The Upstart Macro Index (UMI)

WEBVTT

1
00:00:02.919 --> 00:00:07.839
You're listening to Leaders and Lending from
Upstart, a podcast dedicated to helping consumer

2
00:00:07.919 --> 00:00:13.560
lenders grow their programs and improve their
product offerings. Each week, here decision

3
00:00:13.560 --> 00:00:17.640
makers in the finance industry offer insights
into the future of the lending industry,

4
00:00:18.079 --> 00:00:22.839
best practices around digital transformation, and
more. Let's get into the show.

5
00:00:27.039 --> 00:00:31.079
Welcome to Leaders in Lending. I'm
your host, Jeff Keltner. This week's

6
00:00:31.079 --> 00:00:34.799
episode features my conversation with Paul Goo, the co founder and the head of

7
00:00:34.840 --> 00:00:39.759
product and data Upstart. Actually this
is our second conversation with Paul, and

8
00:00:40.039 --> 00:00:43.640
his first episode is actually still our
most listened to episode of all time,

9
00:00:43.960 --> 00:00:48.439
so we're excited to bring Paul back. It is a really interesting conversation delving

10
00:00:48.439 --> 00:00:54.880
into some recent innovations. Upstart is
made around measuring the impact of macroeconomic conditions

11
00:00:54.920 --> 00:01:00.079
on loan performance and really separating out
individual performance of loans from the acro economy's

12
00:01:00.119 --> 00:01:03.280
impact on those in a very quantifiable
way, which I thought was a fascinating

13
00:01:03.480 --> 00:01:07.079
topic. It's a really interesting area
of research, so we kind of dive

14
00:01:07.120 --> 00:01:11.400
into how it works, why,
how that can be utilized in what up

15
00:01:11.439 --> 00:01:15.480
Start is working on in the future. Towards us. It's a really interesting

16
00:01:15.000 --> 00:01:19.439
topic, one obviously highly relevant in
the current environment, and so I appreciated

17
00:01:19.519 --> 00:01:23.959
having a chance to dive deep with
it with Paul. I will say there's

18
00:01:23.959 --> 00:01:26.760
one correction to make upfront, which
is that Paul mentions what he called the

19
00:01:26.799 --> 00:01:33.359
Upstart Macro Forecast or UMF, and
the correct term is Upstart Macro Factor still

20
00:01:33.480 --> 00:01:37.000
UMF. So just a quick correction
and without further ado, please enjoy this

21
00:01:37.040 --> 00:01:41.840
conversation with Paul Goo. Oh,
welcome back to the podcast. Thanks for

22
00:01:41.920 --> 00:01:45.959
joining us again. Oh, I'm
happy to be back. It's grown quite

23
00:01:45.959 --> 00:01:48.879
a bit since I was last year. It is and yours is one of

24
00:01:48.879 --> 00:01:52.879
the most popular episodes the original that
we've ever had. So apparently alreadience likes

25
00:01:52.879 --> 00:01:57.159
technical discussions. We talked about AI
and lending in a more technical way last

26
00:01:57.159 --> 00:01:59.239
time, so we're gonna try and
give them a little bit more technical stuff.

27
00:02:00.120 --> 00:02:02.920
Great, let's do it well.
The topic of discussion today that I

28
00:02:02.959 --> 00:02:07.719
really wanted to dive into is the
topic of what we call the Upstart Macro

29
00:02:07.840 --> 00:02:10.759
Index or u am I, because
that's been I know, we've talked about

30
00:02:10.759 --> 00:02:14.439
it a bunch and earnings we've talked
about it publicly. We've talked about it

31
00:02:14.439 --> 00:02:16.800
with our partners, but we haven't
disclosed a lot about it or gone in

32
00:02:16.879 --> 00:02:20.039
depth, and so ill I know
some people have gotten one on one,

33
00:02:20.439 --> 00:02:23.199
you know, introductions or explanations,
but I would to spend the day kind

34
00:02:23.199 --> 00:02:24.919
of diving into it. So maybe
the best way to just start is like,

35
00:02:25.520 --> 00:02:30.919
what is you you am I?
Yeah, you AM I stands for

36
00:02:30.080 --> 00:02:36.319
Upstart Macro Index. And before I
describe what it is, I might take

37
00:02:36.319 --> 00:02:38.840
a step back and just describe how
we got here and why we started working

38
00:02:38.879 --> 00:02:44.120
on this thing that it became known
as the u AM I. So if

39
00:02:44.680 --> 00:02:47.439
you if you take a step back
and look at where we were in call

40
00:02:47.520 --> 00:02:55.759
it mid twenty twenty one. At
that time, very loosely speaking, almost

41
00:02:57.000 --> 00:03:00.840
everybody was doing a good job of
repaying their loads. Obviously not everybody.

42
00:03:00.919 --> 00:03:06.319
There were always defaults, but repayment
rates were extremely high compared to historical norms,

43
00:03:06.879 --> 00:03:10.000
and it had looked like at the
time everybody was a genius in lending.

44
00:03:10.560 --> 00:03:15.840
You could have lent to pretty much
any type of consumer and probably you're

45
00:03:15.840 --> 00:03:21.840
outperforming your expectations in terms of repayment
rates and therefore rates of return on debt

46
00:03:21.840 --> 00:03:25.879
capital deployed. Then you fast forward
another year to mid twenty twenty two,

47
00:03:27.120 --> 00:03:32.560
and the story had flipped. Instead
of every type of borrower outperforming, now

48
00:03:32.599 --> 00:03:39.080
you had almost every type of borrower
underperforming. And the change from mid twenty

49
00:03:39.120 --> 00:03:44.439
twenty one to mid twenty twenty two
was dramatic, and you could see that

50
00:03:44.479 --> 00:03:49.080
in a variety of metrics, but
it wasn't totally clear how you would precisely

51
00:03:49.120 --> 00:03:53.840
tease that out and quantify what had
happened in the quote unquote macro environment from

52
00:03:53.879 --> 00:03:57.919
mid twenty one to mid twenty two. And we started to have conversations with

53
00:03:58.039 --> 00:04:01.960
a lot of our lending partners of
the forum, Well, what's going on?

54
00:04:02.080 --> 00:04:04.919
I you know, I all I
see is that delinquency rates have gone

55
00:04:04.960 --> 00:04:09.319
through the roof. I don't know
if it's because the up Start decided that,

56
00:04:10.120 --> 00:04:13.240
you know, has upstarts model has
gone bad? Has Upstart just sort

57
00:04:13.240 --> 00:04:17.720
of open the floodgates to low quality
borrowers? What you know, servicing degraded?

58
00:04:17.800 --> 00:04:21.199
What what in the world is going
on? And when you see a

59
00:04:21.399 --> 00:04:26.199
very dramatic change two x three x, they weren't talking large multiple in a

60
00:04:26.240 --> 00:04:29.959
short period of time. People naturally
get worried and they want to know,

61
00:04:30.000 --> 00:04:32.600
and of course we wanted to know
too, and we had a pretty good

62
00:04:32.600 --> 00:04:38.000
sense from the indirect metrics that we
we had, of course kept our eye

63
00:04:38.040 --> 00:04:41.319
on, but it wasn't precise in
the way that we would like to have

64
00:04:41.439 --> 00:04:44.439
had it, and so we said, well, we really want a precise

65
00:04:44.600 --> 00:04:49.040
quantitative way to tease apart what we
call the macro and the micro effects of

66
00:04:49.480 --> 00:04:54.920
underwriting. And more precisely, what
I mean by that is, if you

67
00:04:54.959 --> 00:04:59.000
think about what an underwriting model does, the one hand, it has to

68
00:04:59.040 --> 00:05:02.519
separate risk, meaning it has to
say this particular applicant is more likely to

69
00:05:02.560 --> 00:05:06.199
default than this other one, and
they're more likely to default by two times.

70
00:05:06.839 --> 00:05:10.040
But another thing it has to do
is it has to figure out the

71
00:05:10.079 --> 00:05:15.399
absolute levels of default risk. So
that first baseline borrower, what does a

72
00:05:15.560 --> 00:05:18.079
nearly risk free borrower look like,
what does a super risky borrower look like?

73
00:05:18.120 --> 00:05:21.120
And what is the average population and
default rate going to look like?

74
00:05:23.279 --> 00:05:28.040
And the reason that those are such
separate problems is that for that second problem,

75
00:05:28.120 --> 00:05:34.040
you really have in some sense very
limited amounts of historical data to train

76
00:05:34.120 --> 00:05:38.040
on in the way that you would
like to train an mL model. Whereas

77
00:05:38.079 --> 00:05:42.319
for the first problem, the risk
separation problem, you have an enormous amount

78
00:05:42.319 --> 00:05:46.079
of data that you could look at. Basically, every single default that's ever

79
00:05:46.120 --> 00:05:48.519
happened is mostly a story of relative
default. It tells you that, within

80
00:05:48.560 --> 00:05:54.839
a given period of time, how
likely was a particular borrower characteristic to default

81
00:05:54.879 --> 00:05:59.879
compared to some other borrower characteristic.
Whereas to look at the average level of

82
00:06:00.000 --> 00:06:01.800
default in say a year, Well, you only have so many years on

83
00:06:01.920 --> 00:06:06.839
record, and you certainly don't have
the thousands or millions of data points that

84
00:06:06.839 --> 00:06:10.560
you would like to have. I
mean, human history barely goes back that

85
00:06:10.639 --> 00:06:13.839
far. And certainly we weren't recording
our loan repayments, you know, going

86
00:06:13.839 --> 00:06:16.279
back to those days. So that
makes it very challenging. And so we

87
00:06:16.319 --> 00:06:21.240
sort of have are just like a
handful of meaningfully different periods recessions and non

88
00:06:21.279 --> 00:06:24.879
recessions. You can think of it
as like, but maybe like twenty data

89
00:06:24.920 --> 00:06:28.279
points in recorded history, and twenty
is just not enough for any kind of

90
00:06:28.319 --> 00:06:33.000
serious machine learning training problem. And
so we separated this sort of micro question

91
00:06:33.079 --> 00:06:36.759
and this macro question, and we
said, okay, how can we just

92
00:06:36.920 --> 00:06:44.519
precisely measure the macro component and know
whether a period of time is seeing a

93
00:06:44.920 --> 00:06:49.079
increase or decrease and of how much
compared to pass periods in its average level

94
00:06:49.199 --> 00:06:54.879
of default across the population. So
that's that's that's what we wanted to know,

95
00:06:54.959 --> 00:06:59.079
and about eighteen months ago we started
pouring a lot of our a lot

96
00:06:59.079 --> 00:07:02.920
of our research science attention to this
problem and the end result is something that

97
00:07:03.040 --> 00:07:08.560
today we call the u am I
the Upstart Macro Index, Yeah, the

98
00:07:08.600 --> 00:07:14.839
Upstart Macro Index. Essentially, it
takes a It aims to tell you if

99
00:07:14.879 --> 00:07:19.839
you had constant borrowers, if you're
just constantly making loans to the exact same

100
00:07:19.959 --> 00:07:26.519
group of people across different periods of
time, what would the delta in their

101
00:07:26.879 --> 00:07:32.319
delinquency rates be in different days,
weeks, months, years, whatever time

102
00:07:32.399 --> 00:07:35.920
frame you want to compare. And
the way it does this is by making

103
00:07:36.040 --> 00:07:40.360
use of what we call a reference
model, because of course you can't actually

104
00:07:40.439 --> 00:07:46.279
find exactly identical borrowers every single year
to make loans too, and no matter

105
00:07:46.279 --> 00:07:48.480
how you control it with one variable
two variable, you might say, well,

106
00:07:48.639 --> 00:07:53.720
let's just look at you borrowers who
are six eighty to seven twenty on

107
00:07:53.759 --> 00:07:56.519
the FICO scale, and that would
be a very simplified way to do this,

108
00:07:56.560 --> 00:08:00.240
but of course there are a million
other things to know about these or

109
00:08:00.319 --> 00:08:03.240
is other than their phycho score,
And so the way to do that in

110
00:08:03.279 --> 00:08:07.519
a more holistic and continuous way is
with what we call a reference model.

111
00:08:07.560 --> 00:08:13.399
So you say, what if we
underwrote all of these these loans with exactly

112
00:08:13.439 --> 00:08:16.040
the same model, because of course, in reality, every loan you have

113
00:08:16.079 --> 00:08:20.399
on your books was underwritten with probably
a different model at a different period of

114
00:08:20.439 --> 00:08:24.079
time. But if you instead underwrote
them with exactly the same model then made

115
00:08:24.079 --> 00:08:31.480
predictions for every single month that you've
subsequently observed, how do the actual delinquency

116
00:08:31.560 --> 00:08:35.240
rates and default rates compared to what
a constant reference model would have predicted.

117
00:08:35.960 --> 00:08:39.480
And what that tells you the reference
models doing the work of controlling for the

118
00:08:39.480 --> 00:08:43.519
consumer. And then what you can
observe is, of course the actual loss

119
00:08:43.600 --> 00:08:48.639
rates in real calendar months compared to
what the model would have predicted on this

120
00:08:48.679 --> 00:08:54.120
sort of abstracted, hypothetically constant kind
of calendar time period. And so the

121
00:08:54.639 --> 00:09:03.320
deltas in those observations compared to a
sort of now controlled expectation of default or

122
00:09:03.360 --> 00:09:09.000
delinquency is going to tell you how
much more or less risky twenty twenty two

123
00:09:09.039 --> 00:09:11.200
was compared to twenty twenty one,
compared to twenty nineteen, compared to twenty

124
00:09:11.200 --> 00:09:16.000
seventeen. And so that's that's what
we've developed. We call it the Upstart

125
00:09:16.080 --> 00:09:20.759
Macro Index, and it does a
really good job for us of both explaining

126
00:09:20.759 --> 00:09:24.360
how sort of past delinquency patterns have
played out and why they've played out that

127
00:09:24.399 --> 00:09:31.360
way, and also actually in being
able to make forward looking statements of the

128
00:09:31.399 --> 00:09:37.120
forum. Well, if the macro
environment gets fifty percent worse, then this

129
00:09:37.200 --> 00:09:41.200
is what will happen to your loss
rates. And being able to make statements

130
00:09:41.200 --> 00:09:46.840
of that form, of course,
then allows us to make useful statements about

131
00:09:46.840 --> 00:09:52.679
how we should be building in buffer
and conservatism to the underwriting for any given

132
00:09:52.720 --> 00:09:58.039
lending partner, you know how aggressive
or conservative their parameters are set in how

133
00:09:58.120 --> 00:10:03.960
much lending they wanted to. So
a really useful both backwards looking and forwards

134
00:10:03.960 --> 00:10:05.840
looking. So I got a number
of questions to dive off this one of

135
00:10:05.840 --> 00:10:07.720
the things you said, But first
I kind of want to be really clear

136
00:10:09.039 --> 00:10:11.919
when when you produce the index,
it's a it's a singular number for any

137
00:10:11.960 --> 00:10:16.159
given month, as I understand it, what is how should still want to

138
00:10:16.159 --> 00:10:16.919
look at the number and say,
what is this? What does this tell

139
00:10:16.919 --> 00:10:18.840
me? If it's a one point
zero, a two point zero, a

140
00:10:18.960 --> 00:10:22.679
zero point file, like, what
what's the actual index? And how do

141
00:10:22.679 --> 00:10:24.159
I think about reading it? And
then I want to dive into a bunch

142
00:10:24.159 --> 00:10:30.200
of things that you said about this. Yeah, so the index is is

143
00:10:31.600 --> 00:10:35.200
condensed down to a single number,
as you just said, and that single

144
00:10:35.279 --> 00:10:39.320
number can be can be applied to
any any length of time. So usually

145
00:10:39.360 --> 00:10:43.120
we talk about it in one month
periods because most of our loans have one

146
00:10:43.159 --> 00:10:46.080
month repayment period, so it makes
a lot of sense to look at it

147
00:10:46.120 --> 00:10:48.840
that way. But you could look
at them in days, weeks, years,

148
00:10:48.919 --> 00:10:52.919
quarters, however you like. But
four a period of time, you're

149
00:10:52.960 --> 00:10:54.600
going to have a number, and
that number is going to range called between

150
00:10:54.679 --> 00:11:00.600
zero and three generally, though of
course there's no theoretical upper meant to to

151
00:11:01.320 --> 00:11:05.919
how high the um I could go. But what we've state, what we've

152
00:11:05.960 --> 00:11:09.639
normalized the numbers to mean, is
that is a one dotto generally is meant

153
00:11:09.679 --> 00:11:13.360
to correspond to a what we think
of as a sort of historically normal time

154
00:11:13.639 --> 00:11:18.159
or at least historically normal time in
our recorded history, but you can generally

155
00:11:18.240 --> 00:11:24.679
think of as being the sort of
five years before covidum, and those those

156
00:11:24.720 --> 00:11:28.960
five years there was relatively sort of
a stable macro environment. We did not

157
00:11:30.080 --> 00:11:33.120
see large changes. You know,
again, taking that reference model approach,

158
00:11:33.159 --> 00:11:37.480
if you're just underwritten all the loans
using the same model, you would basically

159
00:11:37.519 --> 00:11:43.399
see the same sort of results across
across the different calendar years, whereas u

160
00:11:43.799 --> 00:11:46.240
oras after that. Of course,
once you get into the COVID world,

161
00:11:46.279 --> 00:11:50.200
you see both large positive and negative
shocks to that, which is exactly you

162
00:11:50.240 --> 00:11:54.120
know, what this methodology is hoping
to tell us about. Yeah, I

163
00:11:54.120 --> 00:11:58.960
imagine that the COVID period, particularly
the very serious economic impact than the rapid

164
00:12:00.000 --> 00:12:05.080
an input of stimulus and then the
rapid withdrawal of stimulus, was like changing

165
00:12:05.120 --> 00:12:09.080
of macro conditions on steroids, or
at least at fast forward speed in terms

166
00:12:09.080 --> 00:12:11.759
of a normal cycle, which I
think is kind of what was some of

167
00:12:11.799 --> 00:12:13.720
the impetus for this is that kind
of rapidity of change. What we usually

168
00:12:13.720 --> 00:12:18.240
think of is maybe not a super
slow but but not not as rapid of

169
00:12:18.080 --> 00:12:20.080
a of a changing environment as we
experience over the last twenty four months.

170
00:12:20.960 --> 00:12:26.440
Yeah, totally. I mean,
it was almost a perfect couple of years

171
00:12:26.519 --> 00:12:31.200
to showcase just the difficulty of the
macro prediction problem, which you know,

172
00:12:31.200 --> 00:12:37.240
we've we've always a felt was not
really a good a good prediction problem in

173
00:12:37.279 --> 00:12:39.120
the sense that these you have so
few of these events, as we talked

174
00:12:39.120 --> 00:12:43.519
about, and each shock looks really
different, and so trying to predict them

175
00:12:43.600 --> 00:12:46.159
based on you know, pass shocks
looks kind of silly. And I mean

176
00:12:48.080 --> 00:12:52.600
April twenty twenty was probably the most
extreme example of this, where April twenty

177
00:12:52.639 --> 00:12:56.879
twenty hits And of course we have
we have we've built over the years plenty

178
00:12:56.919 --> 00:13:01.279
of what you'd call macro prediction models
that look at things like unemployment rates and

179
00:13:01.559 --> 00:13:05.679
UH and and and UH and various
sort of macro indicators that unemployment generally has

180
00:13:05.919 --> 00:13:09.919
in past shock's been the single biggest
driver. But what happened in April twenty

181
00:13:11.000 --> 00:13:15.519
twenty while unemployment rates went through the
roof a sort of um never before seeing

182
00:13:15.759 --> 00:13:20.159
a rapid rise in unemployment, and
yet what happened to loan repayment, Well,

183
00:13:20.320 --> 00:13:24.480
it turns out that within a very
short time, loan repayment did not

184
00:13:24.519 --> 00:13:30.440
only did not get worse. It
became the best period of loan repayment probably

185
00:13:30.840 --> 00:13:33.320
ever in all of human history.
Um and uh. And it was because,

186
00:13:33.440 --> 00:13:37.519
of course, the you know,
the government responded very forcefully to the

187
00:13:37.679 --> 00:13:41.679
increased unemployment the pandemic and um did
a lot of stimulus in various ways.

188
00:13:41.679 --> 00:13:46.919
And and people had a lot of
extra cash to make their loan payments and

189
00:13:48.039 --> 00:13:50.679
uh and and not as many places
to spend it because they were locked up

190
00:13:50.720 --> 00:13:54.200
at home um and and so you
you've got exactly the opposite effect of what

191
00:13:54.679 --> 00:13:58.960
sort of historical precedent would have told
you. And then if you look at

192
00:14:00.000 --> 00:14:03.399
twenty twenty two, you get almost
the mirror image of that, where you

193
00:14:03.440 --> 00:14:07.679
have extremely low unemployment, basically pushing
records in terms of how low unemployment gets,

194
00:14:09.080 --> 00:14:13.320
and yet you have this rapid rise
in delinquency rates. And it's for

195
00:14:13.399 --> 00:14:16.360
exactly the same reason. The stimulus
ends. People still had jobs, But

196
00:14:16.440 --> 00:14:22.559
it turned out that printing money and
giving it to people had a bigger impact

197
00:14:22.559 --> 00:14:24.639
on whether they would repay their loans
and seeing if they were working or not.

198
00:14:26.919 --> 00:14:30.360
So I want to ask about kind
of how traditional lender. I mean,

199
00:14:30.360 --> 00:14:35.639
this is certainly the topic of how
is the macro economy playing into loan

200
00:14:35.759 --> 00:14:39.159
performance? Every time I talk to, lenders are looking at inflation rates and

201
00:14:39.240 --> 00:14:45.080
interest rates and unemployment rates and trying
to understand what that's going to mean or

202
00:14:45.200 --> 00:14:48.559
is doing to a given portfolio.
So how does how do you think traditional

203
00:14:48.600 --> 00:14:52.320
lenders have looked at the kind of
problem you're trying to model, or at

204
00:14:52.399 --> 00:14:56.320
least the data point of, like
what is the macro environment doing to the

205
00:14:56.399 --> 00:15:00.759
loss rates in a pool? In? What are the as in the way

206
00:15:00.759 --> 00:15:03.080
it's traditionally been done that you were
trying to solve for when you thought about

207
00:15:03.120 --> 00:15:07.360
building out the U and my methodology. Yeah, so I think there's there's

208
00:15:07.399 --> 00:15:11.679
spectrum of approaches, but they share
a couple of things in common. The

209
00:15:11.759 --> 00:15:16.240
first is that because many traditional lenders
don't change their underlying models very much,

210
00:15:16.360 --> 00:15:22.159
or they have very simple ones that
are have a very limited number of variables,

211
00:15:22.399 --> 00:15:28.039
they can kind of make an assumption
and it doesn't seem as crazy as

212
00:15:28.080 --> 00:15:31.960
it does in our world, though
we'll get to why it still suffers from

213
00:15:31.000 --> 00:15:35.559
something the same challenges, but it
doesn't seem as crazy to make the assumption

214
00:15:35.639 --> 00:15:39.200
that, hey, we are underwriting
with a constant model and so the barrow

215
00:15:39.200 --> 00:15:43.639
wars that we produce across different years
should roughly be the same. And if

216
00:15:43.679 --> 00:15:46.720
they're roughly the same, then I
can understand where the macro is just by

217
00:15:46.759 --> 00:15:52.519
looking at the average default rate in
my own portfolio. And this is this

218
00:15:52.559 --> 00:15:56.399
is sort of okay. It's sort
of okay because if you actually don't change

219
00:15:56.399 --> 00:16:00.759
your model in any way, then
actually you do have a fairly sort of

220
00:16:00.759 --> 00:16:06.240
constant model. But of course it
also assumes that your barrower is your barrower

221
00:16:06.240 --> 00:16:11.039
mixes unchanging across time, that your
sort of pricing approaches have not changed across

222
00:16:11.080 --> 00:16:15.799
time, and of course that your
model actually hasn't changed in any way across

223
00:16:15.840 --> 00:16:18.399
time. And of course we know
that even if you just were using say

224
00:16:18.440 --> 00:16:22.519
the Phyco scored down right, of
course the Phyco score distribution that are changing

225
00:16:22.559 --> 00:16:26.519
across time and inputs to it,
you know, are are are getting changed

226
00:16:26.559 --> 00:16:30.600
by you know how medical you know, bankruptcies are handled in sort of different

227
00:16:30.600 --> 00:16:33.200
things like this. So there're actually
a bunch of these changes happening under the

228
00:16:33.200 --> 00:16:36.600
hood. And maybe they're not that
extreme, but but put together, they're

229
00:16:36.600 --> 00:16:38.840
probably each you know, ten twenty
percent type effect, and soon you could

230
00:16:38.840 --> 00:16:42.440
be in this world where you might
be seeing default rates that are fifty percent

231
00:16:42.559 --> 00:16:48.320
higher and now you have to you
have to ascribe that to macro, when

232
00:16:48.399 --> 00:16:52.399
in reality it might just be that
your barrower mix change twenty percent and your

233
00:16:52.840 --> 00:16:56.480
your underwriting model changed twenty percent,
and and there you are with the fifty

234
00:16:56.519 --> 00:17:03.359
percent. So that that's that's that's
pretty inherently challenging to measure macro in And

235
00:17:03.400 --> 00:17:07.880
then I think on the sort of
response side to macro, I think it's

236
00:17:07.920 --> 00:17:14.599
it's also when you don't have h
we don't have sort of a precise factor

237
00:17:14.680 --> 00:17:19.119
that's able to separate what your underwriting
models is doing and what what is happening

238
00:17:19.119 --> 00:17:22.480
in the macro. There's always this
temptation to say, well, we can

239
00:17:22.599 --> 00:17:27.680
fix the problem just by changing the
underwriting model. So if we only focus

240
00:17:27.720 --> 00:17:32.920
on higher phycho barrowers or higher income
borrowers, then then we don't have to

241
00:17:32.960 --> 00:17:37.000
worry about the effects of a recession
or the effects of a macro shock.

242
00:17:37.039 --> 00:17:41.279
And I think that that is an
approach that really can get you into trouble

243
00:17:41.319 --> 00:17:44.240
where you know, you look at
a lot of pass recessions something like two

244
00:17:44.279 --> 00:17:47.559
thousand and eight, and you look
at what happened to credit card delinquency rates

245
00:17:47.559 --> 00:17:49.119
in two thousand and eight or mortgage
delinquency rates in two thousand and eight.

246
00:17:49.160 --> 00:17:56.319
And what you see is actually that
the largest multiple increases to delinquency and default

247
00:17:56.359 --> 00:18:00.480
rates were in high phyco buckets.
Maybe it's a little counter tuitive, but

248
00:18:00.480 --> 00:18:02.400
then if you think about what was
happening in two thousand and eight, of

249
00:18:02.400 --> 00:18:04.400
course, one thing is that the
biggest stress for on people who were buying

250
00:18:04.480 --> 00:18:08.279
large homes and had big, big
mortgages, and and of course that's going

251
00:18:08.319 --> 00:18:12.119
to be a primer consumer um.
And another thing is that, of course

252
00:18:12.400 --> 00:18:15.759
it matters quite a lot, you
know, the kind of shock in the

253
00:18:15.799 --> 00:18:21.799
economy and whether that's something that's going
to impact more prime people or less prime

254
00:18:21.839 --> 00:18:26.000
people. But but some of some
of the less prime people actually m in

255
00:18:26.039 --> 00:18:30.000
some sense are like always in a
recession, they are kind of living paycheck

256
00:18:30.039 --> 00:18:33.720
to paycheck in a way that means
their lass rates are are normally much higher,

257
00:18:33.759 --> 00:18:37.319
but it's also harder for them to
get so much higher. Whereas you

258
00:18:37.319 --> 00:18:41.799
know, when you have someone who
goes from being comfortably sort of like you

259
00:18:41.799 --> 00:18:45.000
know, putting money in the bank
every month to suddenly loses their job.

260
00:18:45.000 --> 00:18:48.880
It's it's sort of a fairly tremendous
change in their ability to pay things.

261
00:18:48.960 --> 00:18:52.920
And of course the dead oppligations are
much larger. There's sort of less kind

262
00:18:52.960 --> 00:18:56.720
of ability to easily make up for
it just by um by you know,

263
00:18:57.079 --> 00:19:03.279
uh going to backup sources of cash
and so there's these different effects that play

264
00:19:03.319 --> 00:19:11.200
out. But but it's it's very
hard, I think nearly impossible to solve

265
00:19:11.279 --> 00:19:15.839
for macro problems by changing the borrower
level characteristics that you want to write too,

266
00:19:17.160 --> 00:19:21.759
because there are no set of characteristics
that are inherently a recession proof or

267
00:19:21.880 --> 00:19:26.440
macro shock proof and um and even
if even if there were, frankly,

268
00:19:26.519 --> 00:19:32.160
I think we don't realistically have the
historical data to identify what those are today.

269
00:19:32.640 --> 00:19:34.720
The other thing that occurred to me
is I was reading through the approach

270
00:19:34.759 --> 00:19:38.640
to you, m I is the
sense that particularly to see things in real

271
00:19:38.680 --> 00:19:41.160
time, you know you not only
mean what you've really said is to do

272
00:19:41.200 --> 00:19:44.440
this right, I need a way
to hold the bar or based constant,

273
00:19:44.480 --> 00:19:48.480
some sort of a constant reference model, the credit model constant. But I

274
00:19:48.480 --> 00:19:52.039
also need really an individualized risk prediction
and ideally a month based risk prediction,

275
00:19:52.079 --> 00:19:55.680
because if I want to see how
it happened, what happened in this month,

276
00:19:55.720 --> 00:19:59.000
I want to see that I expect
this some months or every you know,

277
00:19:59.000 --> 00:20:02.880
every given month of se reasoning that
could be for your loan portfolio exists

278
00:20:02.880 --> 00:20:06.240
in the portfolio somewhere, and you
want to know is this loan that went

279
00:20:06.319 --> 00:20:08.319
delink when a month thirty two?
Was that unexpected or expected? And so

280
00:20:08.400 --> 00:20:12.039
you need really a pretty precise model
of risk to be able to see this

281
00:20:12.359 --> 00:20:15.960
in something like real time, I
mean not real time, but to see

282
00:20:15.960 --> 00:20:18.759
it rapidly and quickly and to be
able to adjust. And so it occurred

283
00:20:18.759 --> 00:20:23.960
to me that the kind of individual
risk prediction and loan level and month level

284
00:20:25.039 --> 00:20:29.359
risk predictions in the reference model are
really critical to being able to see not

285
00:20:29.519 --> 00:20:33.680
just with granularity but with speed what's
actually happening in the environment and how it

286
00:20:33.759 --> 00:20:37.240
might be changing than the underlying from
what the underlying predictions might have expected.

287
00:20:37.559 --> 00:20:44.680
Yeah, that's that's totally right.
If you only have lifetime default probability predictions,

288
00:20:44.759 --> 00:20:48.960
or you only have those for certain
buckets of borrowers, and you're not

289
00:20:48.000 --> 00:20:49.680
going to be able to do this
then, and you won't be able to

290
00:20:49.680 --> 00:20:52.920
control for the borrower mix at all, well with a reference model that doesn't

291
00:20:52.960 --> 00:20:59.359
consider enough individual characteristics. And if
you if you only have a lifetime prediction,

292
00:20:59.400 --> 00:21:02.920
then of course you can't compare what's
going on in different periods of time.

293
00:21:03.160 --> 00:21:04.799
So you really need a model where
you have a lot of confidence that

294
00:21:06.079 --> 00:21:11.640
you know, the month seven prediction
for this particular individual really is a good

295
00:21:11.680 --> 00:21:15.559
prediction, at least in relative terms
relative to your predictions in the other months

296
00:21:15.559 --> 00:21:19.640
and relative to your predictions for other
borrowers in that month. And if you

297
00:21:19.680 --> 00:21:23.480
have confidence in that, if we
think of as you know, if there

298
00:21:23.519 --> 00:21:29.160
are months on the sort of X
axis and borrowers on the Y axis,

299
00:21:29.240 --> 00:21:32.279
you need to feel like both your
axes are really doing a lot of work

300
00:21:32.319 --> 00:21:34.400
for you, that it's sort of
an unbiased estimate. And if that's the

301
00:21:34.440 --> 00:21:38.160
case, then you can actually take
that point on the matrix, if you

302
00:21:38.200 --> 00:21:41.519
will, and compare it to other
points on the matrix, and it tells

303
00:21:41.519 --> 00:21:44.759
you something meaningful. And of course
if you don't have that, then you

304
00:21:44.759 --> 00:21:48.319
can't do that comparison. Yeah,
so I did want to ask you call

305
00:21:48.359 --> 00:21:53.039
this the macron actually talked about the
macro and micro prediction levels of the model.

306
00:21:55.279 --> 00:21:57.759
And it's interesting because macro can be
a reference to the macro economy,

307
00:21:57.799 --> 00:22:00.559
the kind of broader world, but
it's also kind of size metric macro and

308
00:22:00.640 --> 00:22:03.200
micro, And so I kind of
want to just ask the question, like,

309
00:22:03.400 --> 00:22:07.319
how do you think about the relative
importance in any given moment or overall

310
00:22:07.400 --> 00:22:11.559
for the micro portion of what you
might call the core risk modeling that up

311
00:22:11.559 --> 00:22:15.559
starts being focused on and this kind
of macro level that's maybe harder to predict

312
00:22:15.640 --> 00:22:19.559
but now somewhat more measurable through you
am I how do those kind of stack

313
00:22:19.640 --> 00:22:23.000
up in terms of how important they
are to a lender into a lender's portfolio

314
00:22:23.039 --> 00:22:30.400
performance. Yeah, so there,
I might first say they're both really important.

315
00:22:30.200 --> 00:22:36.400
They're both important in the context of
you know, fixed income of any

316
00:22:36.480 --> 00:22:38.480
kind has this property that you can
lose a lot more money than you can

317
00:22:38.519 --> 00:22:42.599
make. That's that's the nature of
fixed income, and that means, you

318
00:22:42.640 --> 00:22:48.559
know, generally, anything that can
meaningfully increase your default rates is something you're

319
00:22:48.599 --> 00:22:51.720
really going to care about. And
certainly both the macro and the micro here

320
00:22:51.759 --> 00:22:55.279
fall into that bucket. If you
get those really wrong, it's going to

321
00:22:55.359 --> 00:22:59.319
really affect the return you were expecting. Now, having said that, if

322
00:22:59.359 --> 00:23:03.480
you were to just force a competition
between the macro and the micro, how

323
00:23:03.559 --> 00:23:07.559
much can losses vary based on macro
and how much can loss is based on

324
00:23:07.599 --> 00:23:11.039
micro? I mean fatly, then
it's not even close. Because the macro,

325
00:23:11.240 --> 00:23:14.039
you know, you can get you
can get a period of time that's

326
00:23:14.119 --> 00:23:18.440
three or four maybe in an extreme
pace, five times as bad as another

327
00:23:18.480 --> 00:23:25.119
period in micro. Of course,
you can get borrowers, and not just

328
00:23:25.200 --> 00:23:29.000
theoretically, but even identifiably. For
example, in our model, you can

329
00:23:29.079 --> 00:23:33.200
identify borrowers who are like eighty percent
likely to default, and you can identify

330
00:23:33.240 --> 00:23:37.720
borrowers who are zero point eight percent
likely to default. So you know,

331
00:23:37.759 --> 00:23:42.079
you're talking about a one hundred x
spread between the sort of highest and lowest

332
00:23:42.200 --> 00:23:48.000
risk in micro, and you're talking
about a five to one spread in the

333
00:23:48.079 --> 00:23:51.319
highest and lowest risk in macro.
And so, of course it's so much

334
00:23:51.359 --> 00:23:53.160
more important to get the micro right
than the macro right. There are so

335
00:23:53.200 --> 00:23:57.240
many more ways that you know,
even in the best of times, you

336
00:23:57.240 --> 00:24:02.759
know, if you try to go
to market and underwrite with with an extremely

337
00:24:02.759 --> 00:24:07.839
poor performing model, just everything will
go wrong. And because not only because

338
00:24:07.920 --> 00:24:11.000
your model will be wrong. You
know, you face adverse selection because other

339
00:24:11.079 --> 00:24:14.480
lenders will pick off at lower aprs
the better borrowers. I mean, just

340
00:24:14.519 --> 00:24:18.240
so many, so many things can
can go wrong in micro, whereas in

341
00:24:18.359 --> 00:24:22.079
macro things can also go wrong,
such as you know, when you go

342
00:24:22.160 --> 00:24:29.079
from a really really unreasonably good period
like twenty twenty one to really a sort

343
00:24:29.119 --> 00:24:33.279
of sharply bad period like twenty twenty
two, that's that's going to significantly and

344
00:24:33.400 --> 00:24:37.759
negatively impact lender returns. But it's
still it's still going to be on you

345
00:24:37.799 --> 00:24:42.559
know, you know, if we're
talking about annualized return impact, I mean

346
00:24:42.599 --> 00:24:45.640
depending on what your risk mixes.
I mean you're still talking. You know,

347
00:24:45.680 --> 00:24:49.400
you can count it on your two
hands, like the number of percentage

348
00:24:49.400 --> 00:24:52.039
points of impact that's going to have. WHEREA if you get the micro wrong,

349
00:24:52.079 --> 00:24:55.480
I mean, there's there's really no
limit to how much money you can

350
00:24:55.519 --> 00:24:57.240
lose doing this. And and of
course on the flip side, you know,

351
00:24:57.240 --> 00:25:02.119
when things are going well, micro
is what enables you and has enabled

352
00:25:02.200 --> 00:25:07.319
the lenders that use our platform over
the years to be able to make dramatically

353
00:25:07.720 --> 00:25:12.079
more more income running a lending business. Because micros what enables you to confidently

354
00:25:12.440 --> 00:25:17.519
expand your approval box while holding your
risk levels constant. So if you can

355
00:25:17.519 --> 00:25:22.799
improve fifty percent or one hundred percent
or two hundred percent as many as many

356
00:25:22.839 --> 00:25:25.960
borrowers as you previously could, but
you get the same risk levels, then

357
00:25:25.960 --> 00:25:30.039
of course you know suddenly now you're
able to meaningfully change the amount of lending

358
00:25:30.079 --> 00:25:33.720
income being general. Yeah, interesting, Well, I want to dive in

359
00:25:33.759 --> 00:25:37.680
a different topic you made this comment
about, like if you can model out

360
00:25:37.720 --> 00:25:41.279
if the world the macro got fifty
percent worse or I'd love to say fifty

361
00:25:41.279 --> 00:25:45.160
percent better. Sometimes I worry we
talk about the macro, we only talk

362
00:25:45.200 --> 00:25:48.079
about bad, talk about macro getting
worse and protecting, and not the fact

363
00:25:48.079 --> 00:25:52.359
that, particularly from where we sit
as we're recording this, there's a lot,

364
00:25:52.519 --> 00:25:56.680
probably more upside for the macro getting
better than the substantially worse over the

365
00:25:56.680 --> 00:26:00.960
next at least twelve or eighteen months. But the interesting thing is you haven't

366
00:26:02.000 --> 00:26:04.680
talked really about what I would consider
traditional macro variables. Right, Like most

367
00:26:04.720 --> 00:26:07.599
people when they talk about how is
the macro doing? You know you have

368
00:26:07.640 --> 00:26:11.359
a um I you'll talk about where
it's at, But most people look at

369
00:26:11.640 --> 00:26:15.000
inflation unemployment, interest rates, maybe
savings rates, things of this nature that

370
00:26:15.039 --> 00:26:19.759
are very clearly tied to what we
think of as a macroeconomy GDP or something

371
00:26:19.839 --> 00:26:23.440
like that. What is there a
relationship or those part of the U and

372
00:26:23.480 --> 00:26:26.559
my model. Talk to me a
little bit about how like traditional views of

373
00:26:26.599 --> 00:26:30.839
what is happening in the macro relate
to the Upstart macro index, or how

374
00:26:30.839 --> 00:26:36.640
you think about the impact of macro
and the lending portfolio. Yeah, yeah,

375
00:26:36.640 --> 00:26:41.359
they're they're they're super correlated. In
short, the upstart measured you and

376
00:26:41.480 --> 00:26:47.960
I. We like it because it
has the nice property that it's directly applicable

377
00:26:48.079 --> 00:26:52.559
to the math you want to do
on lending um, and because it's it's

378
00:26:52.599 --> 00:26:56.960
really specifically designed to tell you this
period is a period in what you can

379
00:26:57.000 --> 00:27:02.359
expect twice as many deligencies the same
borrower pool as that period. You know

380
00:27:02.400 --> 00:27:06.400
that it lets you do math in
a very precise and direct way. But

381
00:27:06.480 --> 00:27:11.559
of course, UM, it would
be really concerning if this index that's meant

382
00:27:11.559 --> 00:27:18.200
to measure the macro is totally unrelated
to the sort of traditional macroeconomic variables,

383
00:27:18.200 --> 00:27:21.319
which, of course you know,
though each one maybe only tells part of

384
00:27:21.359 --> 00:27:23.480
the story. They are collectively,
you know, a pretty good description of

385
00:27:23.519 --> 00:27:30.079
the things that we care about in
the economy and UM and so we have

386
00:27:30.720 --> 00:27:33.680
that that's been an important part of
developing the UUMI is actually building a sister

387
00:27:33.799 --> 00:27:38.720
model that tells us what we We
haven't found a nice, sort of externally

388
00:27:38.839 --> 00:27:42.559
marketable name for it yet, but
if you think about there's sort of um

389
00:27:42.680 --> 00:27:48.720
I as measured in the sort of
collective repayment rates of the of all Upstart

390
00:27:48.759 --> 00:27:52.440
platform originated loans where we have,
you know, access to the data.

391
00:27:52.519 --> 00:27:59.119
But then there's there's uum I as
you would predict based on what you're seeing

392
00:27:59.240 --> 00:28:03.440
from the the sort of official macroeconomic
data, the unemployment rate, the inflation

393
00:28:03.519 --> 00:28:07.119
rate, the personal savings rate,
and and what you want is you want

394
00:28:07.160 --> 00:28:11.559
to see that these two measures,
they sort of internally measured UMI and the

395
00:28:11.680 --> 00:28:19.480
externally inferred UMI are largely moving together, highly correlated and kind of moving together,

396
00:28:19.680 --> 00:28:22.519
and UM and that's that's exactly what
we see. We've been able to

397
00:28:22.559 --> 00:28:26.519
see that, and actually you can
UM you know. The maybe short and

398
00:28:26.559 --> 00:28:32.720
snappy version of this is if you
just want what single variable released by the

399
00:28:32.759 --> 00:28:37.000
FED can tell you roughly where the
UMI is going, and the answer would

400
00:28:37.039 --> 00:28:40.559
be the personal savings rate. That
of course is a bit cheating because the

401
00:28:40.599 --> 00:28:45.640
personal savings rate itself is a bit
of an amalgamation of a bunch of variables.

402
00:28:45.640 --> 00:28:48.359
It's, you know, looking at
personal and broadly looking at personal income

403
00:28:48.400 --> 00:28:52.279
and personal expenditures. But of course
what makes up personal income. You know,

404
00:28:52.319 --> 00:28:55.400
people have to work, so there's
employment rates baked into that, there's

405
00:28:55.599 --> 00:29:00.240
wage wage growth baked into that,
and of course on the personal expenditures side,

406
00:29:00.279 --> 00:29:03.200
there's sort of your real rate of
spending, and then there's there's the

407
00:29:03.240 --> 00:29:06.359
inflation rate. And so you've kind
of got this one variable that the Fed

408
00:29:06.440 --> 00:29:08.680
publishes, you know, the end
of every month, and it tells you

409
00:29:10.559 --> 00:29:15.079
roughly the things that really matter unemployment, inflation, spending, and wage growth.

410
00:29:15.160 --> 00:29:18.359
And the net of that sort of
is like how much money people have

411
00:29:18.720 --> 00:29:22.640
left over, which is a really
good proxy for, you know, how

412
00:29:22.720 --> 00:29:26.599
much money people are going to have
available to service their their debt. And

413
00:29:26.119 --> 00:29:30.119
if you just look at that one
variable, that's probably like you know,

414
00:29:30.279 --> 00:29:36.039
mostly mostly correlated with how we've measured
you and I over the past years.

415
00:29:36.240 --> 00:29:37.599
Yeah, it does seem like a
i won't say a cheating variable, but

416
00:29:37.599 --> 00:29:41.920
it is an amalgamation that that takes
into account a number of the things that

417
00:29:41.960 --> 00:29:47.400
people traditionally think of as pretty impactful
to at least particularly unsecured credit, where

418
00:29:47.440 --> 00:29:49.000
there's you know, it's kind of
the I want to say the riskies,

419
00:29:49.079 --> 00:29:52.519
but the bottom of the payments to
act in any ways, where the first

420
00:29:52.519 --> 00:29:56.000
thing to take take a hit when
one times get tough. So one thing

421
00:29:56.000 --> 00:30:00.559
we haven't talked about today at all, and I'm sure our lending partnership seems

422
00:30:00.559 --> 00:30:03.119
to be very upset with me to
be this long to ask this question,

423
00:30:03.519 --> 00:30:07.799
But you know, I think there's
obviously when you and I goes up,

424
00:30:08.279 --> 00:30:12.119
macro gets riskier. There's this question
of is that a direct indication of loans

425
00:30:12.480 --> 00:30:18.759
under or overperforming? And how does
the sense of the macro play into how

426
00:30:18.759 --> 00:30:22.640
you actually work with a lending partner
to assign a level of risk to a

427
00:30:22.680 --> 00:30:26.279
borrower at time of underwriting and understand
the returns produced by that portfolio. So

428
00:30:26.559 --> 00:30:30.079
I think the thing that I hear
a lot is like does a U am

429
00:30:30.079 --> 00:30:33.759
I above one? Meaning all my
loans are underperforming because the macro is high,

430
00:30:33.799 --> 00:30:36.839
and therefore I'm losing money and that's
a bad thing and I should stop

431
00:30:36.880 --> 00:30:40.720
lending or whatever. So talking about
how you leverage this in the context of

432
00:30:40.880 --> 00:30:45.920
partnerships with originators to kind of calibrate
to what's going on and help them understand

433
00:30:45.960 --> 00:30:48.799
the level of risk in the environment
that's going into their originations. Yep.

434
00:30:49.759 --> 00:30:56.519
Yeah. So in short note,
it does not there's no direct way to

435
00:30:56.559 --> 00:31:00.039
say you am I is a certain
number, therefore my loans are either we

436
00:31:00.119 --> 00:31:02.559
are underperforming. To know that,
you need to know one more thing,

437
00:31:02.680 --> 00:31:07.440
which is essentially what you and I
were you assuming when you made the loans

438
00:31:07.480 --> 00:31:12.759
in question, and that that particular
thing we call the um the macro forecast.

439
00:31:14.000 --> 00:31:19.839
And if you very roughly speaking,
if you just for any batch of

440
00:31:19.920 --> 00:31:26.319
loans, if you take their UMF
and you then look forward and say,

441
00:31:26.359 --> 00:31:30.359
okay, I made this batch of
loans in January of twenty twenty two,

442
00:31:30.559 --> 00:31:34.559
and at the time we had a
UM of one dot O and and now

443
00:31:34.599 --> 00:31:40.160
the UUMI is one point five.
Then that sort of implies that you're going

444
00:31:40.200 --> 00:31:44.880
to have actual losses that are fifty
percent higher than forecasting. In that case,

445
00:31:45.359 --> 00:31:48.839
sort of the example you gave as
valid. Now fast forward to to

446
00:31:48.039 --> 00:31:55.519
more recent times, and many of
our lenders actually have a much higher UM.

447
00:31:55.720 --> 00:32:00.400
Imagine your UMF is two doto,
but the uum I that's in observe

448
00:32:00.480 --> 00:32:02.880
sense is only one point eight.
Well, then, actually, even though

449
00:32:02.920 --> 00:32:07.640
one point eight is a really high
number by historical standards, UM it's actually

450
00:32:07.880 --> 00:32:13.440
twenty or it's actually ten percent lower
than the UM that you assumed. And

451
00:32:13.440 --> 00:32:16.079
and so that means that your loans
were priced as if you were operating in

452
00:32:16.119 --> 00:32:20.480
a two doto type world. But
actually you've been seeing a one dot a

453
00:32:20.559 --> 00:32:24.319
type world. And now you're going
to see defaults and delinquencies that are running

454
00:32:24.599 --> 00:32:29.079
on average ten percent lower than than
than you were expecting, which of course

455
00:32:29.160 --> 00:32:32.720
means uh, you know, more
more sort of net interest, more more

456
00:32:32.720 --> 00:32:37.000
excess returns, and so returns will
be higher than than expected. And of

457
00:32:37.000 --> 00:32:40.039
course you know it, Actually it
matters how this plays out over the whole

458
00:32:40.400 --> 00:32:45.680
sort of term of the loan.
But um but you know, just roughly,

459
00:32:45.720 --> 00:32:46.880
as an approximation, if you think
about a single point in time,

460
00:32:46.920 --> 00:32:50.759
this kind of tells you. This, this ratio between the um I and

461
00:32:50.880 --> 00:32:54.720
UM tells you how things are going
to be. And so if you underwrite

462
00:32:54.759 --> 00:33:00.680
with a fairly high UM, then
of course there's more room for um I

463
00:33:00.799 --> 00:33:05.160
to positively surprise against that. On
the other hand, if you want to

464
00:33:05.200 --> 00:33:07.839
write with a high UMF, you're
going to be passing up a lot of

465
00:33:07.839 --> 00:33:12.559
borrowers and pricing a lot of other
ones out of out of out of converting,

466
00:33:12.680 --> 00:33:15.039
just because you're assuming that you know, the world is in a really

467
00:33:15.079 --> 00:33:19.480
bad place and you want to,
uh, you know, lend to people

468
00:33:19.559 --> 00:33:22.960
who, uh you know, who
are going to be going through a really

469
00:33:22.039 --> 00:33:27.359
hard time. And of course if
if if they aren't, or if you

470
00:33:27.400 --> 00:33:30.680
know other lenders are not seeing it
that way, then those borrowers are going

471
00:33:30.720 --> 00:33:37.119
to go elsewhere. How how can
you help a lender think about I mean,

472
00:33:37.519 --> 00:33:40.000
what a UUMI or UMF of two
or one point eight or one point

473
00:33:40.039 --> 00:33:43.559
five, Like, I feel like
that's not something that at least as of

474
00:33:43.640 --> 00:33:45.759
yet to be great for you.
If that became something about which people had

475
00:33:45.759 --> 00:33:50.359
an intuition and a made a natural
understanding, it was like a number they

476
00:33:50.359 --> 00:33:52.920
were used to kind of thinking about
where it may trend and where it may

477
00:33:52.960 --> 00:33:55.920
fall. But how do you think
about, like what causes a um I

478
00:33:57.000 --> 00:33:59.000
or UMA to get to a certain
level? Can you I mean, I

479
00:33:59.000 --> 00:34:01.079
think that scenario plan. You know, if you think about just the Dot

480
00:34:01.119 --> 00:34:05.559
Frank stress test kind of thing,
if unemployment goes here and somebody goes here,

481
00:34:05.599 --> 00:34:07.400
like, how bad does that get? What will have my portfolio?

482
00:34:07.599 --> 00:34:12.320
How do you think about connecting the
dots between those that the concept of you

483
00:34:12.360 --> 00:34:15.559
know what might happen and what that'll
mean to you know, where these loans

484
00:34:15.559 --> 00:34:17.480
could be performing. Yeah, I
think they're two good ways to build intuition

485
00:34:17.559 --> 00:34:22.840
around at these UUMI index values.
The first is to go back to that

486
00:34:22.960 --> 00:34:25.760
personal savings rate, which is largely
correlated, not perfect but pretty good and

487
00:34:25.840 --> 00:34:30.599
again includes the unemployment and inflation,
all all the pieces, and and what

488
00:34:30.639 --> 00:34:34.800
you can basically see is that you
can think of it as a historically average

489
00:34:34.880 --> 00:34:37.639
level of the personal savings rate.
Let's just be round numbers. Call it

490
00:34:37.679 --> 00:34:43.320
on the order of ten percent,
roughly corresponds to a sort of normal UUMI

491
00:34:43.480 --> 00:34:47.400
condition of around one point zero.
And then you know what happens, of

492
00:34:47.400 --> 00:34:51.400
course, if the personal savings rate
goes up. You know, if the

493
00:34:51.400 --> 00:34:55.199
personal savings rate goes up to something
like twenty percent, which you know basically

494
00:34:55.280 --> 00:35:00.840
never happens except in in twenty one
when you know people are given lots of

495
00:35:00.960 --> 00:35:06.920
money. Yeah exactly. Um,
but you know when you when you get

496
00:35:06.960 --> 00:35:09.400
something like that, then you see
you and I basically get cut in half

497
00:35:09.800 --> 00:35:14.280
down to about zero point five.
And again I'm just speaking in round numbers

498
00:35:14.280 --> 00:35:16.400
here and then um uh. And
then on the on the other hand,

499
00:35:16.559 --> 00:35:21.280
UM, you know you can see
you and my values go uh, you

500
00:35:21.320 --> 00:35:23.599
know, go from one point zero
to two point zero. If you basically,

501
00:35:23.880 --> 00:35:27.800
you know, roughly speaking, cut
your personal savings rate in half,

502
00:35:28.159 --> 00:35:31.079
which is, you know, on
the order of of what we've seen over

503
00:35:31.559 --> 00:35:36.079
over the past eighteen months or so. And then the other sort of way

504
00:35:36.119 --> 00:35:38.000
to build intuition is just to look
at different periods of time. So again

505
00:35:38.320 --> 00:35:43.519
you and I one dot O roughly
corresponds to the five years before the COVID

506
00:35:43.559 --> 00:35:49.599
period. UM in the sort of
um, happy days of free flowing stimulus

507
00:35:49.639 --> 00:35:52.400
money. Um, you know,
you've got you of mys that go down

508
00:35:52.400 --> 00:35:54.719
to nearly zero point five and then
um and then uh, you know,

509
00:35:54.719 --> 00:35:59.239
stimulus running out producing that large shock
takes you, you and I over one

510
00:35:59.400 --> 00:36:02.159
one one one point five. And
then if you were to sort of look

511
00:36:02.159 --> 00:36:06.599
at historical values of personal savings and
you and I sort of asked the question,

512
00:36:06.639 --> 00:36:08.400
well, how bad can it get? At least how bad have we

513
00:36:08.440 --> 00:36:13.079
seen it get historically? Probably the
answer is something like, you know,

514
00:36:13.199 --> 00:36:16.000
in the worst times ever that again
in recent recorded history, we have all

515
00:36:16.039 --> 00:36:20.800
the relevant economic data, maybe something
like the worst month of the Great Procession.

516
00:36:20.880 --> 00:36:23.440
We think you and I would have
been in the high twos. Interesting,

517
00:36:23.719 --> 00:36:27.559
So I get two more things I
want to hit on, I think,

518
00:36:27.719 --> 00:36:30.199
and then I think we've kind of
at least gotten a good intro for

519
00:36:30.239 --> 00:36:32.920
people into this concept. One is, we didn't really talk about this,

520
00:36:34.000 --> 00:36:38.119
but You and I as you've built
it today is really based explicitly on unsecured

521
00:36:38.159 --> 00:36:43.079
installment loans, particularly the kind that
Upstar partners with banks and credit unions to

522
00:36:43.119 --> 00:36:47.800
originate. Are there intentions to build
something like You and I for other asset

523
00:36:47.840 --> 00:36:51.119
classes? Is it applicable to other
assets? Like? How do you think

524
00:36:51.119 --> 00:36:54.199
about the usefulness of it as a
construct or the limitations of it as a

525
00:36:54.239 --> 00:36:59.920
construct given that it's really a measurement
specifically on a particular category of consumer credit.

526
00:37:00.280 --> 00:37:04.320
Yeah, great question. Um so
we don't we don't see um I

527
00:37:04.480 --> 00:37:07.800
as something that you just develop methodology
once and it's done. I mean,

528
00:37:07.880 --> 00:37:10.320
nice if it was like that.
But really it's a very sort of rich

529
00:37:10.599 --> 00:37:15.800
um surface to explore, because there's
not only the question you mentioned, which

530
00:37:15.840 --> 00:37:19.639
is a question of you and my
for different products, but there's actually you

531
00:37:19.719 --> 00:37:23.199
a mid for different types of consumers. And while maybe on average they will

532
00:37:23.239 --> 00:37:27.400
all move together, they don't,
of course exactly all move together. Their

533
00:37:27.400 --> 00:37:30.400
slopes are a little bit different,
their sensitivities are a little different, and

534
00:37:30.199 --> 00:37:35.280
um, and so what we really
are. You can think of the progression

535
00:37:35.400 --> 00:37:38.480
of you and my methodologies as going
from first you sort of assume there is

536
00:37:38.519 --> 00:37:46.079
just one UOM quote unquote working off
one reference model. Then you have one

537
00:37:46.199 --> 00:37:51.960
UUMI that's produced actually by an amalgamation
of reference models, and that sort of

538
00:37:52.000 --> 00:37:53.800
gets, you know, more and
more continuous as you go from you know,

539
00:37:53.840 --> 00:37:58.079
one to two to end and then
and you really get to a methodology

540
00:37:58.079 --> 00:38:00.119
that that's sort of much more continuous
with the back to your reference model.

541
00:38:00.559 --> 00:38:05.519
And then you have a sort of
second dimension of you and my methodology improvement,

542
00:38:05.559 --> 00:38:08.800
which is really about going from assuming
sort of quote unquote the average borrower

543
00:38:08.920 --> 00:38:14.639
to really making it something that where
the you and I is interacting with the

544
00:38:14.679 --> 00:38:19.639
borrower characteristics, not unlike how we
see all the sort of the other work

545
00:38:19.679 --> 00:38:22.960
we do in in in the sort
of prediction realm is there's just interaction effects

546
00:38:22.960 --> 00:38:29.679
going on between almost any given characteristics
of a borrower, and of course those

547
00:38:29.760 --> 00:38:32.760
also interact with macro and so those
should those should be considered. And then

548
00:38:32.800 --> 00:38:36.639
the last sort of dimension is is
the type of loan it that it is,

549
00:38:36.679 --> 00:38:38.360
and that's you know, true of
it, of its size, of

550
00:38:38.360 --> 00:38:42.559
its duration, of its type,
and so that could be it being an

551
00:38:42.559 --> 00:38:44.960
auto loan, a personal loan,
a three year loan, a one year

552
00:38:45.039 --> 00:38:50.320
loan, or um or any sort
of combination of these. And and so

553
00:38:50.400 --> 00:38:53.559
there's there's there's a lot of sort
of continuing research into you and I.

554
00:38:53.679 --> 00:38:57.519
That's that we're doing. And having
said all that, the good news is

555
00:38:57.519 --> 00:39:01.320
that we think even this first of
you and I, as the first approximation

556
00:39:01.360 --> 00:39:07.400
if you just want to answer the
question of under what conditions will my loans

557
00:39:07.480 --> 00:39:12.039
over and underperform and by roughly how
much, you're going to get a very

558
00:39:12.119 --> 00:39:16.280
good approximation just from you and I
V one. That is is true,

559
00:39:16.400 --> 00:39:20.800
it's built on unsecured personal loans,
but you know, we look at it

560
00:39:20.800 --> 00:39:23.599
in the context of our other products
like our Autorefi products, and again it's

561
00:39:23.639 --> 00:39:28.039
it's not quite as precise, but
it does give you a really close first

562
00:39:28.119 --> 00:39:32.480
order approximation of your level of over
and underperformance, again looking at those uom

563
00:39:32.480 --> 00:39:37.360
I and uu F ratios of the
changes across time. So it does a

564
00:39:37.400 --> 00:39:42.480
good job explaining, and that gives
us more confidence in the methodology because it

565
00:39:42.480 --> 00:39:46.719
tells us that there's something very general
happening in addition to the sort of many

566
00:39:46.960 --> 00:39:51.760
smaller but more specific things going on
within each product, within each type of

567
00:39:51.760 --> 00:39:54.320
borrower, and so on and so
forth. Interesting. I guess my last

568
00:39:54.400 --> 00:39:59.320
question for you is it sounds like
this is early days in the development of

569
00:39:59.360 --> 00:40:02.159
you am I like a very solid
foundation, but compared to the work you

570
00:40:02.199 --> 00:40:07.920
put into understanding micro risk is still
a pretty pretty immature in comparison and has

571
00:40:07.920 --> 00:40:09.920
a lot of room for improvement.
And you also mentioned if people want to

572
00:40:09.920 --> 00:40:13.440
get an intuition, they should look
at the numbers. So if people are

573
00:40:13.480 --> 00:40:19.159
interested in looking at the numbers or
following along as the methodology and kind of

574
00:40:19.199 --> 00:40:23.159
the insights coming out of the work
into macro and the u m I evolve,

575
00:40:23.480 --> 00:40:25.880
is there a place they can go
to see those or they publish somewhere,

576
00:40:25.960 --> 00:40:29.559
is there are place they can follow
along what's happening, Like where can

577
00:40:29.599 --> 00:40:31.199
people go to learn more and keep
up to date with you know, what's

578
00:40:31.199 --> 00:40:35.360
happening to um I and what's happening
to the work into looking into this macro

579
00:40:35.480 --> 00:40:42.280
concept. Upstart dot com slash You
am I just out and we're really excited

580
00:40:42.280 --> 00:40:45.119
about it, and it's going to
get better and better. But it's going

581
00:40:45.199 --> 00:40:46.880
to have the core data there and
there's going to be more ways to engage

582
00:40:46.880 --> 00:40:51.559
with and play with it and really
understand the numbers. So yeah, go

583
00:40:51.639 --> 00:40:53.719
go there, check it out,
see how those values have changed across time,

584
00:40:53.840 --> 00:40:57.320
what they were in different periods of
time, And I think that will

585
00:40:57.360 --> 00:41:00.480
really build a good, good level
of intuition for this. Certainly, it's

586
00:41:00.480 --> 00:41:05.760
our hope that this would become a
tool of general interest beyond those who are

587
00:41:06.320 --> 00:41:09.239
either making loans through our platform or
even making loans in general. But really

588
00:41:09.280 --> 00:41:15.840
we think it's a it's a really
it's a really good and useful tool to

589
00:41:15.960 --> 00:41:20.920
just understand what's going on in the
world. Upstart dot com Slash You am

590
00:41:20.920 --> 00:41:23.800
I. It's not a very original
you are l very straightforward and easy to

591
00:41:23.840 --> 00:41:29.000
remember, so I guess that's probably
better than original, and I look forward

592
00:41:29.000 --> 00:41:31.239
to see that. I'm actually really
excited that there's we're going to be publishing

593
00:41:31.239 --> 00:41:35.159
the data set and so kind of
a historical data set. I think there's

594
00:41:35.159 --> 00:41:37.960
a lot of rich surface area for
innovation on top of it, so I'll

595
00:41:37.000 --> 00:41:39.880
be excited to see. And if
any of the listeners have questions or feedback

596
00:41:39.920 --> 00:41:44.159
on what's useful or what you'd like
to learn more about the concept, please

597
00:41:44.480 --> 00:41:46.639
do right to us or right to
just me and Jeff at upstart dot com,

598
00:41:46.679 --> 00:41:51.239
because I think it's actually a really
interesting space and we'd love to have

599
00:41:51.320 --> 00:41:53.320
real input from all of you in
terms of where where we take the research

600
00:41:53.320 --> 00:41:55.679
and what we learn next. All
right, Paul, thanks for taking the

601
00:41:55.679 --> 00:41:59.639
time to joyces Is. I think
a great first dive, and I appreciate

602
00:41:59.639 --> 00:42:02.159
your back for your second time on
the podcast. Awesome, great, thanks

603
00:42:02.159 --> 00:42:07.039
for having me. Upstart partners with
banks and credit unions to help grow their

604
00:42:07.039 --> 00:42:13.199
consumer loan portfolios and deliver a modern, all digital lending experience. As the

605
00:42:13.239 --> 00:42:17.079
average consumer becomes more digitally savvy,
it only makes sense that their bank does

606
00:42:17.079 --> 00:42:23.639
too. Upstarts AI lending platform uses
sophisticated machine learning models to more accurately identify

607
00:42:23.760 --> 00:42:30.079
risk and approve more applicants than traditional
credit models. With fraud rates near zero,

608
00:42:30.760 --> 00:42:36.159
upstarts all digital experience reduces manual processing
for banks and offers a simple and

609
00:42:36.199 --> 00:42:40.559
convenient experience for consumers. Whether you're
looking to grow and enhance your existing personal

610
00:42:40.559 --> 00:42:45.760
and auto lending programs or you're just
getting started, upstart can help. Upstart

611
00:42:45.840 --> 00:42:50.960
offers an end to end solution that
can help you find more credit worthy borrowers

612
00:42:50.960 --> 00:42:55.119
within your risk profile. With all
digital underwriting, onboarding, loan closing,

613
00:42:55.159 --> 00:43:00.079
and servicing, It's all possible with
Upstart in your quarter. Learn more about

614
00:43:00.079 --> 00:43:06.079
finding new borrowers, enhancing your credit
decisioning process, and growing your business by

615
00:43:06.159 --> 00:43:10.960
visiting upstart dot com slash foward dash
Banks. That's upstart dot com slash foward

616
00:43:12.199 --> 00:43:15.719
dash Banks. You've been listening to
Leaders and Lending from Upstart, make sure

617
00:43:15.800 --> 00:43:20.840
you never miss an episode. Subscribe
to Leaders in Lending in your favorite podcast

618
00:43:20.880 --> 00:43:23.480
player using Apple Podcasts. Leave us
a quick rating by tapping the number of

619
00:43:23.480 --> 00:43:28.000
stars you think the show deserves.
Thanks for listening. Until next time,