK being significantly impacted by the variant. Are you guys noticing anything in our data at all, the macro data for the U.S.,

suggesting that the variant is having any kind of impact on the economy, macro or regional impact. Have you seen anything yet? No? So for that, I would look at like open table. Nothing really there yet. The Google mobility, nothing that screams Delta variant.

You know, the weakness in jobless claims, which is Chris's number, that's seasonal adjustment problems. That's not Delta variant. The one thing that's been weak is the home-based data. But I don't know if that is Delta-variant. It might be too early. What's that?

What's the home-based data? data for everybody? So this measures hours worked, a number of people on employment. It's like a good high frequency indicator of the job market. Right. You know, we put that back to normal index together, which is a compilation of a lot of those statistics you just mentioned, including some government data. We do it at the state level.

And I have noticed that our back to normal index, which is equal to 100 right before the pandemic, is fallen back for Florida. If you go back a few weeks, say a month, six weeks ago, Florida. Florida was actually all the way back. It's back to normal index was over 100, back to pre-pandemic. But now it's back down to the low 90s. So I wonder if that might be because Florida has been hit pretty hard by the Delta. And there's a lot of hospital.

In addition to infection, there's a lot of hospitalizations there. So maybe we're seeing some ill effects there. But it's really pretty modest at this point. If there's any effect, it's really on the margin. Yeah. Okay. I think it's a month or two and we'll start to see it in there in other states. because even here in California,

mask mandates are back in L.A. County and in parts of Northern California, and I think other counties here are looking at doing that. And if that happens, right, it's just people are going to be going out less in general. Yeah, what do you think it would take for the Delta variant to become a U.S. macroeconomic issue for it to actually affect the economy in a negative way?

the macro economy, not a state, not a city, but show up in the macro data. I mean, it would have to be widespread business restrictions and lockdowns again, but I can't envision that happening. What if schools did not reopen for in-person learning? Would that be enough? That would certainly show up in the data. Certainly, certainly show up in the data. Yeah.

Just even, you know, all the stuff you guys have been talking about the last few weeks about labor force participation, people staying out of the labor force to care for kids, labor supply shortages, that would certainly worsen those things, I think. You know, guys, I think maybe next week or the week after, we might want to have a podcast on this variant, talk about the, because this is global. You know, I was just talking to Steve Cochran, our chief economist out in APAC, and he was, Asia's in trouble. I mean, Singapore's shutting down again.

Malaysia, Indonesia have real problems with the pandemic. So we may want to revisit this in a week or two. Chris, your statistic is UI claims, and as Ryan alluded to, did something weird last week, what happened? Yeah, they jumped up, up 51,000 from the week before to 419,000. But as Ryan mentioned, there are a lot of measurement issues that may discount that movement. So seasonal factors are certainly one. You do have the auto industry scaling back because of the chip shortages. And then also just definitely I got some questions about this measure from some clients

just in terms of how these, how UI benefit initial claims actually work, right? So clearly there's a case where you get laid off, you apply for benefits. Your claim may be rejected, in which case you may apply again. So you might be counted multiple times. And then also, if you are on UI, and then you're applying for an extension, that also would show up as an initial claim. So you have to take that all into account when you're looking at this number. So it could be inflated, if you will, for those reasons, above and beyond just weakness in the economy overall.

So I think important to watch, the four-week moving average is $385,000. That's still flat. So it's indicating things aren't really getting worse at this point. But, you know, certainly something we want to. So, Ryan, I know you use, or at least historically, use UI claims as one of the key inputs into your estimate of the job number for the month. So do you have an early estimate? I know it's early because this doesn't come out until it's two weeks from now, isn't it? It's two weeks, right.

But claims were for the payroll reference period. Yeah, exactly. So does that, what are you thinking? You have a sense of things for the June number at this point? I have a number. I'm not using claims. So the model is a click now because what was Chris was saying with the seasonal adjustment issues. It's around auto retooling.

Normally this time of year, auto manufacturers shut down, retool their plants. That can cause a lot of, you know, variability in the data. Timing of the July 4th, holiday can throw off claims earlier in the month. So July is one month that, you know, my little things to do list for forecasting appointments, like don't use claims in July. Yeah. So what's your number? your preliminary number, just to get a sense of it? 850.

Oh, okay. A strong number. Because last month we got, what did we get? 850? Yeah, right around there again. Oh, right, right. Okay. You think anything, any particular thing going on?

Any reason why? Yeah, I mean, the strength can be inflated again by residual seasonality. I mean, there's always this tendency for leisure and hospitality to be weak in July. that might not happen this time around. State and government education is always very strong. So these two factors combined, you know, really should juice it. Got it. Okay.

And my statistic I've been following week to week is copper prices, because that's a Dr. Copper window into global economic conditions and inflationary pressures. And that continues to be very strong. $4.35, at least the last time I looked, $4.35 a pound. Anything over $4.4. that's a rip-orne economy.

Three is dollars a pound as typical, anything closer to two, which is what we saw back in the teeth of the pandemic would be consistent with a weak economy. So copper says things are still strong out there. Inflationary pressures are still quite intense. So no abatement there. Okay. Anything else on the statistics before we move on?

Anybody want to bring up? I thought that was pretty good. Okay. Before I dive in, Marissa, did you want to bring up your second statistic? You said it was related to the topic of the philosophy of forecasting. And you are the head of global forecasting. So no better person to have on to talk about the philosophy of forecasting than you. You drive all the trains.

And I will say, you are key to our operations. You make things work. And I will say, if there's one disadvantage to being in Dana Point, Newport Beach, is the time zone, right? Because you feel like, you must feel like a little bit out of sorts there, right? Because if we have a 7.30 a.m. meeting in Westchester on the east coast of the U.S., that's, what, 1230, London. That's PM 1.30 PM, Prague. That's, I don't know, 4.30 p.m. Dubai. That's 7.30 p.m. I'm pretty good at that.

this. Yeah, Singapore. That's 9.30 p.m. Sydney, and that's 4.30 a.m. Dana Point. If you have a 7.30 a.m. meeting in Westchester, I don't attend it. You don't. You shouldn't have told me that. I had no idea. I thought you were there. You never have 7.30 a.m. meetings in Westchester.

Oh, see, she doesn't. In Westchester? Yeah, yeah, no. I do all the time. What are you talking about? No. Well, I'm not invited to them, so it doesn't have. affect me, I guess. No, I work Westchester hours. So, I mean, I'm frequently on meetings at five, five, five, three, six. Oh, you do. Oh, so you, oh, I didn't know that. So you're, okay. So, so when it's, when it's like five, or you don't leave at five, say six p.m. Eastern, you're done. A three. Yeah, at three in the after. I, it's, it's nice. I mean,

I was never for a morning person. I don't think I still would classify myself as a morning person, but it's nice to have the whole afternoon. Now I understand how you become such a good surfer. This is how it happened. You have like... I've never served in my life, Mark. No, I've been told you you're a great surfer. I've seen video.

Scooby diver. I've never surfed in my life. I have scuba dive. I heard you were in the Olympic trials. Someone who's telling me that. No? No, no, that's all wrong? This is definitely not me.

Okay. I do outrigger canoeing. Oh, well, close. That's close. It's in the water, yeah. It's in the water. Oh, anyway, we digress. We digress.

So what's your statistic? Yeah. So my statistic is 18.3%. And that's a seasonally adjusted annualized rate. 18.3% in recent indicator? Recent indicator? It has something to do with forecasting. I wouldn't, no, it's not a recent indicator.

It's not a recent indicator, but it has something to do with forecasting. It is a forecast. Oh, it is a forecast. That number is a forecast. Whose forecast? Yours. Oh, no. This isn't the nasty.

I can feel it. I can feel it already. This is just a segue into the topic that we're going to, going to be talking about and I thought this was a nice. Okay. Well, now I'm starting to, I'm going to have to, you know, I'm going to have to roll up my sleeves here.

Okay. All right. I got to get ready for game. Okay. Is this marks average error on residential housing starts? No. Thank you, Ryan. Wow.

We're coming back to that. We're coming back to that one. He said, I can feel this. This is, they're ganging up on me, guys. It's going to turn into a rose. Yeah, I'm going to.

18. All right. Say it again, 18.9.3%. 18.3%. I thought you guys would get this. Get this right away, huh? 18.8.3%. It's a forecast.

Jeez, do you guys have any clue? A year over year? No, it's not year over you. It's annualized. It's a quarterly annualized percent change. Okay. I don't know. No.

Has something to the stock market? No. Personalico? No. All right. Okay, we give up. What is it? Car prices.

Car prices. I just want to name every variable that we forecast. Yes, exactly. It is. It is the GDP forecast that we made for the second quarter of 2020 in late March, right after the pandemic started. Hey, wait. That's pretty, that was a pretty good.

What was it, was the actual number? Minus 31%. Oh, gee. Oh, this is Q2. This is Q2. This is Q2. This is Q2. That's right.

This is just for. Q3. Hold on. Wait, wait, wait, wait, wait, wait. What? So you're saying in late March. Do you remember back in March, we had, when the pandemic started in March 2020, we did our March forecast.

and as publishing it, like literally the day after we published it, every place went into lockdown. Right. And we had to redo the forecast later at the end of March because we knew that that original forecast was going to be obviously very wrong. So we have two vintages of a March 2020 forecast. Okay. I think that's the only time that ever happened, right?

It is. Yeah, it is. That's right. And what was the first one that we made that decision to do that? Right. And you're saying our GDP forecast for yeah, was eight was 18 positive 18. Oh, because we had this big minus 18.

Oh, I would have gotten it if it was minus. Did I say, did I not say minus? No, you said, well, at least I heard positive 18. I heard positive. I'm so confused. Yeah. Got it. Okay. Maybe we can edit that part out.

I meant to say, there's no one. Talk about forecast error. Geez, Louise. Well, that wasn't a forecast error. That was a mental error. Yeah. Okay, I got it now.

Okay. So you're saying when we, at the end of March, of course, all hell broke loose mid-March, by the end of March, we said, look, the world has changed because of the pandemic. We're going to have to update our forecast. And by the way, for all the listeners, March 2020 was the month in which large financial institutions first adopted the Cecil accounting standard. Remember the Cecil account?

That is a huge, massive accounting change around loan loss provisioning that the whole entire banking system, all the regulators, all the accountants, everyone was really up in arms, nervous about this. And March was the first time we did this. And so in that month, for the first time ever, and we've been doing forecasts for 30 years, we had to run two forecasts in that month, one at the beginning of the month when we normally do it because of the pandemic one and then month. And what you're saying is at the end of the month,

we forecast it for Q2 of 2020, minus 18, and it came in at minus 30. That's what you're saying. Right. Got it. Got it. So we actually, good forecaster, how do you feel about that forecast?

Do you think we did a good job, given what we knew at the time? time or not? Yeah. Yeah. I think it was. And actually, if you look at the annual average for 2020, it doesn't look as bad. Well, I thought you just said that was good. Doesn't look as bad. I said, yeah. Okay. It's all a matter of framing. It could have been better. It could have been better. Yes, but here, here's a point, though. I think, I think, think the most important contribution we made at that point in time was providing the framework for trying to understand what was going on in the economy. Many, not all, but many recessions

are caused by demand side shocks to the economy. Consumers pack it in and stop spending, businesses stop investing. This was a massive supply side shock to the economy. Now that I say and everyone's lived this. This sounds, feels like, oh, yeah, yeah, that's exactly what happened. But back in late March, that was not obvious to people. That took a lot of work and insight. And we had to really do a lot of work trying to understand what was happening to the supply side of the economy, to output and then figuring out what it meant for the rest of

the economy. So I think that was really the key insight that it wasn't, of course, we weren't the only ones, you know, coming to that insight. there was others that came to it as well. But I mean, I think that was providing that framework for our clients to try to understand how to think about this and what it would mean for the for the outlook, I think was most important, not the number per se. But interesting, that was good.

Yeah, I think if you said minus 18 or so I'm pretty sure, right, right, Ryan, don't you think we would have gotten that? Because I really not say it would have gotten it. Yeah. I think she's, I don't know, 18.3%. That's why we were throwing out all these positive. Yeah, yeah, yeah, yeah. No worries. Well, that was a bust. No, no, no, no. I like that. I was very, that was very informative. That was very informative.

Okay, but now we're talking about forecasting. Oh, and I was saying all these nice things about you, too, before that. You are key, you, and all jokes aside, you are critical to our effort. You know, everything revolves around, I mean, I don't know how else to say it, around you. I mean, the energy, effort, work, the thoughtfulness, the care that you put in. Okay, that's the last thing I'm going to say that's nice about you. Thank you. That's very kind. But it's very true.

But you're the right person to have in this conversation around philosophy forecasting. And I want to begin with the question to whoever wants to tackle it first. We've all been doing forecasting a long time. I longer than anybody, but we're all been doing a long time. What forecast are you most proud of? And then I'm going to ask you what forecast are you most disappointed in? But whoever wants to take that, what forecast are you most proud of? Anybody?

Not proud of the forecast? Chris, Ryan, Marissa? Actually, I'm personally proud of many forecasts. I could take another two hours talking about my forecast prowess, but you don't want that. So since Marissa brought up minus 18. I was pretty proud of our high-frequency GDP model throughout the pandemic. It got pretty close. I mean, you know, the forecast errors were much larger than it was pre-pandemic.

Explain what that is. So our high-frequency US GDP model is essentially a bean-counting approach. So all this data that we talk about on the podcast as it comes out, it feeds into this model and it counts up, bottom-up, you know, all the components of GDP and then estimates on the daily basis what GDP is tracking. And, you know, throughout the pandemic, it was getting pretty close. Yeah, I totally agree with you. It's done a marvelous job.

And what's it forecasting for the current quarter? Because we're going to get that number next week, GDP, right? It will change because we get durable goods next week. But right now it's 7.8% annualist. You notice Ryan's a particularly good forecaster because before he gives you his forecast, he gives you all, oh, this, that, you know, we got more data coming in. and, you know, it's going to get more accurate. It's just, it's the truth.

It's, so what's the number in it for, what do you, what is it saying? 7.8% which is down, right? It's come in a little bit. It's come down a lot. It was well over, it was well into the double digits not long ago, I think, right? Yeah, the initial estimate was 10. So what's coming in? What's causing it to come down?

Why so much lower? Equipment spending has come in. I mean, it's still strong, but it was, it was thought to be like, you know, gangbusters. Yeah. The housing data has softened in the retail sales numbers, you know, that kind of thing to consume or spending. Yeah, a little weaker. Yeah, interesting.

Well, since I have you, I know this is tough. No, you're very modest guy. So the tougher question is what you're proud of. So let me ask you, what forecast are you most disappointed in? It was probably one of the first forecasts that I did here. and it would be New Orleans employment after Hurricane Katrina. So Katrina hit and I had just started. I started in July 2005 and this massive hurricane hits and devastates Louisiana.

We had to come up with a forecast of what the recovery is going to look like. And I thought it was going to be this big rebound, like a very V-shaped type recovery. It just never came to fruition. So that is a disappointing. That's a really good one because it's easy to make that mistake, right? Because we rely when we do our forecast on history, right? So in the case of these natural disasters, we go back and look at other natural disasters. Well, what the hell happened after that disaster?

You know, you get insurance money coming in. You get federal aid and you get rebuilding pretty quickly. And so the economy does come roaring back, right? So you use that as the basis for your forecast. But in the case of Katrina and New Orleans, it was so devastating and so many people left New Orleans that never came back. Right. So that. Right. And the flooding.

I didn't factor that in and how long that would take to, you know, get the rebuilding and cleanup effort underway. Yeah. That's very interesting. Very good. And of course, with all our climate risk work we're doing now, we're spending a lot more energy on trying to understand how. acute physical risks, you know, hurricanes and flooding and fires and everything else affect the economy. So we've gotten a lot much, we've gotten a lot better at it since since Katrina in 2005. Okay, Chris, what about you? What, what forecast are you, well,

you go whichever one you want to do first. You're most proud of, at least, at least, comfortable with. Sure. So most proud of, and I'll, I'll focus on my time at Moody's. The one I, the set of forecasts. As opposed to Fannie Mae, you mean? You were, yeah.

Why did you screw up at Fannie Mae? Is it, you didn't cause a financial crisis? Oh, wow. Did he just say he caused a financial crisis? Oh, that, no, okay. Didn't see that coming.

I've got some proud moments there too. I bet you do. Yep. This was one of the first projects I worked on when I, when I joined the team here as well. This was a forecast around HAMP and HARP. Do you remember those programs after the last recession? This is the Housing Affordable Modification Program and Refinance Program.

And we had a project to forecast the take-up. Yeah, so I just found it to be a really interesting problem, right? It's Greenfield. There's no data you can go back historically and look at. And I just enjoyed the process of piecing together the puzzle. And looking back, actually, the forecast were quite accurate. Because HAMP really did not work that well. Because that was kind of you took a borrower.

You ran them through a waterfall. You saw whether on a present value basis made sense to forgive debt. Wow, you were paying attention. Oh, I remember that program really well. Yeah. I remember. I was excited about it because on paper, you go, this should be pretty good. But actually, in practice, it was kind of sort of like the,

rental assistance program that has been in place for the pandemic. It hasn't really worked out that well, at least so far. Yeah. And what advantage is, I think it just bought us time. Bought us time, yeah. And what forecast are you least proud of? All of my stock market predictions over the years. I'm back to that. That's unfair to you. It's unfair to you. What about your crypto forecast? I'm very proud. Very problem. Still going to zero. Okay. Very good. And Marissa, what about you? What are you proud of? Well, mine is similar to Ryan. So it was a forecast I made for the New York City economy right after like Lehman Brothers filed bankruptcy. So I can remember that, you know, we were, this was 2008. We were already in the recession. There was already a housing market correction. We were just starting to understand why, sort of.

And then all these investment banks started getting into trouble. And it was September. Lehman Brothers filed bankruptcy. It was a Monday morning. I remember coming into the office. My phone was blowing up with reporters asking, what's going to happen to New York City? You know, all these thousands and thousands of people are going to get laid off.

So I looked at the forecast that I made the following month. So the October forecast. And over the next five years, I got pretty close to the decline, the initial decline in jobs and the timing. And five years out, I was back to the actual employment figure that actually ended up, you know, unfolding. And I remember, just having every day just adding up these layoff announcements from all these investment banks and financials and just trying to figure out how many are in New York City and what's the odds that these people find jobs and other sectors. So that's, yeah, that's the best forecast. That's a good one. That's a good one. Certainly the financial crisis was a seminal event for a lot of us, you know, as forecasters. What are you least proud of? What forecast, The forecast I made two months later of the same thing was worse. Oh, that is, that's funny.

And I think it's because kind of like Ryan's, I actually, so Ryan was anticipating a big rebound after Hurricane Katrina in New Orleans. I was not anticipating a strong rebound in New York City, right? It was like all this talk about banking and the financial segment just shrinking permanently and firms leaving New York, which always seems to happen after any event happens in New York City. There's always this narrative about how everything's going to leave New York and it never happens. So I had a pretty slow long-term recovery and it ended up being much faster in reality. Part of that, too, was that at that time, and this I think goes back to our,

error in part of not knowing what would happen with the pandemic is not anticipating the extraordinary fiscal intervention there would be to help the banking sector and the financial sector back then, you know, all the two big TARP and all of that. So that came later. And I think that ended up saving a lot more jobs in the financial sector in New York than then otherwise would have happened. Yeah, those are good ones. You want to know which one I'm most proud of, the forecast I'm most proud of and least proud of you guys care?

No, I want to know. Absolutely. Okay. So I'm actually most proud of our house price forecasts back prior to the financial crisis in 2005 and really into 06. And we sold economy.com. Who, so Marissa, you were with me when economists. economy.com. Chris was not. Ryan, were you with me too at economy? I was. So the two,

we were all economy. We sold our company to Moody's in, when was that? November of 05, right? November of 2005. And we were in the middle of writing this house. We had done all this, but we, you know, we had done all this work with Kay Schiller. The Kay Schiller was like the new repeat sales house price index, you know, and it was captured the entire market. market and based on actual transactions, we helped finance their development as a company. We modeled it and then we forecast it. And we were in the study that we put together in 06, really kind of, I think it was around the spring 06, we were calling for house prices nationwide to decline. And we actually did HPI forecasts across lots of markets, you know, adding them up across

all these metro areas and the national house price to, we expected national house prices to decline. And I remember we had just joined Moody's. And this study, of course, no one, Moody's knew us, right? Who are these guys? Who's Zandi? What the hell is he talking about? National house prices have never declined. This is just, this is bogus.

It's not good. This is, you know, I heard all kinds of things. Because the forecasting they were doing was based on these, you know, these kind of Monte Carlo, you know, these are, they're not based on fundamental factors in the housing market. They're based on statistical relationships historically. And you pump out these distributions of forecasts. And basically the middle of the distribution is kind of the average house price forecast over history. And it bears no relationship to the current level of house prices, which if you're in a bubble, that's a big problem, right? because prices are significantly overvalued.

And all you're saying is, oh, they're going to continue to rise at, you know, three, four, or five percent per random. So we come out with this. And by the way, if you have national house price declines, that's a big problem for mortgage-backed securities that are, you know, being, you know, it was the RMBS residential mortgage-backed securities being issued at the time, you know, it was just out of control. There was, you know, hundreds of billions of dollars of securities being issued and rated,

of course. And if you had house price declines, you didn't take a genius to, you know, it was just, to connect the dots and say, oh, there's going to be a problem here, you know, with these securities and the ratings behind the securities. And, of course, there was all kinds of hand-wringing and, you know, debate, got up to the CEO, great guy, Ray McDaniel who has since retired. And, you know, he goes, well, and it was a reasonable question. He goes, he goes, Mark, why is an economist so interested in house prices?

you know, kind of when you think about it. Now everyone says, well, that's a crazy question. Back then, you're going, well, okay, well, what does it have to do with GDP and jobs and all that kind of stuff? You know, and he was not an economist, right? You know, that wasn't what he did. So he just fundamentally didn't quite get why I was paying attention to this. Fortunately, Ben Bernanke, chair of the Fed, wrote a speech at that point in time. The speech's title was subprime mortgage. And by the way, talk about a forecast error. In that speech, he said, don't worry about housing and mortgages. This is not going to be that big a deal. And if it is a problem, don't worry, we'll, you know, we'll solve, you know, we'll step in and solve the problem. Anyway, that was the best thing that ever happened

because once Ray saw that, he goes, oh, okay, now I get it. So publish your paper. And we publish the paper. And it was, you know, created all kinds of stir because we were saying national house prices were going to decline. And I remember it really worked out quite well because when the Financial Inquiry Commission, that's the commission that was established by Congress to investigate why, what happened during the financial crisis, Ray was asked to testify. I testified previously. He testified with Warren Buffett, who was a major shareholder in Moody's, and he brought up to study. He said, look, look, you know, we said that this is going to be a problem. And there was a study.

So I was very proud of that study. But here's the thing I was most disappointed in in terms of forecast, is our forecast was for only a 3% decline in national house prices. Not a 30% decline in national house prices. And this is a good lesson, fantastic lesson, right? because, you know, some of the markets that we were modeling, you know, we had a price, I remember Miami, house price declines. Our model said prices are going to decline 90%.

Think about that for three seconds, one second, 90%. That's what the model said. And I go, I remember talking to Celia, one of our housing analysts who's still with us. I go, Celia, if we come out with that, no one's, everyone's going to think we're nuts. We're crazy. they're not going to pay any attention to anything we say. So we adjusted the model result and said down 30 for Miami. And that got us to down three for the nation. Guess what Miami house prices declined during peak to trough in the financial crisis? 90%. 90%. So what's the what is what is what do you

take away from that? Just your model. Damn right. You know, you have to have the courage of your conviction in your models and use them. And by the, the way, this gets back, Ryan, I'm coming right back at you with the housing starts forecast. So here we are on the other side of the financial crisis, housing starts are, you know, 600K, I don't know, very low. It clearly didn't make any sense relative to underlying demand, which was improving. They can see rates come crashing in, you know, and the model says, hey, we need more housing starts. So, you know, I learned my lesson from the financial crisis. I said, we're not changing this forecast. Yeah, I understand there are things that are not in the model that we, you can't

really measure or forecast like, you know, supply constraints and zoning issues and that kind of thing, permitting costs. But in, but I, you know, I stuck with the, the housing starts forecast. And it was obviously too strong, way too strong. But here we are in, you know, 2021. And we've got an affordable housing crisis, right? The model said, we should have been building. You know, we had the demand to build, you know, but we didn't build for, you know, reasons that we didn't capture in the model. And here we are with an affordable housing crisis.

You know, we have a shortage of affordable homes, you know, to the tune of 1.5, 6 million units, about a year's worth of supply. But anyway, good lesson. Do you think people out there listening to this podcast are going to care about these stories? Do they think that you find that interesting? We find it interesting. Do you think they'll find it interesting? Absolutely. Absolutely. Okay. All right. We wax on. Hey, I have a bunch of other questions, though. Let me ask you this question. Key question. You know, people in business and government, they, you know, they value forecasts. They have to have forecasts. They have, you know, everyone forecasts for budgets for budgets. For budgets. for planning, for risk management, that kind of thing, for loan loss provisioning.

But they really value forecasts, I think, around recessions, right? And, you know, that's when things really become an issue. The economy really matters. But economists are not in, I'd say we're in this camp, too, good at predicting recessions. We just have a hard time doing it. And we don't have the, here we don't have the courage of our convictions and say, hey, recession is coming, you know, a year from now or two years from now. Why do you suppose that is?

Why can't we forecast, we collective we economists, forecast recessions very well. Well, you actually have been saying, you did. You didn't get the cause right, but you did say there could be a recession in, well, you were saying June 2020, right? Yeah. Thank you. Thank you, Marissa. Thank you for pointing that out.

Thank you for pointing out. Chris and I believe that prediction deserves an asterisk, just like Barry Bonds is home run total, big asterisk next to it. Why? What would that, what, okay, we wouldn't have had a recession if it wasn't for a pandemic. I don't know. I don't know. Yield curve did invert, right? So, real curve invert.

Okay, but back to my question. With that aside, Marcia, you're right, I did. And of course, it was a half ingest, right? Because, you know, how could I predict a recession? you know, in June, on June 20th, 2020. I was trying to make a point. The point was, look, there's a boatload of risk out there, and you should, you know, you should behave accordingly, act accordingly, be cautious.

I don't know exactly how we'd end up in a recession on June 20th, 2020, but hey, guys, the preconditions for this are coming into place. But it wasn't half ingest. So why can't economists get it right? What do you think is going on? I've got a theory, but I've been waxing on a lot here. So does anyone else have a theory as to what's going on? There's randomness in the world, right?

We can't predict sporting events either, right? Plenty of data. So there are random shocks and animal spirits, all the things that we've identified in the past. So the specific event occurring is very difficult to predict. But I think we have gotten better, certainly, at predicting conditions, right? to your point, we can look at different indicators and identify the preconditions for a recession, whether or not it actually occurs, right? We can at least send out the warning signals.

And I think that is, I think that is helpful for the market to know. Yeah, that makes sense. I mean, we couldn't have predicted the pandemic, right? So that's a shock. I mean, in that case, I don't think it was, it didn't matter whether the preconditions for recession were in place or not. we would have had a recession because it was such a cataclysmic event. But, you know, 9-11, I mean, if you go back to that recession in 2001, I don't think the Dating Cycle Committee of the National Bureau of Economic Research, the arbiters of recession, would have said that was a recession without 9-11, and that certainly could predict that. Yeah, so you're saying randomness, you're saying what an economist might call an exogenous shock,

something outside the system. some others would call it a black swan, I guess, event, you know, something out there on the tail of the distribution of possible outcomes. It kind of sort of know it's a, you know, we knew that there's pandemics out there, but there's no way to know that you're going to have a pandemic in March of 2020, right? You just that you can't predict. You just can't do that. So I think that's fair.

Any other reasons why we can't or we have a hard time as economists predicting recession? Any other theories? What about, you know, my view, my thought is recessions ultimately at the end of the of the day are a loss of faith, a loss of faith in the economy in your financial well-being. And that's very much an emotional event, right? It's not generally a numbers and sense event. You know, the numbers and sense aren't working all that great. You know that. But, you know, it's when everyone kind of collectively gets to a place where they say, oh, my God, you know, this is going to be really bad. And they all run for the doors or maybe the better analogy,

they run for the bunker at the same time. We all run into the bunker. Stop spending, stop investing, stop hiring, you know, stop doing what we do, you know, when times are, you know, normal, typical. And, you know, we go into recession. And that's why I think, you know, one of the really good, I don't know, it's foolproof, but pretty close, indicated near immediate indicators of recession is when consumer confidence measures, you know, the surveys of consumer confidence,

like the Conference Board survey, the University, goes south in a big way, right? They fall big time definitively, not for one month, but for two, three months, and then you know we're toast, you know, we're in recession because we had a loss of faith. So I think that, you know,

there's animal spirits, right? You know, in that case, the kind of the dark spirits, you know, take over and we go back in recession, we go into recession. So that is inherently, how do you predict that, right? I mean, how do you predict that kind of

emotional, collective emotional response that, you know, happens when, when you go in recessions. Any others? Any other reasons why we might miss it? Or about policy errors? Do you think policyers? Here, I'll give you, I, here's my theory about the financial crisis, a theory. It is probably overstated, but just for didactic purposes. You know, the things were going badly. Housing was overvalued. We had a bubble. It was bursting. Prices were declining. I was doing damage to consumers. They were getting nervous, trying to create problems for the banking system. But maybe we wouldn't have – maybe we'd probably gone into recession,

but maybe we wouldn't have had a cataclysmic financial crisis, if not for a series of policy mistakes. And what I mean by that is as financial institution after financial tuition got into trouble, policymakers, you know, the Treasury Paulson under Bush started with bear Stearns when it failed early on. Each of the institutions that got into trouble was treated differently. You know, the creditors of the institution, the debt holders, the equity holders, the investors in the short, the commercial paper or the short-term funding that was being used.

And at some point, and I think the point was really with Lehman Brothers, when they took over Lehman Brothers and they said, we're not all the creditors of Lehman Brothers are going to lose, all the equity holders, all the debt, we're not bailing out anybody. That's when everyone ran for the door, and that's when we had a financial system collapsed and needed to bail out, and we had a financial crisis. I would view that as a, you know,

I'm not blaming anyone. I mean, I'm not saying anyone could have done any better, you know, under these trying circumstances, but that feels like a policy error to me. You know, we didn't treat these institutions. And in fact, you know, in the Dodd-Frank legislation after the pen, after the financial crisis, there was an effort to put together

a kind of a cookbook of, you know, here's how you, you know, you work with troubled institutions so that we don't have the same kind of policy mistake in the future that we had during the crisis. I think that's another example. I will say the inconsistency of the treatment, right? That's the issue. Consistency of the treatment, the consistency of treatment. If it was well known ahead of time that that would be the outcome, then investors would have

acted differently. We may not have had that outcome. By the way, before the pandemic hit, I was writing a book, the next recession. I really was. I had written three chapters to go into some of these subject matters, but I kind of put it on hold. Got to figure out how to resurrect it now in the wake of the pandemic. I know the recession. Yeah.

Was the next one? Yeah. Oh, well, I just got that question this morning. And the answer I gave was, I don't know. Good answer. Good answer. No, no. Well, none of the preconditions are in place.

You know, yield curve is fine. You know, there's plenty of room to run here. You know, abstracting from the pandemic going down a very dark path here, you know, it's hard to argue that or think that we're going to have a, and everyone's on high alert policy is, you know, fiscal and monetary policies, because they've got their foot flat on the accelerator. So it's premature to call the next recession. But I'll call, ask me a year from now.

That might be very different kind of forecast. Okay. Here's another question. I know we're already getting a little longer in a tooth here. This is actually, for me, very fascinating podcast. I hope it is for everybody else. Do you guys have any kind of forecast rules that you think are important to good forecasting? You know, kind of things that you hold on to when you are doing a forecast that, you know, a discipline that you use.

to make sure that you don't make mistakes that you have in the past. Do you have any rules that you've adopted? I'll start. One rule I have is keep it simple. The complex models are very attractive, right? They got all the bells and whistles, but they're also very, tend to be fragile. So if I'm going to use a model, well, first of all, I won't use just one model. I mean, use multiple models, but within each model, I have a, you know, I have a,

bias towards a very simple model or more parsimonious structure. Yeah, I think that makes a lot of sense. Kind of sort of like a corollary to Occam's razor, sort of. Yeah. You know, the most straight, forward explanation is probably the right explanation. Don't overcomplicate things. Okay, that makes sense. What about, do you rely on your models too?

I mean, do you think people get into, into, start making forecast errors if they don't rely on the models and the tools that they have, or not? I think so. Yeah, I think so. I think the model provides some discipline. And again, it may be your model, maybe multiple models. Ideally, right?

My second rule is three is better than one. Yeah. Looking at a range of models and a range of scenarios, it's going to help you to bound the forecast better. I think you absolutely want to leave some room for jobs. judgment, but the models instill that type of discipline that your story illustrates, right? That you're able to look at things a little bit more scientifically, understanding there's lots of variability, but at least the model helps to organize the information in a logical

format that your judgment or your assumptions on their own may not be able to. Yep. Good. Any, Marissa, any rules that you use when you're doing your forecasting or guiding others when they do their forecasting? Yeah, I agree with what Chris said. I think the default is always to trust the model. And unless you have a really compelling reason to believe that the model is wrong, and that you should override it, then you stick to the model. I think we've seen we were talking about this the other day. I think I've seen that a lot of times the most accurate forecasts are the ones that don't involve any human intervention. are very little. A lot of times when people really start doing a whole bunch of overlays and judgment and looking at the news every day and changing the forecast around, they tend to be

worse than if they just stuck with what the model told them from the beginning. So I kind of think less is more as well, which I think is kind of what Chris was saying. And then the other thing I would say is just when you're trying to forecast a macroeconomic variable, whatever it is, you have to really understand the data. You have to really understand what it is that you're forecasting. What is the source? How is it put together? How is it compiled? What are the downfalls of it? You know, what are its strengths and weaknesses in terms of what it's measuring? Like, I mean, Ryan is the best at this, right? Like, he knows the data inside and out where he's saying, oh, I don't use UI claims every July because of the auto manufacturing retooling. I mean, that takes a lot of experience and

time to understand the data to that extent that you can understand how these individual industries would conspire to move an employment number. I mean, that's getting pretty deep into it. And I think the deeper you can get into it, typically the better you are at forecasting something. Absolutely. Makes total sense. Ryan, any rules that you use? No, I agree with everything Marissa and Chris are saying. I think my rules. rules are a little bit different because we have a significantly different forecast horizon than Marissa, Chris, and you. I'm looking out one month, one week.

So for me, more is better. So I try to get as much information, high frequency information as I can to put into these models. And we quickly had to scramble when the pandemic hit, all the models I had just, they weren't going to work because we've never seen anything like this. We had to turn to alternative data, you know, quickly started using Google. Google mobility, home-based data to forecast some of these high-frequency indicators. So, yeah, I mean, more information is better. Yeah, I got it.

That makes a lot of sense. One of the thing I think is important is a consistency. You know, our clients, and when people use forecasts more broadly, they don't want to be whipsawed. They don't want, oh, it's great this month. It's not so great next month. It's great this month. It's not so. Well, what is it exactly?

So a lot about a lot of the forecast depends on the underlying assumptions that you make, you know, the exogenous inputs, the things that are outside the model that affect the model results in the forecast. So monetary policy, a fiscal policy, these are, you know, inherently assumptions we're making. And we have, I think a very important rule is we have to have a very high, are in our thinking around the assumption before we change them. And I have into my mind this two-third probability rule that I will not change a major assumption in the forecast, you know, the Fed changing monetary when it's going to raise interest rates or this $3.5 trillion fiscal package that Congress is debating. I will not change assumptions around those things until I feel

like that there's a two-thirds probability that we're going to get a different set of assumptions, you know, going forward. So that way we have consistency and stability in our forecast, and we don't typically get whipsawed. And there might be cases, I'm sure there are, but generally we don't. And I think we serve, I think it leads some more accurate forecasts, and I think it also leads to, you know, more useful forecasts for the people who are using them for whatever they're using them for. That at least is another quick question, and that is forecast accuracy.

How do you think about that? Do you think, first of all, you know, it's kind of an epistemological question. What is forecast accuracy? And how important in your mind is that? I know that's a tough question, but I ask tough questions. Anyone got a view on that?

Mercer, do you have a thought around forecast accuracy? Well, I mean, if you're talking about, did you hit the number right on the nose, that's almost never going to happen. Well, Ryan, it happens for Ryan a lot, I guess. But when you're forecasting out a year, you know, a year and a half, three years, you know, the forecast is not going to be right on the nose. So I think it's more about the narrative that you tell. Well, what is, you know, tell me the story about what's going to happen and how this is going to play out and how severe it's going to possibly be. What are the probabilities that it's X versus Y? I think it's, at least with our clients, I think that they're valuing more that the framework in which we're placing the forecast rather than holding us to hitting a specific.

number because most people understand that that's not going to happen. I think that's a great point. Chris, anything you want to add on that? Yeah, just to, I would say it really depends on the application. And I'm thinking of Cecil as one example, right? The accounting rule, the new accounting rule you mentioned. If you read the documentation, the guidelines, it doesn't say anything about accuracy. It's a, it talks about reasonable and supportable, right? The forecast has to be reasonable and support. So that's, that's really the emphasis. You're coming up, there's an acknowledgement that you're not going to hit the number, as Marissa said. And it's really about coming up with an approach that, you know, meets

the smell test is something that is a reasonable expectation of what could happen, understanding there's a lot of uncertainty around that. So I think that's what we actually are aiming for in our forecasting process. Got it. You know, I think forecast accuracy is important and I don't mean to sound slippery here because I do think, you know, it is important that if you're not, if you're consistently not accurate, then you're doing something wrong, right? I mean, the way you're thinking about things, the framework, the model, the data, whatever, you know, you got it wrong. So, you know, accuracy is important. But actually measuring accuracy, that's a pretty slippery concept.

I mean, you know, for many years, I did a lot of regional economic forecasting. That's where this company started. It was really in forecasting what was happening to regional economies. You know, that was before economy.com, we were regional financial associates. So, you know, we were regional. And employment is key. employment data, the employment data that, Mercia, you put together for the world, we relied on and consumed in great detail. But that employment data is subject to mass revision, you know,

particularly when you're talking about a metropolitan area or a state, because it's based on a sample of establishments or households. And it's just a sample. And of course, it's, you know, not the universe. And once a year the BLS goes out and collects the universe based on unemployment insurance records, so-called benchmarks, it's survey-based information to that universal count. And you can get some pretty massive revisions, particularly when the economy's going up and down and all around and get very different perspective. So if you're saying, well, are you, was your employment forecast accurate? Well, accurate in measuring what exactly? The initial estimate that the BLS put out or the second estimate or the third estimate,

or the third estimate or the final estimate. By the way, the final estimate feels like it's never final. It's like five years after the fact it's final kind of thing. So, you know, what's important for the client is, is your forecast giving them a sense of where this economy is headed in the strength of the economy or the weakness of the economy or what is the weakness of the economy? What's going wrong?

And also just telling clients about the data itself, what data should you be using or not be using? And if you're using this data, you know, what should you be thinking about when you're using it in your own revenue forecasting or budgeting or policy analysis or whatever it is? So forecast accuracy is, you know, it's easy to say, but it's a very slippery concept. And what's important is if you're giving people who use those forecasts, you know, are fundamentally a sense of, you know, what's driving it, what's the framework? Is it, as Chris said, reasonable and supportable? And at the end of the day, is it giving you

directionally, you know, the reality of what's going on here. This is a good economy. This is a bad economy. I think that's what's most important. Okay. I know we're running on, but I don't care. We're going to keep going unless you guys have, Marissa doesn't have happy hour for a while. She's going to go surfing pretty soon. So we're going to keep going. And you guys, listeners, if you don't like this, you can tune out. You can tune out. I'm having, I think this has been, this is very informative, but I got a little bit more to talk about here. So I want to know, each of you, whether you are a hedgehog or a fox or something in between. Now, a hedgehog is someone who's got embedded in their minds, a framework for how the economy

works. It's immutable and everything has to fit in that and all my forecasts revolve around that view of the world. Or a fox. Fox is, I'm taking in all this data, all this information. I'm going to change my mind, you know, based on recent data points, you know, kind of bringing it all together, use my intuition, maybe use some models and come up with the forecast. But I don't have this kind of broader, you know, kind of fundamental framework for how the economy works. It's, you know,

it's more based on, you know, what's happening in the recent past. So what are you guys? Or you can come up with a different animal if you want, but, you know, and maybe you have. But, you know, Marissa, are you a hedgehog or a fox or something in between? So you did, so the listeners know, you did tell us you were going to ask this question. I did. An online quiz to find out what I am. And I am three quarters head talk, one quarter fox. Oh, I love that.

I love that. You did an online. Marissa just went second level. Hold on it. What does that mean exactly? You did a survey of whom? I don't know. Just I googled.

I Googled it and there was something, some, you know, business insider website had like a investment that you could take to tell you whether you're a fox or a hedgehog. Oh, I see. There's this, oh, the inside, business insider has a actual tool, survey. You can go on and take it. I think it was business insider. Yeah. Oh, that is really interesting.

That is, I'm going to have to take you. Sir, your three quarters hedgehog. I would have said, I would have forecasted that, by the way. Yeah. As a joke, you guys should be laughing. Where's the soundtrack? That was like a that was a Colbert quality joke right there. We'll ask Ben to slice that in. Yeah, thank you. A little drum roll. A little drum roll. All right. What's Chris? What's your forecast for Chris? Chris is, I'd say he's 50-50. Like most things in life, he's 50-50. Yeah, he's 50-50. He's a 50% hedgehog, 50%. I think, I think that's, what would you say, Ryan? What would you say he is? Maybe a little more

hedgehog. Yeah, I was going to say more hedgehog. Two-thirds, one-third. Yeah. Yeah. Yeah. Chris, what do you? What do you say? I think I'm a mix, but I would actually go with more with the Fox. A little bit more Fox and Hedgehog. Did you see the way he said that? I'm the Fox. That's pretty funny. Really? Okay. So tell me the percentages. And, you know, 6040. Okay, 6040. It's pretty good. And why do you say that?

Well, so as you, you know, as you said, the hedgehog thing is really dedicated to one idea and one thought really focused on it versus the Fox has many ideas. And I think I tend to to look at a problem or look at a forecast from a lot of different angles versus sticking to one worldview. Got it. Got it. I thought you were, okay, I had a joke, but I'm not going to say it because it wasn't Colbert quality. Oh, yeah. Ryan, oh, Ryan, God damn. Oh, sorry, I shouldn't have said that.

Ryan, let me guess. Ryan is, he masquerades as 30% hedgehog, 70% Fox. In reality, deep down when he's, you know, in bed thinking about things, he's 80% hedgehog, 20% Fox. but, you know, he, he masquerades this Fox. Do I have that right, Ryan? No. Yeah, I'm like 90-10 Fox. 90-10 Fox, yeah.

That makes sense. Yep. Okay. I believe that. That makes sense. Because, you know, this is an important point, I think. It really depends on your horizon. Yeah.

Right? Right? I mean, if your horizon is your term, you know, next month, next quarter, I think Fox makes a lot of sense, right? it doesn't really matter what the structure of the economy is. No, no, not at all. If you're forecasting out three to five years, I'd say probably better to be a hedgehog.

Correct. More of a hedgehog than a fox. And if you're thinking out, you know, 25, 30 years, then you better not be either one of those. Definitely not a fox. And I'm not sure a hedgehog's any good either. You know, we've got to come up with a different animal for that kind of forecasting.

Because at that point, technology changes. and, you know, the society changes and all kinds of things change. Very, very difficult to get now that. And by the way, that leads a very interesting point. In my view, I'm curious if you have a different view, I think our accuracy is highest in a horizon that's probably around three to five years, right? Maybe three, five, seven. Because, you know, near-term, you get asset bubbles and policy changes and, you know,

events that's going to scramble things. longer than seven years than you have technological changes and big political upheavals. But three, five, seven, that's kind of the sweet spot, I think, right? Because, you know, you don't have the near-term noise, you know, this long-term structural shift. And that's where our models, I think, work best. Would you agree with that, roughly speaking? I wouldn't. And don't try to change this subject yet.

Yeah, yeah. We've got to guess what you are. Well, I answered the question. I am a, there's a word for, there was this being that evolved over time, you know, changes form depending on the horizon. So if I have to do a near-term forecast, I'm definitely going to be more of a Fox. If I'm going to do a long-term forecast, I'm more of a hedgehog. But if you nail me down and say, Mark, on average, through time, you know, what is the percentage? I'd say I'm more hedgehog than Fox.

I'd say I'm 80, 85% hedgehog. I really am because I believe fundamentally it takes an awful lot to change the structure of this economy. It has not fundamentally changed in the period since World War II, the way it operates, you know, the principles under which it operates the, you know, the mechanisms that drive things, the way we conduct policy and think about the world. I think, you know, it would take an awful lot for me to change, you know, the way I think about things. Is that how, does that how you perceive me? Yeah. Definitely. Yeah. Yeah. Yeah. You have got these strong rules of thumb that you follow as well, right? Tenures going to. I would have sprinkled a little

sloth in there. Right. Because very slow to incorporate changes to the tenure forecast. That's true. Which I could have said was my most disappointing forecast, the 10 year, you know, coming after the financial crisis, but we won't go there. Hey, I do want to end. We got to end this thing. I could go on. This was a fantastic conversation, and we have a lot more to talk about. But you know, you know what really annoys me? It annoys me when people say, I don't forecast. Everybody forecast. There's a great, there's a great word for that. What is it? Bullshit. That's the word. Yeah. Yeah. No, oh, you know, I got a better word for it. Flapdoodle. Do you know what flapdoodle means? It's a great word, by the way. Flapdoodle. Good SAT keyword. Nonsense. Nonsense. That is nonsense. All we do all day long as human beings is forecast. And in fact, I go so far as to say, knowledge is the ability to forecast well. By the way,

I picked the word well, you know, carefully. It's not accurately. It's well. It's forecast well. Because, you know, we all know the sun comes up every day. Why do we know that? Because we got a lot of data points to say it. We've got a theory to suggest why it's going to happen. But we're forecasting that someday it's not going to come up, I assure you. You know, it could be three billion years from now, but it's not coming up. But it's a good forecast. It's a good forecast. And so I think forecasting is a, I'll go so far as to say this. And maybe, you know, this is definitely self-serving, but it makes me just feel good. Forecasting is a noble profession because it takes a lot of, you know, a lot of hard work to do it and a lot of failure to do it. It's very humbling. But at the end of the day, it is the most important thing

that we do, you know, we do economic forecasting. Others do all kinds of other, there's all kinds of forecasting, but that's, that is what all of us are doing. So when someone gets on and says, hey, I don't do a forecast, I, you know, that I find, you know, just, I guess the word's annoying. It's just annoying. That's not true. Not the case.

We all forecast and we all are working hard to make better forecasts. So with that bit of a rant, that was a rant. Thank you. This is a great podcast. And we want to hear from you. Want to know what topics you would like us to tackle next. We've got a poll. You can go to economy.com. See a little button there for Inside Economics and go there and let us know what you want us to talk about next. Can't wait to next week. Talk to you soon. Take care.