本字幕由TME AI技术生成
Npr robbis smith
Uh
Why the long face i read something on the internet
That i disagree with give you haven't gotten used to this sped up
I know i know i know
But this was someone bad mouthing
The page book of all things
No
The paebook is maybe a favorite of schoill government report
It comes from the fitter reserve containing short eectes without the economy
When you know who doesn't agree with us
Journalists mat aggacious
No usually has these hot takes i enjoy
But this time on twitter
His hot take was about
Well i'm just going to have matt read it for us
What is the point to the beijbook
We supplement our statistically valid survey with another survey that's worse
Why you know here at the inicicator we love the beijbook
Because you know it tells little beaubeautistories about the economy
Uh
So why do you hate beauty
I just think it's
It's funny
Because the bashbook is supposed to be an alternative to the kind of blooddless data
Uh
But the data itself is just compiled of surveys
And so you don't want to hear like how a car dealerships doing in minnesota or
Or some hotel in keep cuard
I mean
I am in ananalyst
And i understand the value of human stories that can illustrate trends up
But it's hard to know how to
How to make anything out of these kind of randoomanic tives
Oh
Hh will make something out of the myriglacious watch us
Um
It's the basiawards are eight times a year salute to the art and science of telling stories about the economy
I'm roabtt smooh
And i'm teing woooods 're 're going to do something specispecial today just to prove matt wrwewe ll be giving a special technical achievement
Watch the economists and data scientists who have taking those so called led andomomacdotes in the beiebook
And using them appear into the future data magic
After the break support for this podcast and the following message come from wise
The app that makes managing your money in different currencies
Easy with wise you can send and spend money internationally at the middmarket exchange rate
No gueas work and no hidden fees
Learn more about how wise could work for you at wise that com it's a high stakes election year
So it's not enough to just follow along
You need to understand what's happening
So you are fully informed to come november
Every weeday on the empie politics podcast
Our political reporters breakdown important stories and and back stories from the campaign trail
So you understand why it matters to you listening to empmpe politics poadcast
Whererever you get your podcast in this country
Some truths aren't self evident in npr black stories
Black truths
A collection of stories as wide ranging and real as the people
Who tell them we celebrate the black experience for all its soul enrichness search npr black stories
Black truths
Wherever you get podcasts so usually on the basiia awards
We read all the anecctes in the beijbook
Published by the twelve regional banks in the photo reserve
And we given award to the writer
That has the most exciting story
But today we are doing something different
We 're presenting eecds to some economists to a doing research on the beijbook
Itself doing those exciting ecctes into data code had numbers
Or i'm so excited
Here is the envoludarian
I'll let you the the honours
The basiiaward for technical achievement in beijbooky goes to the clievan fed accepting the award for his team from eclivvin fed economist
James mitchell congratulations jodents
Thank you very much
I'd like to thank another
Another always your moderl listening
I heard so james 's his mother by the way lives in gilfford ari outside of london
Where james grew up
James in his team are winning the beijy for a joint research project of the cleveland fed and washington university and sanet lewis
They used machine learning
A kind of ai to go back and read all of the baijbooks
Every single one respect and tell if i get this right james
Every classifies each sentence as positive about the economy or neutral or negative about economy
That's exactly
Exactly rise
So we look at each paagebook from nineteen sevenseventhrough through to the the latest facebook at the time of speaking
And we look at those facebook sentence by sentences
And then for each paagebook publication
We taught on the number of positive sentences
And attract from that the number of negative sentences
What if the basebook look like in the nineteen seventies
I mean was it like you know a sales of bll bottooms are down uh financial conditions are
Are grooy y all right men
And for sure
The basebook has changed
Over time indeed it was the red book
It had a red cover through the ninety seventies
Until ninenineteen
Eighty three
But over time the basebook ook become come structured and indedeed a little
It llonger so a positive sentence in the basebook might read housing markets continued to strengthen
And a negative sentence might be spending on goods remains sluggish
So once they attend sententces of the basebook to numbers
Positive
Neutral
Negative
They could look at how this corresponded to something we all care about very much
The business cycle
The upsent downs of the economy
The team found that
When the basebook starts to use more negative sentences
That correlates with the downturn in the economy
Autumline is that
We find that
That these
These machine learning methods appear to work
Work Can
Can
The basebook preicicted recession
It can be helpefully
So it's a great
But we believe additional source of information that can be helpful in anticipating in particularly in now casting
Whether the us economy is in an expansiary or a recessions
Ary fades of the business cycle helpful and additional source of information
And
And he doesn't seem to be saying it's a crystal ball
Yeah
These are all correlations in the past
But no economist is going to say definitively that this is going to happen
I read the paper dyian
And there is one more interesting thing
Some regions of the country are slightly more helpful
And telling us when the economy will turn those regions
Pildolphia
Richmd
Severt
Csco
Chicago and miniopalas go to in buoves
Yeah
Why those cities well ai doesn't really explain
Itself it just finds correlations
You know when we do these awards
We always like to see how the beijbook plays out in the real world
And some investors are already using this kind of ai analysis to beit on the future for instance
Right now the economic world has one big question
When with the federal reserve lower its interest rates
And i found someone who's using this ai analysis to help investors with that question
My name is irojersey
I'm the chief
Us straight strategists for glumbger intelligence
Which is the research arm of blumber gl p
Iraa has developed a computer model
That can get usable information out of the bichbook
He's already been doing this for a couple of years on transcopes of the federal resves press conferences and minutes of their meetings
He analyzes the sentiments of the words
And then compare them with actions set the federal reserve
Then takes i asked
Ira can't i just listen to the press converters and get the zim information
And he said you can
You should
But you're probably missing something so for example
Um a lot of times
The media might focus in on one phrase
That's different or new or sounds
Exciting instead of looking at the totality of what's actually said
And that's where the machine is a little bit better than we are in terms of determining
That because computers don't have the same kind of emotions
I think that's a big part of it
Um
I think suspect that there's um
There's your own personal biases right
If expect a positive speech from jappoal
Maybe he sounds negative
When i hear it
Even when he's not iira has just started to put in the beijbook data to his model
And it is a little a more complicated
He says He
He decided to just use the summaries of the basebook
Not the ecdotes that we all love so much in the past
Some of the regional districts had much more verbose reports
Yeah
You're told me that much what have to read through
But they could get long
Yeah Ulo
Yeah Well
And it's not as long
It used
Used be because now everyone kind of sticks to the same format more or less
Ira in his bloomberer team have already found moments
Where a computer can see something in the basebook
That the rest of us missed for incence in twenty
Twenty one
When the fed was talking about how the inflation was transitory and about to cool off and analysis of the basebook show that the economy was quite strong an inflation rising
So what's the basebook telling gus right now i ran last month spoook through the model
And it's no longer showing a knowing economy
The fed stoarts
He says is neutral or beige
If you will but stay tuned when something changes
Maybe the computer will spot it first and experra
Baji
Wooter
James mitchell from the clevveland fed congratulations again for your award
For your technical achievement
I do worry a little bit about your research
About that perhaps
If they are enough numbers association with the basebook
Who
Who would actually sit down in read it covered to over
Where it's great to be here to accept the war
But i think that's the text and data should always exist together
So perhaps
Our numbers but are helpful compliment to that great rich text but is the benchbook take that hatters
Who still got a job
Rob it this episope was produced by corry bridges
With engineering from code tucha a uki
Chenivan was feccacked by sarahoadis cake and cadam edits
The show and the indicator is a production of npr
All that sitting and swiping your body is adapting to your technology learn how and what you can do about it
I really felt like the cloud in my brain kind of dissipated
Once i started realizing what a different seasonlittle breaks were making there's no turning back from me take np s body electric challenge
Listen to the series
Wherever you get your podcasts
What does it meto be black in america
Nnpr black stories
Black truths
A collection of stories as varied
Nuanced and dynamic as the black experience
You'll hear it means everything search np r black stories
Black truth
Wherever you get your podcasts
What does it mean to be black in america
An npr s black stories
Black truth
A collection of story is a very nuonce and dynamic as black experiences
They'll hear it means everything
Search npr black stories
Black truth
Wherever you get your pockets
