December 24, 2020 • 63 mins
FRED is a data service provided by the St. Louis Federal Reserve. FRED is updated daily and allows 24/7 access to over 800,000 financial and economic data series from more than 100 public and proprietary sources. From kindergartners to economists to small business, the vast amounts of data allows us all to make more informed decisions. This week, we welcome Yvetta Fortova, the manager of FRED and FRED-family products (ALFRED, GeoFRED) in the Research Division, where she oversees the collection, organization, and publication of data found in FRED. We also have Maria A Arias who works on the FRED team developing and maintaining the data update process. She also helps lead FRED workshops for new and existing audiences to market new features and help users make the most of FRED.
Get your questions answered on the upcoming show by posting your questions in our community: https://bit.ly/ddre-26
00:00 The Data Driven Real Estate Podcast welcomes Yvetta Fortova and Maria Arias from
@Federal Reserve Bank of St. Louis Federal Reserve Economic Data (FRED) team
02:16 When was FRED launched?
03:34 How many data time series are housed within FRED?
09:36 What is GeoFRED?
12:03 How big is the FRED team dealing with 800,000 data series?
16:20 Kindergarders use FRED?
18:56 Who decides the data sets that are featured in FRED? Who are some of the data partners?
33:23 How frequently is the data updated?
39:40 Does access to data at FRED inspire others, like states, to do a better job curating their own data?
47:45 Example of a real estate investor exploring a completely new market. What kind of data is available on things like the economy, demographics, and real estate?
57:10 What is FRED working on for 2021?
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Aaron Norris (00:05):
Welcome back to the Data Driven Real Estate
podcast, the podcast for realestate professionals dedicated
to driving business using data.
I'm Aaron Norris, and along withco-host Sean O'Toole, CEO of
PropertyRadar, this is Episode26. And we get to interview
FRED. FRED is housed within theSt. Louis Federal Reserve and we
are very excited to interview aswe've been longtime users and
huge fans for a number of years.
(00:27):
This week, we've got YvettaFortova, she is the manager of
FRED and FRED-family ofproducts, including ALFRED and
GeoFRED, which I didn't evenknow existed until this
interview. We also have MariaArias who works on the FRED team
developing and maintaining thedata update process, the data
process on hundreds of thousandsof different data series. We
cover everything on what that'slike, and the different data
(00:48):
sets that are available, howthey inspire everybody from a
kindergartener to a PhD, how toexplore the data and make sense
of it and release their ownresearch, and how you could even
become a forecaster and getworldwide recognition. You won't
want to miss this week. So, hey,Maria, and Yvette, we really
appreciate you being here today.
And I guess the question I wantto start with is why data? What
(01:09):
keeps you excited about thisindustry? What do you love?
Yvetta Fortova (01:18):
Hello, Aaron, delighted to be here. Just
before we start, we would liketo say that the views expressed
here are the views of our ownand not the views of the, or
the, of the Federal Reserve Bankof St. Louis, or the Federal
Reserve System. But data is veryexciting to us. And we work on a
(01:42):
FRED team, which is a teamspecializing in dissemination of
Federal Reserve economic data.
And really, data is very uniquebecause we really like the fact
that our website can deliverdata service to users. And data
can mean stories. Because whenyou look at data, you can see
(02:05):
trends in data, you can see whenthe data goes ups and downs. And
that always comes with aninteresting story.
Sean O'Toole (02:16):
I've been a longtime FRED user, you know,
probably since it firstlaunched, when did it first
launch?
Maria Arias (02:26):
Yeah, that's a really interesting story,
actually, officially on the web,FRED launched in 1991. But he
ideal FRED actually started inthe early 60s. So, the research
director at the time his namewas is Homer Jones. He wanted to
(02:46):
share some monetary data withother policymakers, but also
with the general public. And so,he started disseminating a memo
that contained these three datatables. And so, for a long time,
it was just like a memo that wasupdated, of course, like once
people received that they wantedto keep receiving it with an
updated value every, every monthor every week, however often
(03:09):
this was. And eventually, in theearly 90s, FRED became a dial up
bulletin board, and it's justgrown from there.
Sean O'Toole (03:18):
Wow. Okay, so, I'm not nearly as early as I thought
I was on that.
Aaron Norris (03:25):
And when you say grown, can you describe sort of
the vast amounts of data? Likein your library?
Yvetta Fortova (03:34):
Yes, FRED has definitely growing
exponentially. We've, as Mariasaid, in 1991, when FRED first
started, we had about 30 timesseries. And when you talk about
time series, think aboutunemployment rate, or gross
domestic product, those are thetype of time series that were
(03:54):
available. And then we'vereally, with technologies and
with a lot of automations. In2013, we reached our milestone
with 100,000, 100,000 series.
And today, we are almost at an800,000 time series that we
continuously maintain and updatein FRED.
Sean O'Toole (04:22):
Just again, I want to keep like help orient our
readers, right? So, this is abasically a service of the St.
Louis Federal Reserve, right,that makes that available. And
it's at FRED.StLouis, tell, tellus where people go to accesses
if they haven't before.
Maria Arias (04:41):
Yes, FRED is a data aggregator, which means that we
get data from public websiteslike government institutions,
international organizations, andthen some private institutions
and academic resources. And wekeep that up to date in our in
our website, and so the way thatthis works is we're putting all
(05:02):
of this data in one place forusers to be able to combine and
compare different time seriesfrom different sources. So, for
example, if you want to compareemployment growth and economic
growth using parallel employmentand the Gross Domestic Product,
those are, both of those seriesare produced by different
(05:25):
government agencies, the Bureauof Labor Statistics and the
Bureau of Economic Analysis. Butin FRED without leaving our
website, you can plot both ofthose in a graph very quickly,
and then compare themimmediately. And again, for
everyone who has not visitedFRED yet. It's that
FRED.stlouisfed.org.
Sean O'Toole (05:45):
Perfect. Yeah, that's great. One of the other
things that's, you know, youknow, really impressive to me
about it. And it's not, itwasn't perfect, right,,like so.
And maybe you could talk aboutthis, but you're getting this
data from all these differentsources, right, and now you're
gonna lay it on a chart andcompare it. And that presents,
as a data guy, that presentssome real challenges, right? Are
(06:05):
you getting the same timeslices, you know, quarterly
versus monthly versus annually.
And how much of your work goesinto just getting these things
to align so that they, they workon a chart together.
Yvetta Fortova (06:19):
So, over the years, we developed a process
where the notion of the timeseries is well defined in our
back end. And really, it's alldependent on the frequency of
the data. So, you can have data,like interest rates, there's
your interest rates, who arepublished, with a daily cadence,
(06:40):
and the data is on a dailyfrequency. And then you can have
data like labor markets, whichare published by Bureau of Labor
Statistics, and those data aremonthly. And really, when all of
this gets put together on ourwebsite, users can combine these
data together on a graph. Andthere is a kind of like a magic
(07:04):
in the, in the, in the in, thebag that happens that is able to
translate that this data, whileit's the treasure data, while
it's daily, and the labor data,which is monthly can end up on
one graph. And we have a highlyinteractive graphs, so users can
see their values and downloadthe data from the graph. And
(07:26):
really, that is, what makes FREDvery unique is the fact that we
are trying to provide this dataservice to public and, and
recognize the value in beingable to give up to date
information to our users.
Maria Arias (07:44):
Right. And just to add a little bit on to that some
of our tools that we provide, aswell as, while you can add
multiple series to the graph,you can also edit the graph
right then and there. So, forexample, if you want to compare,
like you mentioned quarterly andmonthly data, you can convert
the monthly data to a quarterlyfrequency, as well. And so
that's all the calculations areall done in the background. And
(08:06):
then you can pick if you wantlike an average for those
monthly values converted to aquarterly frequency, or if you
want a different type ofaggregation, things like that.
So, it can really allow you tocreate a customized graphs for
whatever your needs are, orwhatever data you're interested
in.
Aaron Norris (08:24):
How did this, how did this line at the St. Louis
Federal Reserve, or the otherbanks doing something similar,
or they're just like St. Louishas got this.
Yvetta Fortova (08:33):
So, FRED originated in St. Louis, and we
are part of a researchdepartment. And as, Maria kind
of diluted that, really the inthe 60s, the big vision for
Homer Jones, who was a, who wasa research director, at that
time, he really liked the visionof allowing data to be available
(08:59):
to public and really, thisproduct has since been
organically evolving over theyears with multiple other side
products and creating this FRED,FRED-family of products. So,
FRED is the mainstream of thedata but then we have a mappable
(09:19):
data in the in a GeoFRED andthen we have also real time data
in ALFRED so we're really tryingto capture as much as much
pieces and angles of the lookingat the data as possible.
Sean O'Toole (09:36):
Somehow I completely missed GeoFRED like
I've been a FRED user for Ican't tell you how many
presentations of mine have atleast one FRED chart in it, if
not quite a few. So um, but Icompletely miss GeoFRED. I just
found it this morning. And Ionly got a chance to play with
it after Aaron mentioned it. Andso, I don't know how I possibly
(10:01):
missed that. How long is thatwe've been around?
Yvetta Fortova (10:05):
Quite a bit. We, and what GeoFRED is just a way
for users to see cross sectionalcomparisons. So, it's just a way
of looking at the data on a map.
So, in comparing the States inthe US, for example, and if,
you're right that it wasdeveloped on a side, and maybe
(10:26):
it was not as well, as wellprovided to our users, but we
have been trying to, over theyears trying to incorporate the
maps on the FRED website in abetter way. So, nowadays, if
users would be able to, would beable to see View Map button next
(10:49):
to their graphs, that's anindication that they can also
examine and then analyze thedata on a map. And in addition,
we also provide globes, imagesof globes below the graph, in
the related content sectionwhere users can also see data in
GeoFRED.
Sean O'Toole (11:10):
Okay, I was trying to do I'm sorry, Aaron, you go
ahead.
Aaron Norris (11:14):
Visualizations are just very powerful. So, I have a
lot of playing to do. I too,have been a long term user of
FRED and as pulling togetherdifferent data from different
governmental agencies has beenvery tricky. And then I found
FRED and it was a godsend. Ican't tell you how much time you
guys have saved me over theyears.
Sean O'Toole (11:34):
We're total fan boys just to be clear. We're
definitely your fanboys.
Aaron Norris (11:39):
Sean, I already asked Maria. I'm like, Where can
we get those sweatshirts?
Sean O'Toole (11:43):
I know, I totally want one.
Maria Arias (11:46):
Yeah. The FRED swag is available at the St. Louis,
St. Louis Feds, the economymuseum. Unfortunately, it's
closed right now. But as soon asit open, we're expecting to have
an influx of FRED fans coming toget some FRED swag.
Aaron Norris (12:03):
How many how big is your team, like when you're
dealing with 800,000 timeseries? What kind of manpower
does that take?
Yvetta Fortova (12:14):
Our team is, is relatively small, we have a team
of about 10 people. And theyconsist of the developers who
are working on making sure ourwebsite functions properly and
developing new features and newtools to FRED. And then the
other part of our team, our dataengineers who specialize in
(12:36):
making sure our data and contentin FRED is, is up to date. And
because FRED is so popular, wealso have a lot of help outside
of our team within the researchdepartment to help us with other
other things related to dataand, and content.
Aaron Norris (13:01):
Do you have any idea how many users are on your
site any, any specific time ofyear?
Yvetta Fortova (13:10):
Guess, our users fluctuate. And we have kind of
like a seasonality in our, inour user, in our users. And that
is, that is correlated a lotthat education, education and
semesters at schools because oneof our core users are academia,
(13:33):
professors and students whoutilize our product to learn
about economy, money andbanking. FRED is in a textbooks
for students to learn about dataand do homework. But overall, we
do have over over millions ofusers annually that come to our
(13:56):
sites from many countries, andare hungry for data and for
information about data and whatand graphs.
Maria Arias (14:08):
Yeah, something that we think is really cool
about FRED users is at any pointin time, we could have a person
who is just now learning abouteconomics, it's their first time
looking at a graph, as well as aPhD economist or even Nobel
Prize winning economist on thewebsite. So, again, we try to
make sure that our features areeasy to use, the data is
(14:30):
accessible for everyone to learnabout, but also that we have
tools that more advanced userscan take advantage of and also,
you know, like everyone inbetween. So, we're, again really
happy to to hear when peoplereally like our product, and we
try to keep developing newfeatures for you guys out there
(14:52):
who are using it to collect thedifferent data that you need.
Sean O'Toole (14:57):
I personally think it should be you know, there's a
whole class on it in highschool, maybe elementary school,
right? Like for every personbefore they're allowed to
graduate before they're allowedto vote, like they should know
how to use FRED, like, it wouldclean up half the crazy ideas
that I see on, on Facebook andTwitter, and all the wrongheaded
(15:19):
ideas out there. Because it'sjust such, you know, you can ask
these questions rather than, youknow, guess about them. And it's
just, it's really an amazingresource.
Maria Arias (15:28):
Right, even thinking one step back, we, one
of our core missions is topromote data literacy. And
that's something that is veryimportant to us. So, just going
back to understanding, you know,what is the frequency of this
data? What is the ,what are theunits, you know, how can you
compare two different series,they have to be in the same
(15:49):
units in order for you tocompare apples to apples and not
apples to oranges. So, we workvery closely with another group
in the research department. Thatis the Economic Education Group,
and they create a lot ofdifferent materials from
kindergarten all the way to highschool and college level
educational resources for, forstudents out there and teaching
(16:13):
assistant teaching materials aswell.
Aaron Norris (16:17):
Really?
Sean O'Toole (16:17):
That's amazing. I didn't know that, that's, that's
great to hear.
Aaron Norris (16:20):
I'm gonna have to find links to that. That sounds
really cool. Kindergarten?
Maria Arias (16:24):
Yes, down to kindergarten, there are some,
actually there stories that areread out loud. And so, you could
just sit your kid in front of atablet and play this story that
teaches them a lesson aboutfinance, or education or
economics, things like that, andyou find resources at
econlowdown.org.
Sean O'Toole (16:44):
Might have to start with that kindergarten,
one for my 18 year old.
Aaron Norris (16:49):
And build your way up.That's amazing.
Sean O'Toole (16:50):
Build his way up?
Yes. He's actually pretty good.
He's had to put up with me ashis dad. So, he's not that bad.
Aaron Norris (16:56):
Now, somehow, I don't think he's gotten away
with scot-free for sure.
Sean O'Toole (17:00):
No, yeah.
Aaron Norris (17:01):
Maria, you have to deal with a lot on the on the
back end. And I really wantpeople to appreciate what it
means on the data side. So, whatis your day to day look like
when data is coming into thesystem? How does that work?
Maria Arias (17:13):
Right. So, for our process, currently, a lot of it
is automated. And so, we havedeveloped a Python-based
pipeline where we, when the datacomes out, we reach out to the
source's website, collect thatdata and process it and save it
onto our database. And at thatpoint, it's available
immediately on our website. Butwe have a set of like
(17:37):
notifications and things likethat throughout, the throughout
the day with these processes,making sure that everything is
the data is updatingsuccessfully. And if there are
issues, we would go in and checkeach like specific data set or
release to see what what ishappening. And then this is kind
of for, like maybe 90% of ourdata is automated, updated in an
(18:02):
automated way. And then we also,for some, like for the remainder
of that we either manually haveto go and check to see if the
data has updated on the source'swebsite. Or maybe we wait to
receive an email fromsubscription, that the data is
available, things like that. So,there are a few kind of outliers
(18:23):
out there that involve some sortof manual intervention to get
the data updated. But kind ofeven before we get to that
stuff, we as we add new data toFRED, we have to add them to
this data pipeline. So, thatinvolves us, as data engineers
understanding the data and howthe data is provided by this
(18:44):
particular source or theparticular data set that we're
adding, so that we can processthat data and kind of
standardize the data so that wecan save it to a database.
Aaron Norris (18:56):
Who decides what data gets to be featured in the
app?
Yvetta Fortova (19:01):
So, for us, we have lot a long list of data
requests that comes from variousplaces, our users, the data
providers themselves, who wouldlike their data added to FRED,
and as we mentioned, our team isrelatively lean. So, what we do
(19:22):
is we have internally have acommittee that meets on a
regular basis. And then we, weselect data that we think are
relevant to our users, and theyare reliable as well from the
originating sources. So, thatwe, we can continuously provide
(19:48):
new content to users and and wecan, we can, we can give our
users additional data to workwith. So, for example, recently,
we have added weekly pandemicunemployment claims, which has
been developed by Department ofLabor. And we have also worked
(20:15):
on addition of housing datafrom, from Optimal Blue and
realtor.com, which are alsoavailable on our website. So,
anybody who's interested, can goand find them.
Sean O'Toole (20:29):
I saw one on GeoFRED it was market hotness.
That seemed to be a real estaterelated.
Maria Arias (20:38):
Yes, but as a data set from realtor.com, as well,
they have a set of indicatorsthat they define as market
hotness indicators. So, it's adifferent different topics that
kind of like how many days ahouse has been or like, average
(20:59):
of how many days houses havebeen on the market in this
region, things like that. Andso, that's a data set that they
put together and we publish inFRED.
Sean O'Toole (21:08):
On GeoFRED, it had very few counties populated from
what I saw, yeah. Is that stilljust a work in progress? Is that
the reason or did they just nothave data on most of the
country? Or do you have anyideas there?
Maria Arias (21:19):
Yeah, so, that's a we have their complete dataset
in FRED for the market hotnessdata, they have the only the
largest. I don't know how many Ithink it's the largest, like 100
MSA. And then I think thecounties that are included in
that data set are from those MSAand surrounding areas as well.
So, it's not a county by countyfor the US for this particular
(21:43):
data set.
Sean O'Toole (21:44):
I happen to pull down County, if I pulled down
MSA, it probably would havefilled in a little nicer for
those MSAs versus County. It'slike only when the MSA maps to
the county, would it work?
Right?
Maria Arias (21:56):
So yeah, what depends on what data they have
available for that particulardata set. And yeah, I know that
for that one. In particular,it's just the certain number of
the largest MSA and then thosesurrounding counties.
Sean O'Toole (22:10):
We have a lot of data users. And I want to come
back to your your data pipeline,right. So, in that, in, for that
data pipeline is, you know, I'veused obviously, the website and
the graphical tools, right?
Folks that want to get access toraw data, is your data pipeline,
open source? Do you guys have anAPI to the raw data? What are
(22:33):
the, is there any options there,they've got to go do all the
same hard work you did.
Maria Arias (22:40):
Right. So, the work that we do, hopefully will make
everyone else's work a loteasier. We are doing the hard
work of going to, you know, theCensus and the BLS and all of
these other organizations andparsing their data that is
provided in different formats.
And putting it all in like onestandardized format in FRED. And
so, for all of you who want tomaybe automate your workflows or
(23:05):
things like that, we also have,we have a FRED API. And we also
have an add in for Excel thatyou could use. So, that allows
you to, again, automate yourworkflows in different ways and
combine it with the other datathat you already have. So, if
you like programming for theAPI, in particular, you just
(23:25):
have to create a free useraccount and request an API key.
And there are third party kindof wrappers or packages that are
already out there that otherpeople have created. And they
just kind of wrap around theFRED API and make it easier to
use. And then that's in additionto you know, if you go to the
FRED website, you can downloadthe data as a CSV or an Excel
(23:48):
file. And even if you've addedmultiple lines to the graph, if
you download the data from thatgraph, it'll include all of the
data that you have. So, that'salso another way that, that you
could access the raw data. Andthen
Sean O'Toole (24:04):
All the above love it.
Maria Arias (24:06):
Awesome. And then you have for the GeoFRED data as
well. If you want to downloadall of the say, like all of the
county level data for the entireUS, that's probably the best way
to download all of that data isthrough GeoFRED for that
particular map.
Sean O'Toole (24:21):
And all the data itself, basically at least is
provided from you is opensource, and people can use it in
their applications and that kindof stuff for their license
limitations at all?
Yvetta Fortova (24:33):
So, most of the data we have come from
governments or sources orinternational organizations,
and.
Sean O'Toole (24:43):
Public record.
Yvetta Fortova (24:44):
Public domain data, correct. But we do have
small percentage of data thatcomes with copyrights. And users
will be able to identify thosedata because we have a note
under the graph that specifieswhat the copyright, the
restrictions are. And we alsohave in a notes, what that means
(25:07):
to the end user. And if theyneed any additional actions to
utilize the data, but overall,from the whole scheme of data
you have in front, it's a smallpercentage of data.
Aaron Norris (25:23):
I like to, Yvetta what you said about data telling
stories. And part of that is,what I like about data is
there's a trust factor to it.
It's, it's just the data, thedata is the data, data doesn't
get political, it tells a story.
How important is the datavetting process, when you
consider new partners? I was, Ihad no idea that you were
pulling in like a NationalAssociation of Realtors, which
(25:45):
is an association group. Whatdoes that data vetting cycle
look like?
Yvetta Fortova (25:53):
When we are looking on, where like, what
kind of sources, the data comefrom? We are we are looking on,
how much data would be availableto us? And also how, what the
data look like how is the dataconstructed? If it's data that
is academic data, we are lookingfor a ways to see whether there
(26:20):
is some peer review work and orpaper that has been published
for construction of the data?
And also, if there are ways thatthis data, how is the data
disseminated all in all. So, inwhat way we, as a team can
maintain the data, which is alsoa big piece. Because we are
trying to look for ways of, ongetting the data in an automated
(26:45):
way. And cut down on the manual,manual work. So, there are
pieces of of that. Andobviously, it also comes down
with private sources to havingproper licensing agreements in
place.
Aaron Norris (27:05):
I know you've been on the research team before. So,
do the data sources come in andthen your research team gets a
list of all the new data sourcesand they dream up what's next?
What triggers some of theresearch that your team does?
Yvetta Fortova (27:20):
So, without our, within our team we do like our
team does not do research. We'rejust solely doing breathing and
living from FRED.
Aaron Norris (27:30):
Yeah.
Yvetta Fortova (27:32):
But the economists who are in the
research department, they, theypretty much have, have really a
choice in whatever they wouldlike to publish. And, and and
write about. And their process.
I'm not familiar with theirprocess in details, but.
Aaron Norris (27:58):
That's okay. I was just curious, like, I can't
imagine having access to 800,000data series, and you just wake
up in the morning like, 'Huh,what comedy work I got to put
together today?' That's amazing.
Sean O'Toole (28:10):
Do either of you have any favorite data series?
Maria Arias (28:16):
It's a, it's hard to pick with so many of them.
But I didn't have a...
Sean O'Toole (28:22):
To look at in the morning or when the new, when
the new drop comes that like,'Oh, that's the one I want to
see.'
Yvetta Fortova (28:29):
Yeah, we've definitely become kind of a data
nerds over the years workingwith the data. And there, some,
some data are definitelyinteresting to talk about. For
example, we do have data in FREDthat is the longest series that
(28:54):
we publish, and it's ahistorical data for population
of England, that goes back downback to 1086. So, it's really
depicting a plague that has,that impacted population. And
then you have some reallyinteresting data on orders of
(29:18):
sinks, kitchen sinks and toiletsduring industrialization period,
and the, the peak in indoorplumbing that has happened
during that time. So, really,those are, those are our kind of
like, nice dimensions. Butreally, when it comes to
(29:38):
favorites, it changes all thetime because we, data is really
interesting, when there aretrends and when there are
changes in the data as you asyou deliver it to the to just
spikes and drops in data. So,really looking at business
cycles and seeing how the datais changes fit the business
(30:01):
cycles or after the businesscycles is, I think the most, the
most interesting part oflooking, looking analyzing the
data.
Aaron Norris (30:10):
Data is such a hard business too and I'm sure,
as you've made more available, alot of people are very
demanding, I think one of myfavorite series has always been
migration. And especially withCOVID, you hear a lot of,
there's been a lot of talk inthe real estate space and
different data series to tracknot having to wait for a census,
are you guys getting a lot ofpressure to provide more series
(30:32):
that are a lot more timely,instead of waiting for
government, census and things ofthat nature.
Maria Arias (30:39):
We've seen a lot more demand for more, not not
only more timely series, butalso data that are more frequent
or that have like a higherfrequency. So, say daily or
weekly data instead of likequarterly or annual data. And
so, especially this year, when,you know, if you think about GDP
(31:02):
and the state of the economy,you have to wait for the quarter
to be finished. And then onemore month for the initial
release of the previous quartersGDP to come out, right and ask
like things are changing in thecountry, it's hard to tell what
the state of the economy is atthat point. So, we've definitely
seen an increase in demand.
Unfortunately, we're also kindof bound by, you know, we want
(31:23):
to continue to provide highquality data to users. And we
can't just add everything rightaway either. And so, it's, it's
been kind of a, trying to seewhat the data providers are
creating for like new releases,for example, the census created
a, they created a experimentaldata set from their quarterly
(31:50):
business formation data, theystarted calculating weekly
business formation data. And assoon as they started making that
available, we added that to FREDin the summer or early summer,
late spring, something likethat. And so, now there's this
weekly level, business formationstatistics that has been really,
really popular.
Sean O'Toole (32:12):
Is there what's the Geo-granularity on that?
Maria Arias (32:15):
Um, that is available at the state level?
I'm not sure if it's alsoavailable at the MSA level.
Aaron Norris (32:22):
It's so, funny, Sean. We were just talking about
that with Doug Duncan.
Sean O'Toole (32:25):
Doug Duncan, Chief Economist for Fannie Mae was
just, he was just saying that'sthe one data set he would most
like to have right now, butreally, probably more down at
the county or MSA level? I'mgona have to dive into that one,
for sure. And we'll have toshare it with Doug as well.
(32:46):
Yeah, so, out of you, you know,you mentioned 800,000 time
series, like, give us, give ussome idea, you know, a guess is
fine. But like, what's thefastest data set you're getting?
You're not like a Bloombergterminal where it's, you know,
sub, sub second updates on onstocks or anything like that.
But the fastest is maybe daily,hourly? And then the worst is,
(33:10):
is annual or is it some of themeven less often than that?
Yvetta Fortova (33:17):
Oh, are you? Are you referring to how quickly we
can get the data to FRED? Orwhat kind of data we have?
Sean O'Toole (33:23):
I think there's two separate questions there.
And let's do both. Let's do thefirst one is how often, right?
So, out of your 800,000? Howmany of those are annual versus
quarterly versus monthly versusdaily versus hourly, right? Just
a guess. And then, and then, youknow, she mentioned, Maria
mentioned the, you know, some ofthese things have a month delay.
(33:45):
And so, both of those things arereally interesting issues,
right, like, that I don't thinkpeople think enough about we
have to think about it a lot,because we provide public
records data on real estate,right. And in a small county, we
might go get the data everyother month, right. And in LA
County, which is the largest inthe United States, right? We
(34:07):
have stuff, you know, very, veryquickly, so we get it daily.
And, you know, we usually haveit published to our site within
just a couple of days, versusAlpine county might get it every
two months still publish to oursite within a couple of days,
but there's big differencesthere. So, can you speak to that
kind of distribution among yourdatasets?
Yvetta Fortova (34:30):
The data we have in FRED are anywhere from daily,
weekly, monthly, quarterly,semi-annual annual, and then we
have some five year frequencydata. And really the
distribution among what kind ofdata we have in FRED is more or
less, we have mostly, most datafor on annual frequency and then
(34:54):
it it kind of goes down fromannual to quarterly, monthly And
daily. So, that is kind of likethe distribution of the data we
have in FRED. But then there arealso aspects of how, how, how
the data updates. So, when we doour updates, let's say that an
(35:16):
employment data, the employmentsituation that is published by
the Bureau of Labor Statisticsis published on the first Friday
of every month. And it comes outaround the 7:30 Central time.
So, we really are trying to hitthe, the update process,
(35:38):
immediately or minutes after thedata is released. And then
obviously depends on the bit onthe size of the data itself. So,
if employment situation hascouple 1000 series, it takes
some time to get all the newinformation into FRED. But we
are trying to optimize ourprocess to really get into it in
(36:01):
minutes or from what the actualrelease of the data happens. And
then there are also aspects ofthe data itself. So, when the
data is released, there arethere is the lack of the
information, because whathappens is, first Friday of the
month, there is a publishpublication of the data. But
(36:24):
actually, the data that'spublished represents the values
for the previous month. And thenthat goes all the all the way
back down to the origin eatingagencies who are working on the,
on obtaining the underlyingdata, and perhaps the micro
data. So, they need time to geta good size sample of the data,
(36:49):
which then allows them to modelthe data and make the aggregated
values available. So, those inmost data, if we have in FRED of
work on on this typical lackwhere you're not going to see
data for December, monthly datafor December, if the December
(37:12):
isn't over yet. And again, itcomes back down to, you know,
the originating agencies havingtime to prepare those estimates.
On the other hand, we also havea little bit of advanced data or
data that is forecasting thatare looking into the future.
(37:35):
Which users can also also use toto look at what the data will
look like. And last not butleast because all of this
process of disseminating thosedata and publishing data is
really based on a lot of timeson surveys and on incomplete
(37:56):
samples. That's why agenciesover time, as they are
publishing new values, newmonthly data, let's say they are
also going back and revisingexisting data. And for that
reason, we also have ALFRED,which is an archival, FRED.
Aaron Norris (38:16):
Interesting.
Sean O'Toole (38:19):
Yeah, And that was Alfred was another one I wasn't
familiar with. And, you know,Aaron mentioned this morning
when we were doing our pregame,you know, prep. And, you know,
he mentioned, he had a dataseries that changed quite
dramatically, you know, and oneof the things that he was
following and, and he was ableto find that the historical data
(38:40):
and ALFRED and see why, youknow, see that difference. So,
that's, that's pretty cool thatthe archival piece too,
valuable.
Yvetta Fortova (38:49):
Yes, I think the the best example that we like to
give is from gross domesticproduct. So, if you think about
gross domestic product, the dataitself, is a quarterly frequency
data, but then the values areannualized. And then the data
(39:10):
updates monthly, and then thevalues to revise could revise
every month. And on top of thatthere, there is an annual
revision that happens every yearin summer. And on top of that,
every three to five years,there's a benchmark revision
which may change everythingunder the GDP.
Aaron Norris (39:32):
I did not know that, wow.
Sean O'Toole (39:33):
GDP doesn't mean anything. It's just a random
guess. Just kidding.
Aaron Norris (39:40):
I.. do you find that your work has inspired
states to get a little bit moreserious about collecting data. I
noticed just in California as anexample in the last year, I've
noticed a lot of data collectionsort of budget line items in
legislation talking aboutdigitizing and starting to track
(40:00):
things. And can you point tosome of your work being so
helpful that states are like wegot an upper data game?
Maria Arias (40:08):
We would like to think that we inspire other
government and publicinstitutions to improve our data
collection. In some cases, whereespecially like smaller
government institutions havereached out to us about, you
know, having their their dataand FRED and it's not
necessarily available in amachine readable form, we've
(40:31):
been able to work with, withthem, and even some of the
academic data that we've added,that seems to be pretty popular.
And FRED, we've been able towork with them to help them put
their date make their dataavailable on their website in a,
an easier form for us to parsebut also for other people. So,
(40:52):
like machine readable forums,standardized tables, things like
that. And so, we have seen alittle bit of that. But we don't
work with like, we don't workwith every government
institution out there that wecollect the data from, because
again, all of this data ispublicly available. So, we just
(41:12):
collected from their website.
But we would really hope thatmore and more, especially
smaller government entities, tocollect their data and make
their data publicly available inan easy format, because it's not
only for us to provide a servicefor other users, but also, you
know, everyone around them, andin your communities, you know,
(41:37):
you know, what's important toyour community. And you know,
what industries are what drivethe community forward, things
like that. And so, if thegovernment's local governments
were able to help collect thatdata and make that data more
available, it would be easier toalso analyze the state of the
economy and these kind ofsmaller geographical regions,
(41:57):
like you were talking aboutearlier, it's really nice to
have county level data, it wouldbe even nicer to have it at a
smaller geographical level foreveryone who's doing research
and trying to understand what'shappening.
Sean O'Toole (42:14):
Our company, the one that sponsors, the Data
Driven real estate podcast,PropertyRadar, really only
exists, because of how difficultit is to get public records data
on real estate assessors data,recorder data, you know, etc.
And I would, it would, it wouldmake me so happy to shut this
(42:35):
company down, because the datawas directly available. On an
easier format, like out of allthe things that we spend money
on as a government, you hearbillion dollars here a billion
dollars, they're like, why yourgroup and the group's in all
these different levels aren'tfunded to the tune of billions
of dollars? You know, I have tosay, I was a little dismayed to
(42:58):
hear how small your team was,and I kudos for what you guys
accomplished with that team.
But, you know, if you ever needsomebody like you to write a
letter on why you should getmore funding, like you can count
me in.
Aaron Norris (43:15):
Yeah, the actionable side of data I serve
on the county board for 211 outhere, and you know, a lot of
counties will do is to
Sean O'Toole (43:22):
What is 211 Aaron?
Aaron Norris (43:23):
Oh, sorry, 211 is a Health and Human Services
hotline. It's, it's nationwide,it's sort of like the 411 for
help. It's it has a lot ofnonprofit data government
assistance programs. And I thinkwhy I joined was the data is
that you don't have to wait ayear for to spend $100,000 on
what people need in thecommunity. This 211 hotline is
(43:44):
tracking the need. So ,30% ofthe need in our community for
the last three years has beenconsistently been housing, you
don't need to spend $100,000 inwait a year, it's almost
immediate. So, it's, it's datadoes tell a story. And sometimes
it's really important becausegovernment entities can redirect
resources, where it's needed,instead of wasting time, and
(44:04):
you're a year late. So, I'm justcurious, what are some of the
data that you're most excitedabout in the last few years that
you've seen cross your desk thatyou wish people knew about?
Maria Arias (44:20):
So since.
Sean O'Toole (44:20):
800,000, there's too many.
Maria Arias (44:23):
I'm just trying to think of what we've added this
year, um, something that we didthis year, and it's real estate
related, so I definitely wantedto bring it up is the Optimal
Blue mortgage market indices.
So, they have mortgage data thatis broken down by different FICO
scores and by different types ofloans. So that's a, again,
(44:43):
another private institution thatreached out to us to have their
data added to FRED and it's a,it was added this year. But then
we also have some other likeyou've mentioned the weekly
pandemics claim pandemic claimsdata. That was a really
interesting one. We also addeddaily FOMC target rates that
(45:05):
were digitized by one of oureconomists at the St. Louis Fed
and some of his co authors goingback to the early 1900s. So, now
you can get, like FOMC, targetrates starting 1900s. And then
like slowly put the time seriestogether, up until present day.
(45:25):
And then what else? Oh, there'sanother one by the a group of
economists this one and this isgoing back to kind of some of
the changes that we've seen thisyear have more demand for high
frequency data is a group ofeconomists created the weekly
economic index by last names ofLouis Martin stock. And this is
(45:48):
a kind of like a now cast, butnot necessarily an outcast, per
se. But they combine severalhigh frequency data that are
available at the national level,and they create like a GDP
forecast more or less at theweekly level. And that is
(46:09):
actually updated twice in aweek. And so, that's been really
interesting to see as well. Andagain, just kind of making the,
kind of having like a morefrequent update of state of the
economy and keeping thatconversation going.
Aaron Norris (46:33):
Is it I haven't looked yet, HMDA data I know is
a I'm a, I have my mortgagelicense. And the amount of
detail that the CFPB hasmandated that we start
collecting at the loan level isquite expansive, is that
available on FRED yet?
Yvetta Fortova (46:50):
Oh, we do not have humped up data in FRED.
Aaron Norris (46:55):
Okay.
Yvetta Fortova (46:56):
But I would say for anybody who is interested to
learn about what data we haveadded to FRED, we would
recommend to sign up for FREDnewsletter, which is in bottom
of our web pages, and we letusers know what new changes in
(47:16):
data and what new data we editthat way.
Maria Arias (47:18):
A new feature sets...
Sean O'Toole (47:19):
Right there, that's, that's, that's my tip of
the whole podcast, right? Like,because I've been a FRED user
forever, and I'm not signed up.
And I don't know why. So,, I'msigning up right now.
Aaron Norris (47:31):
I'll make sure to post some extra links to to make
sure in the show notes thatpeople can find this stuff
really easy. So, I think in thereal estate space, there's a lot
of like exploring differenteconomies, and maybe some people
they've never approached FREDbefore. And so, maybe we can
give a little bit more industryspecific. And I know this is
your personal opinion, if youwere going to explore a state
(47:52):
that you didn't live in, and youwere interested in, you know,
the economy, real estate, Idon't know what kind of things
would you explore in FRED peoplemight not know exist.
Yvetta Fortova (48:05):
So, definitely, for someone who is new to FRED,
we would say, to go and try tosearch for information and see
what kind of data we haveavailable for, for given states
or MSA's or counties. On theregional level, we'd, and they
will quickly realize that we dohave a lot of major economic
(48:30):
indicators like labor data andgross domestic product. But then
we also have data on propertyprices, like house price
indices, and we have rates onmortgage rates. So, there's
really a wealth of informationavailable. And it's a, it's a
(48:52):
really matter for people to goand try it and see what we have,
you know, from housing stars andconsumer price index, to prices
and wages. And it's kind oflike, when you go shopping, and
you feel like you've reallygood, got a good bargain at the
end of the day. So, that'sreally what we are trying to
(49:15):
give users with FRED.
Maria Arias (49:19):
Right. And then I would just add on to that, that
if you find some graphs that youfind really interesting, or that
you would like to come back to.
This is just kind of a plug forsome very useful FRED features.
And especially if you don't havea FRED user account yet, right
at the bottom of the, of thegraph, you can save graphs to
your user accounts. But then youcan also add graphs to a
(49:41):
dashboard. So, as you add graphsof say, a similar topic that you
are saving to your user account.
You can also put all of thosegraphs or maps onto a dashboard
and add other like snippets,like notes or like a single
number or something like that.
And then all of these graphs canbe saved so that they are
(50:02):
updated automatically. So, thenext time the data comes out,
you just go back to yourdashboard or to your saved
graph, and the data isautomatically updated for that.
So, this is a very usefulfeature that if you're not
familiar with FRED, or if youdon't have a user account, I
highly recommend you check out,especially again, if you have
these kind of several graphs orindicators that you'd like to go
(50:24):
back to. And then together withthat, you can make your
dashboards public and share themwith your colleagues or share
them with other people in theindustry, to just, again,
simplify sharing data andsharing content.
Sean O'Toole (50:42):
But, you know, one thing I haven't looked at, and
I'll just ask is, can you embed?
Can you get like an embed codefor those graphs, so you can
embed them in a website orwhatever, so that they're
constantly updated on your siteas well?
Maria Arias (50:55):
Yes, just below the graph, there are some shared
links, and account tools. So,under the account tools, that's
where you can save those to youraccount. But then the share
links, you can share a URL to agraph. And so, whatever
modifications you've made to thegraph, and you share that, using
the URL provided under thatbutton, the person who accesses
(51:18):
that will see exactly the graphthat you created. And then same
with the embed code, if weprovide like a pre created
snippet that you could just putonto your website and embedded
that way.
Sean O'Toole (51:30):
In my, you know, one of the things I really like
is the, the bars that showrecessions, right, um, and,
excuse me here, if I like, don'tknow, that, in terms of those
bars, that kind of underlieright recessions is, is that the
only one or there are someothers?
Yvetta Fortova (51:51):
We only have recessions at this point on the
graphs. And there are other waysyou can, you can take those
recessions off from the graph,if, if there's no need for them.
But we also have a way tomanipulate the data of it in the
(52:12):
graph itself. So, for example,if you have, if you have a data,
you can then index the data todifferent time of the year, or
different period in time and,and it will set the values to
100 of a given period. And thenyou can see changes to the in
(52:32):
the data. So, that is also oneway how someone can kind of take
the data and, and make it workto whatever they are trying to
save at the data.
Sean O'Toole (52:48):
I want to just expand on that a little bit to
make sure our folks understand,right. So, sometimes you get
these data sets that are wildlydifferent in their values,
right? Instead, in some, youknow, one way to do that in
charts to show one value on thisside, one shot value and this
side. But when you're looking atthe relative change over time,
you guys allow you to set a dateat which you basically normalize
(53:09):
them all. And then you can seehow the change relative to each
other from that, even if theyhave completely different
scales. And that's a superuseful tool. I gotta say, the
one thing that I do a fair bitis, is, is underlie those events
besides recessions, right, like,one of the things that that
(53:31):
folks in the real estatecommunity worry about a lot is
changes in administration, andchanges in the makeup, excuse
me, of the, you know, like thehouse in the Senate. Right. So,
like, everybody obviously isworried right now about this or
thinking about this Senate race,and Georgia. And I'm sorry. And
(53:58):
so, one of the most interestingones I've done over the times is
I go back and kind of like therecession bars. ,I you know,
it's republican versusdemocratic president right, and
then a divided Congress versus,you know, the same Congress and
is it all three? Or is it one intwo like that, like, I would
(54:19):
love some more of those, youknow, different things, that
particular one, you know,administration would be great,
but I'd love to see some morelike that. Because I do think
that is that's a really, youknow, a really cool feature.
Yvetta Fortova (54:35):
Yeah, that sounds really interesting. So
and thank you for thesuggestions, suggestion we can,
we can look on. But what it willlook like and how we can, how we
can approach to have a betteroptions for users and then maybe
have ways for displayingdifferent administration's as
(54:57):
you said, and one more thingthat I also wanted to add is
that if, if it needs to be userscan create draw lines on the
graph, as well. So, there is away to create user defined
lines. So, that is probably theclosest we can get you with the,
(55:18):
with the situation's you'refacing. But definitely, it's a,
it's a place for improvement forus, so thank you.
Sean O'Toole (55:27):
It was an eye opening to me on, on, you know,
what we think about in terms ofwhich party is fiscally
conservative and which oneisn't. But it was not what I
expected.
Aaron Norris (55:37):
And it's good to tell that story in data, it
makes it less political, well,maybe it doesn't. But if you're
a real estate in real estate,and you've never explored some
of the tools, I really liked therecession, because when you go
into an area, and you can see arecession, and how that local
economy, unemployment, householdincome, median price, really
(55:58):
fared through like a recession,it's really interesting to drill
down as you're sort of makingmoves, or you're helping clients
that want to move out of stateand they're exploring a new
area, I just the easiest thingto do is go on the website and
type in the name of a county andjust start playing and see all
the different data series thatare available. It's it's almost
overwhelming. Just have to getin there and explore if Sean, do
(56:22):
you have any other favorite onesthat you use?
Sean O'Toole (56:26):
Oh, you know, the biggest one I use the most often
is I look at the different typesof debt, right? So, mortgage
debt, corporate debt, total, youknow, the the federal debt, you
know, etc. student debt is justa student that that chart is
really scary bad, like, youknow, especially if you take all
(56:48):
of those different types of debtand put them at 100. You know,
say and 2000. And look at howthey change over time, it's, we
clearly have a huge student debtproblem. And I know you guys
probably can't comment on the,on, on those things. But that's,
that's probably one of myfavorites that I look at the
most.
Aaron Norris (57:10):
Well, we are about out of time, I'm trying to think
if there's any other things thatwe should cover, like maybe new
features that you're working on,that you could share.
Yvetta Fortova (57:21):
So, we are working on working on surprises
for 2021. And next year, FRED isgoing to be 30 years old. So,
we're going to be looking insome ways how we can make FRED
more mature looking. But thereare definitely going to be more
(57:44):
data edit throughout next year.
So, that comes up that we triedto do that on a regular basis.
So, so, that we don't loseinterest in you guys. And, and
other other aspects here, we aregoing to try to look at is, we
would like to improve our searchabilities in FRED. And we are
(58:06):
looking into ways how we canexpand on providing users with
some industry break down on thedata.
Aaron Norris (58:19):
Sort of categorized things and in
buckets, or something?
Sean O'Toole (58:23):
I would love I mean, with 800,000, this is a
big ask, but like a littledeeper descriptions on each one.
You know, because a lot of timesyou'll see like, you know,
you'll type something in thatyou're interested in, you know,
inflation or whatever. Andthere's 30 or 40 time series,
and you're like, Oh my gosh,which one of these do I use? I
(58:45):
just wanted this simple thing.
And it's that piece, the number,I mean, the number of series can
sometimes get a littleoverwhelming in terms of which
one should I use. So yeah, any,any suggestions there on how
best to parse that when you getand you're like, Oh my gosh,
there's 30 ways to look atmortgage debt or you know,
whatever, like any anysuggestions on how people can
(59:07):
get to the right one and notmake bad decisions?
Yvetta Fortova (59:13):
Oh, we, one of the things we tried to do and in
our search is to really work themost popular and relevant series
to the top of the searchresults. So, for novice users,
that may be an easy way to, toknow that based on their search,
(59:33):
that the, the series at the topare are going to be the, the
best to use. And also, we try toprovide additional filtering
options on the side of the havethe data and to subset the
searches. But if all else fails,we really and there's really not
(59:56):
a good description in a data. Wereally tried to ask users to see
if they will be able to go backto the originating source and,
and, and see if they have someother descriptions to to make
the decision of which series isbetter.
Sean O'Toole (01:00:16):
I do that a lot I go in. And it's a you know, it's
an amazing chance to learn,right? Because you see others
these different ways of lookingat it. And so, then I'll take
the time to go research thosedifferent ways. And I come out
smarter at the end of the day.
So, it's not it's not all allbad the approach you take, and
Google's your friend on thatright Wikipedia and those things
(01:00:36):
for sure. So great,
Maria Arias (01:00:41):
Right. And part of the filtering options as well is
if you can filter by source. Andso, if you know you want, say
inflation or consumer prices bya particular source, you can
filter by source and that wouldsometimes help you find the
right one or close to whatyou're looking for.
Sean O'Toole (01:01:01):
What else did we not cover that you guys would
like to to share or mentionbefore we wrap up.
Yvetta Fortova (01:01:06):
Um, I would say for anybody who is interested to
do some playing with the data,we didn't mention that we also
have a forecasting game calledFREDcast. And we, if you have a
(01:01:26):
user, have a user account, youalso have access to FREDguests.
And what it is, is, users can bepart of a public leak, and they
can set up their own leaks. Andthey can create forecasts for,
for economic indicators. Theycan forecast unemployment rate,
(01:01:47):
GDP, payroll employments andinflation as a CPI gross, and
every every month, the scoregets calculated, and then you're
ranked among other peopleplaying. And you can see how
good you have scored and whatyour average error is towards
(01:02:12):
other players.
Maria Arias (01:02:13):
You earn public recognition and also earned
badges. So it's pretty fun.
Sean O'Toole (01:02:18):
I'm all about the badge, I was a boyscout, so I
had lots of badges.
Aaron Norris (01:02:22):
That's amazing.
I'm so glad you brought that up.
Maria, do you have anything elseyou'd like to add?
Maria Arias (01:02:29):
Just go in there and search for whether it's your
topic that you're interested, orthe region and the country or
the world that you're interestedin and see what you can find.
Yeah, there's lots of differentways to search for data in FRED,
right under the big search bar,and in the middle, there's also
(01:02:50):
Release Calendar. So, you cansee what data releases are
scheduled to be released todayor this week. And then you can
also search by category. So,that's kind of all of the data
in FRED are grouped intodifferent categories and that's
also a good, a good way ifyou're interested in a
particular topic to kind of godown the different breakdowns of
(01:03:10):
the data.
Aaron Norris (01:03:12):
Fantastic. Well, thank you so much for your time
today. This has been a lot offun.
Maria Arias (01:03:17):
Thank you so much for having us.
Yvetta Fortova (01:03:19):
Thank you.
Sean O'Toole (01:03:20):
Yeah, thank you so much, really appreciate it. Like
I said, we're definitely fans,so keep up the good work.
Aaron Norris (01:03:26):
Thank you for listening to the Data Driven
Real Estate Podcast, you canfind show notes and links to
some of the resources mentionedin the show at
datadrivenrealestate.com. Clickthat join the community, and
you'll be forwarded to thePropertyRadar community where
you can ask questions about thecurrent show and even see
upcoming guests and askquestions there. We'd love to
engage with you in thecommunity. So check it out.
(01:03:46):
Please don't forget to like,favorite, subscribe and share on
your favorite platform whereyou're listening to the show. It
helps us out a great deal.
Thanks for listening, and we'llsee you next week.
Welcome back to the Data Driven Real Estate
podcast, the podcast for realestate professionals dedicated
to driving business using data.
I'm Aaron Norris, and along withco-host Sean O'Toole, CEO of
PropertyRadar, this is Episode26. And we get to interview
FRED. FRED is housed within theSt. Louis Federal Reserve and we
are very excited to interview aswe've been longtime users and
huge fans for a number of years.
(00:27):
This week, we've got YvettaFortova, she is the manager of
FRED and FRED-family ofproducts, including ALFRED and
GeoFRED, which I didn't evenknow existed until this
interview. We also have MariaArias who works on the FRED team
developing and maintaining thedata update process, the data
process on hundreds of thousandsof different data series. We
cover everything on what that'slike, and the different data
(00:48):
sets that are available, howthey inspire everybody from a
kindergartener to a PhD, how toexplore the data and make sense
of it and release their ownresearch, and how you could even
become a forecaster and getworldwide recognition. You won't
want to miss this week. So, hey,Maria, and Yvette, we really
appreciate you being here today.
And I guess the question I wantto start with is why data? What
(01:09):
keeps you excited about thisindustry? What do you love?
Yvetta Fortova (01:18):
Hello, Aaron, delighted to be here. Just
before we start, we would liketo say that the views expressed
here are the views of our ownand not the views of the, or
the, of the Federal Reserve Bankof St. Louis, or the Federal
Reserve System. But data is veryexciting to us. And we work on a
(01:42):
FRED team, which is a teamspecializing in dissemination of
Federal Reserve economic data.
And really, data is very uniquebecause we really like the fact
that our website can deliverdata service to users. And data
can mean stories. Because whenyou look at data, you can see
(02:05):
trends in data, you can see whenthe data goes ups and downs. And
that always comes with aninteresting story.
Sean O'Toole (02:16):
I've been a longtime FRED user, you know,
probably since it firstlaunched, when did it first
launch?
Maria Arias (02:26):
Yeah, that's a really interesting story,
actually, officially on the web,FRED launched in 1991. But he
ideal FRED actually started inthe early 60s. So, the research
director at the time his namewas is Homer Jones. He wanted to
(02:46):
share some monetary data withother policymakers, but also
with the general public. And so,he started disseminating a memo
that contained these three datatables. And so, for a long time,
it was just like a memo that wasupdated, of course, like once
people received that they wantedto keep receiving it with an
updated value every, every monthor every week, however often
(03:09):
this was. And eventually, in theearly 90s, FRED became a dial up
bulletin board, and it's justgrown from there.
Sean O'Toole (03:18):
Wow. Okay, so, I'm not nearly as early as I thought
I was on that.
Aaron Norris (03:25):
And when you say grown, can you describe sort of
the vast amounts of data? Likein your library?
Yvetta Fortova (03:34):
Yes, FRED has definitely growing
exponentially. We've, as Mariasaid, in 1991, when FRED first
started, we had about 30 timesseries. And when you talk about
time series, think aboutunemployment rate, or gross
domestic product, those are thetype of time series that were
(03:54):
available. And then we'vereally, with technologies and
with a lot of automations. In2013, we reached our milestone
with 100,000, 100,000 series.
And today, we are almost at an800,000 time series that we
continuously maintain and updatein FRED.
Sean O'Toole (04:22):
Just again, I want to keep like help orient our
readers, right? So, this is abasically a service of the St.
Louis Federal Reserve, right,that makes that available. And
it's at FRED.StLouis, tell, tellus where people go to accesses
if they haven't before.
Maria Arias (04:41):
Yes, FRED is a data aggregator, which means that we
get data from public websiteslike government institutions,
international organizations, andthen some private institutions
and academic resources. And wekeep that up to date in our in
our website, and so the way thatthis works is we're putting all
(05:02):
of this data in one place forusers to be able to combine and
compare different time seriesfrom different sources. So, for
example, if you want to compareemployment growth and economic
growth using parallel employmentand the Gross Domestic Product,
those are, both of those seriesare produced by different
(05:25):
government agencies, the Bureauof Labor Statistics and the
Bureau of Economic Analysis. Butin FRED without leaving our
website, you can plot both ofthose in a graph very quickly,
and then compare themimmediately. And again, for
everyone who has not visitedFRED yet. It's that
FRED.stlouisfed.org.
Sean O'Toole (05:45):
Perfect. Yeah, that's great. One of the other
things that's, you know, youknow, really impressive to me
about it. And it's not, itwasn't perfect, right,,like so.
And maybe you could talk aboutthis, but you're getting this
data from all these differentsources, right, and now you're
gonna lay it on a chart andcompare it. And that presents,
as a data guy, that presentssome real challenges, right? Are
(06:05):
you getting the same timeslices, you know, quarterly
versus monthly versus annually.
And how much of your work goesinto just getting these things
to align so that they, they workon a chart together.
Yvetta Fortova (06:19):
So, over the years, we developed a process
where the notion of the timeseries is well defined in our
back end. And really, it's alldependent on the frequency of
the data. So, you can have data,like interest rates, there's
your interest rates, who arepublished, with a daily cadence,
(06:40):
and the data is on a dailyfrequency. And then you can have
data like labor markets, whichare published by Bureau of Labor
Statistics, and those data aremonthly. And really, when all of
this gets put together on ourwebsite, users can combine these
data together on a graph. Andthere is a kind of like a magic
(07:04):
in the, in the, in the in, thebag that happens that is able to
translate that this data, whileit's the treasure data, while
it's daily, and the labor data,which is monthly can end up on
one graph. And we have a highlyinteractive graphs, so users can
see their values and downloadthe data from the graph. And
(07:26):
really, that is, what makes FREDvery unique is the fact that we
are trying to provide this dataservice to public and, and
recognize the value in beingable to give up to date
information to our users.
Maria Arias (07:44):
Right. And just to add a little bit on to that some
of our tools that we provide, aswell as, while you can add
multiple series to the graph,you can also edit the graph
right then and there. So, forexample, if you want to compare,
like you mentioned quarterly andmonthly data, you can convert
the monthly data to a quarterlyfrequency, as well. And so
that's all the calculations areall done in the background. And
(08:06):
then you can pick if you wantlike an average for those
monthly values converted to aquarterly frequency, or if you
want a different type ofaggregation, things like that.
So, it can really allow you tocreate a customized graphs for
whatever your needs are, orwhatever data you're interested
in.
Aaron Norris (08:24):
How did this, how did this line at the St. Louis
Federal Reserve, or the otherbanks doing something similar,
or they're just like St. Louishas got this.
Yvetta Fortova (08:33):
So, FRED originated in St. Louis, and we
are part of a researchdepartment. And as, Maria kind
of diluted that, really the inthe 60s, the big vision for
Homer Jones, who was a, who wasa research director, at that
time, he really liked the visionof allowing data to be available
(08:59):
to public and really, thisproduct has since been
organically evolving over theyears with multiple other side
products and creating this FRED,FRED-family of products. So,
FRED is the mainstream of thedata but then we have a mappable
(09:19):
data in the in a GeoFRED andthen we have also real time data
in ALFRED so we're really tryingto capture as much as much
pieces and angles of the lookingat the data as possible.
Sean O'Toole (09:36):
Somehow I completely missed GeoFRED like
I've been a FRED user for Ican't tell you how many
presentations of mine have atleast one FRED chart in it, if
not quite a few. So um, but Icompletely miss GeoFRED. I just
found it this morning. And Ionly got a chance to play with
it after Aaron mentioned it. Andso, I don't know how I possibly
(10:01):
missed that. How long is thatwe've been around?
Yvetta Fortova (10:05):
Quite a bit. We, and what GeoFRED is just a way
for users to see cross sectionalcomparisons. So, it's just a way
of looking at the data on a map.
So, in comparing the States inthe US, for example, and if,
you're right that it wasdeveloped on a side, and maybe
(10:26):
it was not as well, as wellprovided to our users, but we
have been trying to, over theyears trying to incorporate the
maps on the FRED website in abetter way. So, nowadays, if
users would be able to, would beable to see View Map button next
(10:49):
to their graphs, that's anindication that they can also
examine and then analyze thedata on a map. And in addition,
we also provide globes, imagesof globes below the graph, in
the related content sectionwhere users can also see data in
GeoFRED.
Sean O'Toole (11:10):
Okay, I was trying to do I'm sorry, Aaron, you go
ahead.
Aaron Norris (11:14):
Visualizations are just very powerful. So, I have a
lot of playing to do. I too,have been a long term user of
FRED and as pulling togetherdifferent data from different
governmental agencies has beenvery tricky. And then I found
FRED and it was a godsend. Ican't tell you how much time you
guys have saved me over theyears.
Sean O'Toole (11:34):
We're total fan boys just to be clear. We're
definitely your fanboys.
Aaron Norris (11:39):
Sean, I already asked Maria. I'm like, Where can
we get those sweatshirts?
Sean O'Toole (11:43):
I know, I totally want one.
Maria Arias (11:46):
Yeah. The FRED swag is available at the St. Louis,
St. Louis Feds, the economymuseum. Unfortunately, it's
closed right now. But as soon asit open, we're expecting to have
an influx of FRED fans coming toget some FRED swag.
Aaron Norris (12:03):
How many how big is your team, like when you're
dealing with 800,000 timeseries? What kind of manpower
does that take?
Yvetta Fortova (12:14):
Our team is, is relatively small, we have a team
of about 10 people. And theyconsist of the developers who
are working on making sure ourwebsite functions properly and
developing new features and newtools to FRED. And then the
other part of our team, our dataengineers who specialize in
(12:36):
making sure our data and contentin FRED is, is up to date. And
because FRED is so popular, wealso have a lot of help outside
of our team within the researchdepartment to help us with other
other things related to dataand, and content.
Aaron Norris (13:01):
Do you have any idea how many users are on your
site any, any specific time ofyear?
Yvetta Fortova (13:10):
Guess, our users fluctuate. And we have kind of
like a seasonality in our, inour user, in our users. And that
is, that is correlated a lotthat education, education and
semesters at schools because oneof our core users are academia,
(13:33):
professors and students whoutilize our product to learn
about economy, money andbanking. FRED is in a textbooks
for students to learn about dataand do homework. But overall, we
do have over over millions ofusers annually that come to our
(13:56):
sites from many countries, andare hungry for data and for
information about data and whatand graphs.
Maria Arias (14:08):
Yeah, something that we think is really cool
about FRED users is at any pointin time, we could have a person
who is just now learning abouteconomics, it's their first time
looking at a graph, as well as aPhD economist or even Nobel
Prize winning economist on thewebsite. So, again, we try to
make sure that our features areeasy to use, the data is
(14:30):
accessible for everyone to learnabout, but also that we have
tools that more advanced userscan take advantage of and also,
you know, like everyone inbetween. So, we're, again really
happy to to hear when peoplereally like our product, and we
try to keep developing newfeatures for you guys out there
(14:52):
who are using it to collect thedifferent data that you need.
Sean O'Toole (14:57):
I personally think it should be you know, there's a
whole class on it in highschool, maybe elementary school,
right? Like for every personbefore they're allowed to
graduate before they're allowedto vote, like they should know
how to use FRED, like, it wouldclean up half the crazy ideas
that I see on, on Facebook andTwitter, and all the wrongheaded
(15:19):
ideas out there. Because it'sjust such, you know, you can ask
these questions rather than, youknow, guess about them. And it's
just, it's really an amazingresource.
Maria Arias (15:28):
Right, even thinking one step back, we, one
of our core missions is topromote data literacy. And
that's something that is veryimportant to us. So, just going
back to understanding, you know,what is the frequency of this
data? What is the ,what are theunits, you know, how can you
compare two different series,they have to be in the same
(15:49):
units in order for you tocompare apples to apples and not
apples to oranges. So, we workvery closely with another group
in the research department. Thatis the Economic Education Group,
and they create a lot ofdifferent materials from
kindergarten all the way to highschool and college level
educational resources for, forstudents out there and teaching
(16:13):
assistant teaching materials aswell.
Aaron Norris (16:17):
Really?
Sean O'Toole (16:17):
That's amazing. I didn't know that, that's, that's
great to hear.
Aaron Norris (16:20):
I'm gonna have to find links to that. That sounds
really cool. Kindergarten?
Maria Arias (16:24):
Yes, down to kindergarten, there are some,
actually there stories that areread out loud. And so, you could
just sit your kid in front of atablet and play this story that
teaches them a lesson aboutfinance, or education or
economics, things like that, andyou find resources at
econlowdown.org.
Sean O'Toole (16:44):
Might have to start with that kindergarten,
one for my 18 year old.
Aaron Norris (16:49):
And build your way up.That's amazing.
Sean O'Toole (16:50):
Build his way up?
Yes. He's actually pretty good.
He's had to put up with me ashis dad. So, he's not that bad.
Aaron Norris (16:56):
Now, somehow, I don't think he's gotten away
with scot-free for sure.
Sean O'Toole (17:00):
No, yeah.
Aaron Norris (17:01):
Maria, you have to deal with a lot on the on the
back end. And I really wantpeople to appreciate what it
means on the data side. So, whatis your day to day look like
when data is coming into thesystem? How does that work?
Maria Arias (17:13):
Right. So, for our process, currently, a lot of it
is automated. And so, we havedeveloped a Python-based
pipeline where we, when the datacomes out, we reach out to the
source's website, collect thatdata and process it and save it
onto our database. And at thatpoint, it's available
immediately on our website. Butwe have a set of like
(17:37):
notifications and things likethat throughout, the throughout
the day with these processes,making sure that everything is
the data is updatingsuccessfully. And if there are
issues, we would go in and checkeach like specific data set or
release to see what what ishappening. And then this is kind
of for, like maybe 90% of ourdata is automated, updated in an
(18:02):
automated way. And then we also,for some, like for the remainder
of that we either manually haveto go and check to see if the
data has updated on the source'swebsite. Or maybe we wait to
receive an email fromsubscription, that the data is
available, things like that. So,there are a few kind of outliers
(18:23):
out there that involve some sortof manual intervention to get
the data updated. But kind ofeven before we get to that
stuff, we as we add new data toFRED, we have to add them to
this data pipeline. So, thatinvolves us, as data engineers
understanding the data and howthe data is provided by this
(18:44):
particular source or theparticular data set that we're
adding, so that we can processthat data and kind of
standardize the data so that wecan save it to a database.
Aaron Norris (18:56):
Who decides what data gets to be featured in the
app?
Yvetta Fortova (19:01):
So, for us, we have lot a long list of data
requests that comes from variousplaces, our users, the data
providers themselves, who wouldlike their data added to FRED,
and as we mentioned, our team isrelatively lean. So, what we do
(19:22):
is we have internally have acommittee that meets on a
regular basis. And then we, weselect data that we think are
relevant to our users, and theyare reliable as well from the
originating sources. So, thatwe, we can continuously provide
(19:48):
new content to users and and wecan, we can, we can give our
users additional data to workwith. So, for example, recently,
we have added weekly pandemicunemployment claims, which has
been developed by Department ofLabor. And we have also worked
(20:15):
on addition of housing datafrom, from Optimal Blue and
realtor.com, which are alsoavailable on our website. So,
anybody who's interested, can goand find them.
Sean O'Toole (20:29):
I saw one on GeoFRED it was market hotness.
That seemed to be a real estaterelated.
Maria Arias (20:38):
Yes, but as a data set from realtor.com, as well,
they have a set of indicatorsthat they define as market
hotness indicators. So, it's adifferent different topics that
kind of like how many days ahouse has been or like, average
(20:59):
of how many days houses havebeen on the market in this
region, things like that. Andso, that's a data set that they
put together and we publish inFRED.
Sean O'Toole (21:08):
On GeoFRED, it had very few counties populated from
what I saw, yeah. Is that stilljust a work in progress? Is that
the reason or did they just nothave data on most of the
country? Or do you have anyideas there?
Maria Arias (21:19):
Yeah, so, that's a we have their complete dataset
in FRED for the market hotnessdata, they have the only the
largest. I don't know how many Ithink it's the largest, like 100
MSA. And then I think thecounties that are included in
that data set are from those MSAand surrounding areas as well.
So, it's not a county by countyfor the US for this particular
(21:43):
data set.
Sean O'Toole (21:44):
I happen to pull down County, if I pulled down
MSA, it probably would havefilled in a little nicer for
those MSAs versus County. It'slike only when the MSA maps to
the county, would it work?
Right?
Maria Arias (21:56):
So yeah, what depends on what data they have
available for that particulardata set. And yeah, I know that
for that one. In particular,it's just the certain number of
the largest MSA and then thosesurrounding counties.
Sean O'Toole (22:10):
We have a lot of data users. And I want to come
back to your your data pipeline,right. So, in that, in, for that
data pipeline is, you know, I'veused obviously, the website and
the graphical tools, right?
Folks that want to get access toraw data, is your data pipeline,
open source? Do you guys have anAPI to the raw data? What are
(22:33):
the, is there any options there,they've got to go do all the
same hard work you did.
Maria Arias (22:40):
Right. So, the work that we do, hopefully will make
everyone else's work a loteasier. We are doing the hard
work of going to, you know, theCensus and the BLS and all of
these other organizations andparsing their data that is
provided in different formats.
And putting it all in like onestandardized format in FRED. And
so, for all of you who want tomaybe automate your workflows or
(23:05):
things like that, we also have,we have a FRED API. And we also
have an add in for Excel thatyou could use. So, that allows
you to, again, automate yourworkflows in different ways and
combine it with the other datathat you already have. So, if
you like programming for theAPI, in particular, you just
(23:25):
have to create a free useraccount and request an API key.
And there are third party kindof wrappers or packages that are
already out there that otherpeople have created. And they
just kind of wrap around theFRED API and make it easier to
use. And then that's in additionto you know, if you go to the
FRED website, you can downloadthe data as a CSV or an Excel
(23:48):
file. And even if you've addedmultiple lines to the graph, if
you download the data from thatgraph, it'll include all of the
data that you have. So, that'salso another way that, that you
could access the raw data. Andthen
Sean O'Toole (24:04):
All the above love it.
Maria Arias (24:06):
Awesome. And then you have for the GeoFRED data as
well. If you want to downloadall of the say, like all of the
county level data for the entireUS, that's probably the best way
to download all of that data isthrough GeoFRED for that
particular map.
Sean O'Toole (24:21):
And all the data itself, basically at least is
provided from you is opensource, and people can use it in
their applications and that kindof stuff for their license
limitations at all?
Yvetta Fortova (24:33):
So, most of the data we have come from
governments or sources orinternational organizations,
and.
Sean O'Toole (24:43):
Public record.
Yvetta Fortova (24:44):
Public domain data, correct. But we do have
small percentage of data thatcomes with copyrights. And users
will be able to identify thosedata because we have a note
under the graph that specifieswhat the copyright, the
restrictions are. And we alsohave in a notes, what that means
(25:07):
to the end user. And if theyneed any additional actions to
utilize the data, but overall,from the whole scheme of data
you have in front, it's a smallpercentage of data.
Aaron Norris (25:23):
I like to, Yvetta what you said about data telling
stories. And part of that is,what I like about data is
there's a trust factor to it.
It's, it's just the data, thedata is the data, data doesn't
get political, it tells a story.
How important is the datavetting process, when you
consider new partners? I was, Ihad no idea that you were
pulling in like a NationalAssociation of Realtors, which
(25:45):
is an association group. Whatdoes that data vetting cycle
look like?
Yvetta Fortova (25:53):
When we are looking on, where like, what
kind of sources, the data comefrom? We are we are looking on,
how much data would be availableto us? And also how, what the
data look like how is the dataconstructed? If it's data that
is academic data, we are lookingfor a ways to see whether there
(26:20):
is some peer review work and orpaper that has been published
for construction of the data?
And also, if there are ways thatthis data, how is the data
disseminated all in all. So, inwhat way we, as a team can
maintain the data, which is alsoa big piece. Because we are
trying to look for ways of, ongetting the data in an automated
(26:45):
way. And cut down on the manual,manual work. So, there are
pieces of of that. Andobviously, it also comes down
with private sources to havingproper licensing agreements in
place.
Aaron Norris (27:05):
I know you've been on the research team before. So,
do the data sources come in andthen your research team gets a
list of all the new data sourcesand they dream up what's next?
What triggers some of theresearch that your team does?
Yvetta Fortova (27:20):
So, without our, within our team we do like our
team does not do research. We'rejust solely doing breathing and
living from FRED.
Aaron Norris (27:30):
Yeah.
Yvetta Fortova (27:32):
But the economists who are in the
research department, they, theypretty much have, have really a
choice in whatever they wouldlike to publish. And, and and
write about. And their process.
I'm not familiar with theirprocess in details, but.
Aaron Norris (27:58):
That's okay. I was just curious, like, I can't
imagine having access to 800,000data series, and you just wake
up in the morning like, 'Huh,what comedy work I got to put
together today?' That's amazing.
Sean O'Toole (28:10):
Do either of you have any favorite data series?
Maria Arias (28:16):
It's a, it's hard to pick with so many of them.
But I didn't have a...
Sean O'Toole (28:22):
To look at in the morning or when the new, when
the new drop comes that like,'Oh, that's the one I want to
see.'
Yvetta Fortova (28:29):
Yeah, we've definitely become kind of a data
nerds over the years workingwith the data. And there, some,
some data are definitelyinteresting to talk about. For
example, we do have data in FREDthat is the longest series that
(28:54):
we publish, and it's ahistorical data for population
of England, that goes back downback to 1086. So, it's really
depicting a plague that has,that impacted population. And
then you have some reallyinteresting data on orders of
(29:18):
sinks, kitchen sinks and toiletsduring industrialization period,
and the, the peak in indoorplumbing that has happened
during that time. So, really,those are, those are our kind of
like, nice dimensions. Butreally, when it comes to
(29:38):
favorites, it changes all thetime because we, data is really
interesting, when there aretrends and when there are
changes in the data as you asyou deliver it to the to just
spikes and drops in data. So,really looking at business
cycles and seeing how the datais changes fit the business
(30:01):
cycles or after the businesscycles is, I think the most, the
most interesting part oflooking, looking analyzing the
data.
Aaron Norris (30:10):
Data is such a hard business too and I'm sure,
as you've made more available, alot of people are very
demanding, I think one of myfavorite series has always been
migration. And especially withCOVID, you hear a lot of,
there's been a lot of talk inthe real estate space and
different data series to tracknot having to wait for a census,
are you guys getting a lot ofpressure to provide more series
(30:32):
that are a lot more timely,instead of waiting for
government, census and things ofthat nature.
Maria Arias (30:39):
We've seen a lot more demand for more, not not
only more timely series, butalso data that are more frequent
or that have like a higherfrequency. So, say daily or
weekly data instead of likequarterly or annual data. And
so, especially this year, when,you know, if you think about GDP
(31:02):
and the state of the economy,you have to wait for the quarter
to be finished. And then onemore month for the initial
release of the previous quartersGDP to come out, right and ask
like things are changing in thecountry, it's hard to tell what
the state of the economy is atthat point. So, we've definitely
seen an increase in demand.
Unfortunately, we're also kindof bound by, you know, we want
(31:23):
to continue to provide highquality data to users. And we
can't just add everything rightaway either. And so, it's, it's
been kind of a, trying to seewhat the data providers are
creating for like new releases,for example, the census created
a, they created a experimentaldata set from their quarterly
(31:50):
business formation data, theystarted calculating weekly
business formation data. And assoon as they started making that
available, we added that to FREDin the summer or early summer,
late spring, something likethat. And so, now there's this
weekly level, business formationstatistics that has been really,
really popular.
Sean O'Toole (32:12):
Is there what's the Geo-granularity on that?
Maria Arias (32:15):
Um, that is available at the state level?
I'm not sure if it's alsoavailable at the MSA level.
Aaron Norris (32:22):
It's so, funny, Sean. We were just talking about
that with Doug Duncan.
Sean O'Toole (32:25):
Doug Duncan, Chief Economist for Fannie Mae was
just, he was just saying that'sthe one data set he would most
like to have right now, butreally, probably more down at
the county or MSA level? I'mgona have to dive into that one,
for sure. And we'll have toshare it with Doug as well.
(32:46):
Yeah, so, out of you, you know,you mentioned 800,000 time
series, like, give us, give ussome idea, you know, a guess is
fine. But like, what's thefastest data set you're getting?
You're not like a Bloombergterminal where it's, you know,
sub, sub second updates on onstocks or anything like that.
But the fastest is maybe daily,hourly? And then the worst is,
(33:10):
is annual or is it some of themeven less often than that?
Yvetta Fortova (33:17):
Oh, are you? Are you referring to how quickly we
can get the data to FRED? Orwhat kind of data we have?
Sean O'Toole (33:23):
I think there's two separate questions there.
And let's do both. Let's do thefirst one is how often, right?
So, out of your 800,000? Howmany of those are annual versus
quarterly versus monthly versusdaily versus hourly, right? Just
a guess. And then, and then, youknow, she mentioned, Maria
mentioned the, you know, some ofthese things have a month delay.
(33:45):
And so, both of those things arereally interesting issues,
right, like, that I don't thinkpeople think enough about we
have to think about it a lot,because we provide public
records data on real estate,right. And in a small county, we
might go get the data everyother month, right. And in LA
County, which is the largest inthe United States, right? We
(34:07):
have stuff, you know, very, veryquickly, so we get it daily.
And, you know, we usually haveit published to our site within
just a couple of days, versusAlpine county might get it every
two months still publish to oursite within a couple of days,
but there's big differencesthere. So, can you speak to that
kind of distribution among yourdatasets?
Yvetta Fortova (34:30):
The data we have in FRED are anywhere from daily,
weekly, monthly, quarterly,semi-annual annual, and then we
have some five year frequencydata. And really the
distribution among what kind ofdata we have in FRED is more or
less, we have mostly, most datafor on annual frequency and then
(34:54):
it it kind of goes down fromannual to quarterly, monthly And
daily. So, that is kind of likethe distribution of the data we
have in FRED. But then there arealso aspects of how, how, how
the data updates. So, when we doour updates, let's say that an
(35:16):
employment data, the employmentsituation that is published by
the Bureau of Labor Statisticsis published on the first Friday
of every month. And it comes outaround the 7:30 Central time.
So, we really are trying to hitthe, the update process,
(35:38):
immediately or minutes after thedata is released. And then
obviously depends on the bit onthe size of the data itself. So,
if employment situation hascouple 1000 series, it takes
some time to get all the newinformation into FRED. But we
are trying to optimize ourprocess to really get into it in
(36:01):
minutes or from what the actualrelease of the data happens. And
then there are also aspects ofthe data itself. So, when the
data is released, there arethere is the lack of the
information, because whathappens is, first Friday of the
month, there is a publishpublication of the data. But
(36:24):
actually, the data that'spublished represents the values
for the previous month. And thenthat goes all the all the way
back down to the origin eatingagencies who are working on the,
on obtaining the underlyingdata, and perhaps the micro
data. So, they need time to geta good size sample of the data,
(36:49):
which then allows them to modelthe data and make the aggregated
values available. So, those inmost data, if we have in FRED of
work on on this typical lackwhere you're not going to see
data for December, monthly datafor December, if the December
(37:12):
isn't over yet. And again, itcomes back down to, you know,
the originating agencies havingtime to prepare those estimates.
On the other hand, we also havea little bit of advanced data or
data that is forecasting thatare looking into the future.
(37:35):
Which users can also also use toto look at what the data will
look like. And last not butleast because all of this
process of disseminating thosedata and publishing data is
really based on a lot of timeson surveys and on incomplete
(37:56):
samples. That's why agenciesover time, as they are
publishing new values, newmonthly data, let's say they are
also going back and revisingexisting data. And for that
reason, we also have ALFRED,which is an archival, FRED.
Aaron Norris (38:16):
Interesting.
Sean O'Toole (38:19):
Yeah, And that was Alfred was another one I wasn't
familiar with. And, you know,Aaron mentioned this morning
when we were doing our pregame,you know, prep. And, you know,
he mentioned, he had a dataseries that changed quite
dramatically, you know, and oneof the things that he was
following and, and he was ableto find that the historical data
(38:40):
and ALFRED and see why, youknow, see that difference. So,
that's, that's pretty cool thatthe archival piece too,
valuable.
Yvetta Fortova (38:49):
Yes, I think the the best example that we like to
give is from gross domesticproduct. So, if you think about
gross domestic product, the dataitself, is a quarterly frequency
data, but then the values areannualized. And then the data
(39:10):
updates monthly, and then thevalues to revise could revise
every month. And on top of thatthere, there is an annual
revision that happens every yearin summer. And on top of that,
every three to five years,there's a benchmark revision
which may change everythingunder the GDP.
Aaron Norris (39:32):
I did not know that, wow.
Sean O'Toole (39:33):
GDP doesn't mean anything. It's just a random
guess. Just kidding.
Aaron Norris (39:40):
I.. do you find that your work has inspired
states to get a little bit moreserious about collecting data. I
noticed just in California as anexample in the last year, I've
noticed a lot of data collectionsort of budget line items in
legislation talking aboutdigitizing and starting to track
(40:00):
things. And can you point tosome of your work being so
helpful that states are like wegot an upper data game?
Maria Arias (40:08):
We would like to think that we inspire other
government and publicinstitutions to improve our data
collection. In some cases, whereespecially like smaller
government institutions havereached out to us about, you
know, having their their dataand FRED and it's not
necessarily available in amachine readable form, we've
(40:31):
been able to work with, withthem, and even some of the
academic data that we've added,that seems to be pretty popular.
And FRED, we've been able towork with them to help them put
their date make their dataavailable on their website in a,
an easier form for us to parsebut also for other people. So,
(40:52):
like machine readable forums,standardized tables, things like
that. And so, we have seen alittle bit of that. But we don't
work with like, we don't workwith every government
institution out there that wecollect the data from, because
again, all of this data ispublicly available. So, we just
(41:12):
collected from their website.
But we would really hope thatmore and more, especially
smaller government entities, tocollect their data and make
their data publicly available inan easy format, because it's not
only for us to provide a servicefor other users, but also, you
know, everyone around them, andin your communities, you know,
(41:37):
you know, what's important toyour community. And you know,
what industries are what drivethe community forward, things
like that. And so, if thegovernment's local governments
were able to help collect thatdata and make that data more
available, it would be easier toalso analyze the state of the
economy and these kind ofsmaller geographical regions,
(41:57):
like you were talking aboutearlier, it's really nice to
have county level data, it wouldbe even nicer to have it at a
smaller geographical level foreveryone who's doing research
and trying to understand what'shappening.
Sean O'Toole (42:14):
Our company, the one that sponsors, the Data
Driven real estate podcast,PropertyRadar, really only
exists, because of how difficultit is to get public records data
on real estate assessors data,recorder data, you know, etc.
And I would, it would, it wouldmake me so happy to shut this
(42:35):
company down, because the datawas directly available. On an
easier format, like out of allthe things that we spend money
on as a government, you hearbillion dollars here a billion
dollars, they're like, why yourgroup and the group's in all
these different levels aren'tfunded to the tune of billions
of dollars? You know, I have tosay, I was a little dismayed to
(42:58):
hear how small your team was,and I kudos for what you guys
accomplished with that team.
But, you know, if you ever needsomebody like you to write a
letter on why you should getmore funding, like you can count
me in.
Aaron Norris (43:15):
Yeah, the actionable side of data I serve
on the county board for 211 outhere, and you know, a lot of
counties will do is to
Sean O'Toole (43:22):
What is 211 Aaron?
Aaron Norris (43:23):
Oh, sorry, 211 is a Health and Human Services
hotline. It's, it's nationwide,it's sort of like the 411 for
help. It's it has a lot ofnonprofit data government
assistance programs. And I thinkwhy I joined was the data is
that you don't have to wait ayear for to spend $100,000 on
what people need in thecommunity. This 211 hotline is
(43:44):
tracking the need. So ,30% ofthe need in our community for
the last three years has beenconsistently been housing, you
don't need to spend $100,000 inwait a year, it's almost
immediate. So, it's, it's datadoes tell a story. And sometimes
it's really important becausegovernment entities can redirect
resources, where it's needed,instead of wasting time, and
(44:04):
you're a year late. So, I'm justcurious, what are some of the
data that you're most excitedabout in the last few years that
you've seen cross your desk thatyou wish people knew about?
Maria Arias (44:20):
So since.
Sean O'Toole (44:20):
800,000, there's too many.
Maria Arias (44:23):
I'm just trying to think of what we've added this
year, um, something that we didthis year, and it's real estate
related, so I definitely wantedto bring it up is the Optimal
Blue mortgage market indices.
So, they have mortgage data thatis broken down by different FICO
scores and by different types ofloans. So that's a, again,
(44:43):
another private institution thatreached out to us to have their
data added to FRED and it's a,it was added this year. But then
we also have some other likeyou've mentioned the weekly
pandemics claim pandemic claimsdata. That was a really
interesting one. We also addeddaily FOMC target rates that
(45:05):
were digitized by one of oureconomists at the St. Louis Fed
and some of his co authors goingback to the early 1900s. So, now
you can get, like FOMC, targetrates starting 1900s. And then
like slowly put the time seriestogether, up until present day.
(45:25):
And then what else? Oh, there'sanother one by the a group of
economists this one and this isgoing back to kind of some of
the changes that we've seen thisyear have more demand for high
frequency data is a group ofeconomists created the weekly
economic index by last names ofLouis Martin stock. And this is
(45:48):
a kind of like a now cast, butnot necessarily an outcast, per
se. But they combine severalhigh frequency data that are
available at the national level,and they create like a GDP
forecast more or less at theweekly level. And that is
(46:09):
actually updated twice in aweek. And so, that's been really
interesting to see as well. Andagain, just kind of making the,
kind of having like a morefrequent update of state of the
economy and keeping thatconversation going.
Aaron Norris (46:33):
Is it I haven't looked yet, HMDA data I know is
a I'm a, I have my mortgagelicense. And the amount of
detail that the CFPB hasmandated that we start
collecting at the loan level isquite expansive, is that
available on FRED yet?
Yvetta Fortova (46:50):
Oh, we do not have humped up data in FRED.
Aaron Norris (46:55):
Okay.
Yvetta Fortova (46:56):
But I would say for anybody who is interested to
learn about what data we haveadded to FRED, we would
recommend to sign up for FREDnewsletter, which is in bottom
of our web pages, and we letusers know what new changes in
(47:16):
data and what new data we editthat way.
Maria Arias (47:18):
A new feature sets...
Sean O'Toole (47:19):
Right there, that's, that's, that's my tip of
the whole podcast, right? Like,because I've been a FRED user
forever, and I'm not signed up.
And I don't know why. So,, I'msigning up right now.
Aaron Norris (47:31):
I'll make sure to post some extra links to to make
sure in the show notes thatpeople can find this stuff
really easy. So, I think in thereal estate space, there's a lot
of like exploring differenteconomies, and maybe some people
they've never approached FREDbefore. And so, maybe we can
give a little bit more industryspecific. And I know this is
your personal opinion, if youwere going to explore a state
(47:52):
that you didn't live in, and youwere interested in, you know,
the economy, real estate, Idon't know what kind of things
would you explore in FRED peoplemight not know exist.
Yvetta Fortova (48:05):
So, definitely, for someone who is new to FRED,
we would say, to go and try tosearch for information and see
what kind of data we haveavailable for, for given states
or MSA's or counties. On theregional level, we'd, and they
will quickly realize that we dohave a lot of major economic
(48:30):
indicators like labor data andgross domestic product. But then
we also have data on propertyprices, like house price
indices, and we have rates onmortgage rates. So, there's
really a wealth of informationavailable. And it's a, it's a
(48:52):
really matter for people to goand try it and see what we have,
you know, from housing stars andconsumer price index, to prices
and wages. And it's kind oflike, when you go shopping, and
you feel like you've reallygood, got a good bargain at the
end of the day. So, that'sreally what we are trying to
(49:15):
give users with FRED.
Maria Arias (49:19):
Right. And then I would just add on to that, that
if you find some graphs that youfind really interesting, or that
you would like to come back to.
This is just kind of a plug forsome very useful FRED features.
And especially if you don't havea FRED user account yet, right
at the bottom of the, of thegraph, you can save graphs to
your user accounts. But then youcan also add graphs to a
(49:41):
dashboard. So, as you add graphsof say, a similar topic that you
are saving to your user account.
You can also put all of thosegraphs or maps onto a dashboard
and add other like snippets,like notes or like a single
number or something like that.
And then all of these graphs canbe saved so that they are
(50:02):
updated automatically. So, thenext time the data comes out,
you just go back to yourdashboard or to your saved
graph, and the data isautomatically updated for that.
So, this is a very usefulfeature that if you're not
familiar with FRED, or if youdon't have a user account, I
highly recommend you check out,especially again, if you have
these kind of several graphs orindicators that you'd like to go
(50:24):
back to. And then together withthat, you can make your
dashboards public and share themwith your colleagues or share
them with other people in theindustry, to just, again,
simplify sharing data andsharing content.
Sean O'Toole (50:42):
But, you know, one thing I haven't looked at, and
I'll just ask is, can you embed?
Can you get like an embed codefor those graphs, so you can
embed them in a website orwhatever, so that they're
constantly updated on your siteas well?
Maria Arias (50:55):
Yes, just below the graph, there are some shared
links, and account tools. So,under the account tools, that's
where you can save those to youraccount. But then the share
links, you can share a URL to agraph. And so, whatever
modifications you've made to thegraph, and you share that, using
the URL provided under thatbutton, the person who accesses
(51:18):
that will see exactly the graphthat you created. And then same
with the embed code, if weprovide like a pre created
snippet that you could just putonto your website and embedded
that way.
Sean O'Toole (51:30):
In my, you know, one of the things I really like
is the, the bars that showrecessions, right, um, and,
excuse me here, if I like, don'tknow, that, in terms of those
bars, that kind of underlieright recessions is, is that the
only one or there are someothers?
Yvetta Fortova (51:51):
We only have recessions at this point on the
graphs. And there are other waysyou can, you can take those
recessions off from the graph,if, if there's no need for them.
But we also have a way tomanipulate the data of it in the
(52:12):
graph itself. So, for example,if you have, if you have a data,
you can then index the data todifferent time of the year, or
different period in time and,and it will set the values to
100 of a given period. And thenyou can see changes to the in
(52:32):
the data. So, that is also oneway how someone can kind of take
the data and, and make it workto whatever they are trying to
save at the data.
Sean O'Toole (52:48):
I want to just expand on that a little bit to
make sure our folks understand,right. So, sometimes you get
these data sets that are wildlydifferent in their values,
right? Instead, in some, youknow, one way to do that in
charts to show one value on thisside, one shot value and this
side. But when you're looking atthe relative change over time,
you guys allow you to set a dateat which you basically normalize
(53:09):
them all. And then you can seehow the change relative to each
other from that, even if theyhave completely different
scales. And that's a superuseful tool. I gotta say, the
one thing that I do a fair bitis, is, is underlie those events
besides recessions, right, like,one of the things that that
(53:31):
folks in the real estatecommunity worry about a lot is
changes in administration, andchanges in the makeup, excuse
me, of the, you know, like thehouse in the Senate. Right. So,
like, everybody obviously isworried right now about this or
thinking about this Senate race,and Georgia. And I'm sorry. And
(53:58):
so, one of the most interestingones I've done over the times is
I go back and kind of like therecession bars. ,I you know,
it's republican versusdemocratic president right, and
then a divided Congress versus,you know, the same Congress and
is it all three? Or is it one intwo like that, like, I would
(54:19):
love some more of those, youknow, different things, that
particular one, you know,administration would be great,
but I'd love to see some morelike that. Because I do think
that is that's a really, youknow, a really cool feature.
Yvetta Fortova (54:35):
Yeah, that sounds really interesting. So
and thank you for thesuggestions, suggestion we can,
we can look on. But what it willlook like and how we can, how we
can approach to have a betteroptions for users and then maybe
have ways for displayingdifferent administration's as
(54:57):
you said, and one more thingthat I also wanted to add is
that if, if it needs to be userscan create draw lines on the
graph, as well. So, there is away to create user defined
lines. So, that is probably theclosest we can get you with the,
(55:18):
with the situation's you'refacing. But definitely, it's a,
it's a place for improvement forus, so thank you.
Sean O'Toole (55:27):
It was an eye opening to me on, on, you know,
what we think about in terms ofwhich party is fiscally
conservative and which oneisn't. But it was not what I
expected.
Aaron Norris (55:37):
And it's good to tell that story in data, it
makes it less political, well,maybe it doesn't. But if you're
a real estate in real estate,and you've never explored some
of the tools, I really liked therecession, because when you go
into an area, and you can see arecession, and how that local
economy, unemployment, householdincome, median price, really
(55:58):
fared through like a recession,it's really interesting to drill
down as you're sort of makingmoves, or you're helping clients
that want to move out of stateand they're exploring a new
area, I just the easiest thingto do is go on the website and
type in the name of a county andjust start playing and see all
the different data series thatare available. It's it's almost
overwhelming. Just have to getin there and explore if Sean, do
(56:22):
you have any other favorite onesthat you use?
Sean O'Toole (56:26):
Oh, you know, the biggest one I use the most often
is I look at the different typesof debt, right? So, mortgage
debt, corporate debt, total, youknow, the the federal debt, you
know, etc. student debt is justa student that that chart is
really scary bad, like, youknow, especially if you take all
(56:48):
of those different types of debtand put them at 100. You know,
say and 2000. And look at howthey change over time, it's, we
clearly have a huge student debtproblem. And I know you guys
probably can't comment on the,on, on those things. But that's,
that's probably one of myfavorites that I look at the
most.
Aaron Norris (57:10):
Well, we are about out of time, I'm trying to think
if there's any other things thatwe should cover, like maybe new
features that you're working on,that you could share.
Yvetta Fortova (57:21):
So, we are working on working on surprises
for 2021. And next year, FRED isgoing to be 30 years old. So,
we're going to be looking insome ways how we can make FRED
more mature looking. But thereare definitely going to be more
(57:44):
data edit throughout next year.
So, that comes up that we triedto do that on a regular basis.
So, so, that we don't loseinterest in you guys. And, and
other other aspects here, we aregoing to try to look at is, we
would like to improve our searchabilities in FRED. And we are
(58:06):
looking into ways how we canexpand on providing users with
some industry break down on thedata.
Aaron Norris (58:19):
Sort of categorized things and in
buckets, or something?
Sean O'Toole (58:23):
I would love I mean, with 800,000, this is a
big ask, but like a littledeeper descriptions on each one.
You know, because a lot of timesyou'll see like, you know,
you'll type something in thatyou're interested in, you know,
inflation or whatever. Andthere's 30 or 40 time series,
and you're like, Oh my gosh,which one of these do I use? I
(58:45):
just wanted this simple thing.
And it's that piece, the number,I mean, the number of series can
sometimes get a littleoverwhelming in terms of which
one should I use. So yeah, any,any suggestions there on how
best to parse that when you getand you're like, Oh my gosh,
there's 30 ways to look atmortgage debt or you know,
whatever, like any anysuggestions on how people can
(59:07):
get to the right one and notmake bad decisions?
Yvetta Fortova (59:13):
Oh, we, one of the things we tried to do and in
our search is to really work themost popular and relevant series
to the top of the searchresults. So, for novice users,
that may be an easy way to, toknow that based on their search,
(59:33):
that the, the series at the topare are going to be the, the
best to use. And also, we try toprovide additional filtering
options on the side of the havethe data and to subset the
searches. But if all else fails,we really and there's really not
(59:56):
a good description in a data. Wereally tried to ask users to see
if they will be able to go backto the originating source and,
and, and see if they have someother descriptions to to make
the decision of which series isbetter.
Sean O'Toole (01:00:16):
I do that a lot I go in. And it's a you know, it's
an amazing chance to learn,right? Because you see others
these different ways of lookingat it. And so, then I'll take
the time to go research thosedifferent ways. And I come out
smarter at the end of the day.
So, it's not it's not all allbad the approach you take, and
Google's your friend on thatright Wikipedia and those things
(01:00:36):
for sure. So great,
Maria Arias (01:00:41):
Right. And part of the filtering options as well is
if you can filter by source. Andso, if you know you want, say
inflation or consumer prices bya particular source, you can
filter by source and that wouldsometimes help you find the
right one or close to whatyou're looking for.
Sean O'Toole (01:01:01):
What else did we not cover that you guys would
like to to share or mentionbefore we wrap up.
Yvetta Fortova (01:01:06):
Um, I would say for anybody who is interested to
do some playing with the data,we didn't mention that we also
have a forecasting game calledFREDcast. And we, if you have a
(01:01:26):
user, have a user account, youalso have access to FREDguests.
And what it is, is, users can bepart of a public leak, and they
can set up their own leaks. Andthey can create forecasts for,
for economic indicators. Theycan forecast unemployment rate,
(01:01:47):
GDP, payroll employments andinflation as a CPI gross, and
every every month, the scoregets calculated, and then you're
ranked among other peopleplaying. And you can see how
good you have scored and whatyour average error is towards
(01:02:12):
other players.
Maria Arias (01:02:13):
You earn public recognition and also earned
badges. So it's pretty fun.
Sean O'Toole (01:02:18):
I'm all about the badge, I was a boyscout, so I
had lots of badges.
Aaron Norris (01:02:22):
That's amazing.
I'm so glad you brought that up.
Maria, do you have anything elseyou'd like to add?
Maria Arias (01:02:29):
Just go in there and search for whether it's your
topic that you're interested, orthe region and the country or
the world that you're interestedin and see what you can find.
Yeah, there's lots of differentways to search for data in FRED,
right under the big search bar,and in the middle, there's also
(01:02:50):
Release Calendar. So, you cansee what data releases are
scheduled to be released todayor this week. And then you can
also search by category. So,that's kind of all of the data
in FRED are grouped intodifferent categories and that's
also a good, a good way ifyou're interested in a
particular topic to kind of godown the different breakdowns of
(01:03:10):
the data.
Aaron Norris (01:03:12):
Fantastic. Well, thank you so much for your time
today. This has been a lot offun.
Maria Arias (01:03:17):
Thank you so much for having us.
Yvetta Fortova (01:03:19):
Thank you.
Sean O'Toole (01:03:20):
Yeah, thank you so much, really appreciate it. Like
I said, we're definitely fans,so keep up the good work.
Aaron Norris (01:03:26):
Thank you for listening to the Data Driven
Real Estate Podcast, you canfind show notes and links to
some of the resources mentionedin the show at
datadrivenrealestate.com. Clickthat join the community, and
you'll be forwarded to thePropertyRadar community where
you can ask questions about thecurrent show and even see
upcoming guests and askquestions there. We'd love to
engage with you in thecommunity. So check it out.
(01:03:46):
Please don't forget to like,favorite, subscribe and share on
your favorite platform whereyou're listening to the show. It
helps us out a great deal.
Thanks for listening, and we'llsee you next week.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com