0:02

Welcome to Macrohive Conversations with Balal Hafiz.

0:05

Macrohive uses natural and artificial intelligence to

0:07

educate investors and provide actional of insights

0:09

for all markets from rates to effects

0:12

to equities. For our latest insights, visit

0:14

macrohive.com. Before I start my conversation with

0:16

this episode's guest, I have three requests.

0:19

First, please make sure to subscribe to

0:21

this podcast. Apple Spotify or if you

0:23

listen to podcast, leave some nice feedback

0:26

and let your friends know about the

0:28

show. My second request is that you

0:30

sign up to a free weekly

0:32

newsletter that contains market insights and

0:35

unlocked content. You can sign up

0:37

for that at macrohigh.com. Finally, and

0:39

my third request is that if

0:42

you are a professional or institutional

0:44

investor, do get in touch with

0:46

me. We have a very

0:49

high-octane research and analytics offering

0:51

that includes access to our

0:53

world-class research team, trade ideas,

0:55

AI models, and much, much more.

0:58

You can email me at Balal

1:00

at macrohigh.com, or you can message

1:02

me on Bloomberg for more details.

1:04

Vania is an assistant professor of

1:06

economics at the London Business School and

1:09

has a PhD from Harvard. She has

1:11

extensively studied exchange rate determination in how

1:13

monetary policy, macroeconomic news and beliefs impact

1:16

exchange rate movements. She has also produced

1:18

innovative research on the global network of

1:20

equity holdings and their impact on equity

1:23

fixing them and exchange rate markets. Now

1:25

on to our conversation. Greetings and welcome

1:27

Vania. It's fantastic to have you on the

1:29

podcast show. Thank you so much for having me. It's

1:31

a pleasure to be here. Now before we start

1:33

the meat of our conversation I do like to

1:36

ask my guest something about their background so it'll

1:38

be great to know something about your origin story

1:40

what did you study at university and was inevitable

1:42

you'd end up in academia. So thank you for

1:45

this question. So I grew up in

1:47

Bulgaria I'm originally from Bulgaria and I'm

1:49

the last generation that lived through communism

1:51

and as a result through all the

1:53

crisis following communism, hyperinflation, sovereign debt crisis,

1:55

banking crisis, which is why I was

1:57

interested in interesting economics. So when I

1:59

went into college. wanted to study economics,

2:01

particularly international finance. I also studied math,

2:03

as every diligent Eastern European is advised

2:06

to study, which thankfully I did. And

2:08

then I spent two years at the

2:10

Brookings Institute, followed by my PhD at

2:12

Harvard. Great. Fantastic. And I see that in

2:15

recent years, you've done a lot of work

2:17

on currency market for exchange. Why currencies, rather

2:19

than any of the market? So exchange rates

2:21

was always a passion of mine. Actually,

2:23

my first FX paper was after undergrad

2:25

when I met Ken Rogofet Brookings and

2:27

we wrote a paper on an example

2:30

of exchange rate forecasting. And of course,

2:32

exchange trade is one of the most

2:34

interesting and important prices in international finance,

2:36

right? So, and it's one of the

2:38

least understood prices in financial markets, which

2:40

is I like a challenge, which is

2:42

why I studied it. So indeed, I

2:44

have studied exchange from various angles throughout

2:46

my career, and I have a number

2:48

of papers on it. If you would

2:50

like me, I can just give you

2:53

a brief overview of what are the

2:55

different ways to think about effects

2:57

determination, and I'm going to

2:59

bring my research throughout the

3:01

discussion. Yeah, maybe we could

3:04

frame it. You mentioned Ken

3:06

Rogoff, and he, of course,

3:08

wrote that famous paper in the

3:10

early 80s, like Dick Macy and Ken

3:12

Rogoff, on our inability to forecast currencies,

3:15

trying to not just... show that exchange

3:17

rates are forecasted by out of sample,

3:19

which has been incredibly hard. But also

3:21

in sample, it was shown for many

3:24

years that one count five reconnect between

3:26

exchange rates and pretty much almost any

3:28

variables. So the good news is that

3:31

the in sample reconnect has been shown

3:33

to exist now, and actually my work

3:35

has contributed towards that, particularly with macroeconomic

3:38

surprises, and I can tell you a

3:40

bit more about that. Of course, out

3:42

of sample forecasting remains challenging, and I

3:44

am exploring it in ongoing work of

3:46

mine. Yeah, great. Okay, so let's go back

3:48

to what you were saying earlier about the

3:51

determinants of exchange rates. Yes, there are two

3:53

ways to think about exchange determination determination.

3:55

Essentially one of the standard ways

3:57

is to take what's called the

4:00

nor arbitrage condition in our finance models,

4:02

which we call the Euler equation, right?

4:04

And then you can express any asset

4:07

price, including exchange rates as a function

4:09

of forward-looking variables. So what do I

4:11

mean by this? You can show that

4:13

the exchange rate can be exchanged over

4:16

any two periods, can be expressed as

4:18

the interest rate differential in period T,

4:20

if you're studying the T plus to

4:23

T plus one change, for example, into

4:25

currency risk premium as a PFT. the

4:27

revisions in expectations over the relative nominal

4:29

policy rate paths, revisions and expectations

4:32

over the relative nominal essentially inflation

4:34

rate paths across the two countries,

4:36

and revisions in expectations over the

4:38

current city's period. So in my

4:40

research with Jenny Tang at the

4:42

Boston Fed, we show essentially constructing

4:44

these components of this decomposition to

4:46

do that we actually use survey

4:48

data in order to back out

4:50

the whole expected path of inflation

4:52

expectations, interest rate expectations, and exchange

4:54

rate expectations, and exchange rate expectations.

4:56

We show that the forward-looking components,

4:58

not surprisingly, are the main driver

5:01

of exchange rates. In particular... And on the data,

5:03

where do you get the survey data? So we

5:05

use consensus economics, blue chips data,

5:07

you can effectively use their survey

5:09

data and the approach the methodology

5:11

that we use is essentially you

5:13

take your standard way to construct

5:15

expectations of variables, which would be

5:18

you run some kind of vector

5:20

regressive essentially process. But then instead

5:22

of just stopping there, you are

5:24

disciplining the expectations produced by this

5:26

model using survey data. Right. So

5:28

the idea is that. we know

5:30

that running a simple VAR is

5:32

going to generate massive downward bias

5:34

and very flat expectations two years ahead.

5:36

So for example, if you're familiar with

5:39

the Campbellshire decompositions, which is a similar

5:41

decomposition for stocks, for those researchers that

5:43

back in the day that would use

5:45

a simple VAR to get the expectations

5:47

going forward, they realized that's garbage, right?

5:49

So the literature moved away from this

5:51

kind of simple VAR-based approach, and one

5:53

of the ways to remedy this problem

5:55

is to discipline the expectations using survey

5:57

data, which has been done in the...

6:00

fixed income space, for example, to decompose,

6:02

yields into expectation, her policy and term

6:04

premium, we were the first to do

6:06

it for effects. So we have a

6:08

series of- And just for our audience,

6:10

when you say VAR, you mean a

6:12

vector auto-regression functions as an econometric, you

6:14

know, equation, rather than value risk. You

6:16

usually run a VAR, but most people

6:19

are going to end there, right? So

6:21

what we say is you're going to

6:23

get a massively biased forecast forecast by

6:25

running a simple VAR. unless you use

6:27

some econometric techniques to address the small

6:29

sample bias. But in even better ways

6:31

to say, well, actually, if we believe

6:34

that the survey data captures fairly well

6:36

the expectations of the market, there is

6:38

pricing, currently asset prices, right? So whatever

6:40

expectations they have. factors in and the

6:42

trading decisions, etc. And that's what we

6:44

do. So we amend the VAR approach

6:47

so that the VAR-based expectations are fairly

6:49

close to the survey-based expectations. So that's

6:51

kind of the methodology of the procedure.

6:54

So using that procedure, you see that

6:56

it is the currency risk premium, right?

6:58

That's kind of the main driver. So

7:00

then when I told you that now

7:03

we have a reconnect between macro variables,

7:05

you might wonder, okay, usually currency is

7:07

perceived to be driven by stuff like

7:10

risk aversion. We have shown, you know,

7:12

to a large degree, connection these days

7:14

between exchange rates and this kind

7:17

of more financial variables, if

7:19

we wish, that measure. Market

7:21

sentiment risk aversion, you know,

7:23

liquidity issues in financial markets,

7:25

etc. CAP deviation is doing an

7:28

other important variable. Now, what we

7:30

show in our world that, yes,

7:32

it is the case that this

7:34

forward-looking currency is a premier component

7:37

is the same macro news. that

7:39

are linked to inflation surprises and

7:41

what I mean by surprises is

7:43

realization of inflation CPI inflation minus

7:45

let's say expectations prior to the

7:47

release of the CPI. So macro surprises

7:49

you're saying drive risk premier as much as

7:52

big as these are the factors? Yes that's

7:54

how you can think about it right so so

7:56

why is it the case the model that tends

7:58

to explain not just if but we've applied

8:00

it also to stock markets in

8:03

some of our work is risk

8:05

aversion is very tightly linked to

8:07

macro fundamental so your risk on

8:09

period can be explained actually the

8:11

global financial cycle in your risk

8:13

on period actually in my in

8:15

my some of my latest work

8:17

we show that can be linked

8:19

to US-centered macro news in particular

8:21

right both contemporaneous and like macro

8:23

news not just contemporaneous. In recent

8:25

years, there has always been a

8:27

factor that drives risk. That's a

8:29

very good question. So sensitivity to

8:31

macro news clearly has been always

8:34

there, right? I mean, you turn

8:36

on Bloomberg and they discuss whatever

8:38

macro release there is on that

8:40

date, but the sensitivity varies over

8:42

time. So some of our research

8:44

actually points in the direction of

8:46

when we would expect macro news

8:48

to matter. So we find that

8:50

macro news matter more during periods

8:52

of kind uncertainty. Right. So that's

8:54

when markets are learning more through

8:56

what we call public signals or

8:58

common signals, right? So you essentially

9:00

pretty much putting a larger weight

9:02

on macro signals. So high uncertainty

9:05

recessions disappear. It's where macro news

9:07

tend to matter more. Now we

9:09

also find. In our world, that

9:11

it's lag macaronous riding contemporaneous. So

9:13

that's truly the novel contribution of

9:15

a lot of my research is

9:17

to focus on lag riding contemporaneous.

9:19

So that's something that was missed

9:21

in the literature for a long

9:23

time. For example, we find that

9:25

lag macaronous account in sample, that's

9:27

not a half sample, for about

9:29

50 to 60% of the fluctuations

9:31

of exchange rates at monthly and

9:33

quarterly frequency, is because actually this

9:36

lag pardon and contemporaneous news. So

9:38

the reason why often... both academic

9:40

research, but often also financial practitioners

9:42

focus on contemporaneous, is because they

9:44

say, oh, like information is priced

9:46

in. But that's not necessarily the

9:48

right way to think about it,

9:50

because at every single point of

9:52

time, markets are learning about the

9:54

state of the economy. So let

9:56

me give you a concrete example.

9:58

So imagine you have CPI reduced.

10:00

and then you find that the

10:02

inflation number was higher than the

10:04

consensus, right? We have a positive

10:07

surprise. In order for you to

10:09

interpret... What the impact of this positive

10:11

surprise will be on any asset price

10:13

doesn't have to be affects, it has

10:15

to be stocks, you know, fixed income.

10:17

You need to take it in context

10:20

to what was the latest GDP release,

10:22

right? And was the latest GDP surprise

10:24

on non-farm payroll. So why am I

10:26

saying that? Well, for example, if we

10:28

live in a world where we have

10:31

a positive CPI surprise today, but then

10:33

in the last quarter we had a

10:35

negative GDP surprise, markets extrapolating based on

10:37

this information regime. Right. So you're using

10:39

today's releasing context to previous releases to

10:41

figure out is the variance of supply

10:44

shocks bigger than demand shocks, right? Is

10:46

the shop persistent or not? So we're

10:48

constantly learning, right? So in economics, formally

10:50

you're going to write a patient updating

10:52

model to do that, but the model

10:55

is so complex that clearly no financial

10:57

market participant is kind of solving any

10:59

model, but they're doing it in the

11:01

back of their head, right? They're trying

11:03

to figure out, okay. What does this

11:05

surprise mean in the context of kind

11:08

of the latest news or the history

11:10

of news? So the moment you start

11:12

learning about what we call second or

11:14

the moment, not just the mean of

11:16

supply and demand, right, but also the

11:19

variance, you know, persistence, then the lax

11:21

matter, right? Because you're putting a weight

11:23

on the lack surprises. In a sense,

11:25

that's what kind of the model generates,

11:27

and that is one of the justifications

11:29

why LACS is so important. Now, the

11:32

question arises, can you make money out

11:34

of it, right? So the answer is

11:36

that to the extent that the interpretation

11:38

effectively changing the interpretation of the LAC

11:40

surprises, if you wish, because you decide

11:43

that now we are in a supply

11:45

right in demand region, right? Now. In

11:47

order for you to make money out

11:49

of it, granted that the interpretation changes

11:51

due to the new information that arrives

11:54

today, right, is the interaction of the

11:56

two, you need essentially markets not to

11:58

be good at optimally updating their beliefs,

12:00

right? So they need to be rational

12:02

in some way, which clearly it is

12:04

the case. No one can solve this

12:07

complicated basis. actually it's hard for us

12:09

to solve it. We need to use

12:11

some tricks to do it in our

12:13

research. So granted that markets are not

12:15

kind of doing the optimal Bayesian updating,

12:18

yes, you can exploit that in order

12:20

to effectively develop forecasting models based on

12:22

this landmark on use, which I have

12:24

started doing. So there is some hope

12:26

then in terms of finding proper strategy.

12:28

Yes, I started exploiting that. Is the

12:31

work course of this rate differentials, so

12:33

carry trades, is that the engine one

12:35

would look at, you know, because then

12:37

you're looking to uncovering interest like parity,

12:39

you carry trades? Differentials definitely play a

12:42

role and usually people Think about interest

12:44

rate differential from the fixed income market,

12:46

right? In the sense that you have

12:48

that, for example, the interest rate in

12:50

the US increases relative to Europe, and

12:52

then people doing the carry trade, they

12:55

borrow in Europe, you know, now the

12:57

care trade becomes more profitable, you borrow

12:59

in Europe, you know, now the care

13:01

trade becomes more profitable, you borrow in

13:03

Europe, you borrow in Europe, invest in

13:06

the US, you put market pressure in

13:08

the door from the fixed income market.

13:10

arbitrage conditions, where is this going to

13:12

show up? It's going to show up

13:14

as a link between the current city's

13:17

premium and interest rate differential, which of

13:19

course has been shown to exist, right?

13:21

So you need to find the same

13:23

relationship between exchange rates and interest rate

13:25

differential. But the more intuitive way to

13:27

think about this for me, the carry

13:30

trade in particular is by thinking about

13:32

dispositioning models, right, and how market pressure

13:34

due to changing the positioning with respect

13:36

to doors coming from the fixed income

13:38

market is putting an upper passion on

13:41

the door, for example. Okay, so in

13:43

terms of that position then, let's delve

13:45

a bit more deeply into that. So

13:47

the second part of my research agenda

13:49

focuses on effectively how do we think

13:51

about holdings and positioning if you wish

13:54

in financial markets is drivers of asset

13:56

prices. There I developed a completely double

13:58

decomposition, which actually first it was I

14:00

developed it with the goal of studying

14:02

effects, but it just so happens that

14:05

it works beautifully for equities as well.

14:07

So now I have a lot of

14:09

research on equities as well, but they

14:11

did there is. Actually, let me start

14:13

a little bit with a brief description

14:16

on, you have to explain how it

14:18

works well, which is in order to

14:20

be able to explain how it works

14:22

for effects. And there is also a

14:24

mapping to fixed income as well. But

14:26

the idea there is you take your

14:29

what we call market theory and condition

14:31

nominal supply has to economical holdings, right,

14:33

or nominal demand. So in a market-based

14:35

economy, this has to hold. If you

14:37

observe the universe of holdings of a

14:40

particular stock, let's take in this case,

14:42

Apple, or given ice and for Apple,

14:44

then. If you have the universe of

14:46

holding swap, this has to add up

14:48

to nominal supply. What is nominal supply?

14:50

It's the price of apples times the

14:53

shares issued. Right. Now the shares issued

14:55

at IC and level is pretty much

14:57

driven by buybacks and it doesn't explain

14:59

much of the variance if you wish

15:01

of the price itself. It's not a

15:04

very volatile variable. So the nominal supply

15:06

is just the change in the price

15:08

of apples. which is kind of what

15:10

we're trying to explain. So that's what

15:12

you have on the left hand side.

15:14

What do you have on the right

15:17

hand side? Well, we have the change

15:19

in holdings which are summed up across

15:21

every single investor that buys Apple. So

15:23

this would be from hedge funds to

15:25

pension funds to mutual funds, etc. Right.

15:28

Now you can decompose this change of

15:30

holdings. can be broken down into components

15:32

with very intuitive interpretation that then later

15:34

can lead you into the direction of

15:36

thinking about causality and how do we

15:39

develop investment strategies etc. So what are

15:41

those components? The price of Apple can

15:43

be moving. because of rebalancing, right? Portfolio

15:45

way changes. So this was one of

15:47

the components of the change in holdings.

15:49

The second component is the growth rate

15:52

of assets in the management. So imagine

15:54

a given fund now increases their assets

15:56

in the management on average, their long

15:58

Apple, and then of course this is

16:00

going to also put upward pressure for

16:03

Apple, the most mechanical thing being index

16:05

funds, right? Money is fleeing into index

16:07

funds, which is putting an upper pressure

16:09

on Apple. The growth rate of assets

16:11

in the management can be further broken

16:13

down into the net of the net

16:16

of returns component. So essentially you make

16:18

profits on your portfolio that get to

16:20

invest. including in Apple, right, or final

16:22

inflows and outflows into the fund. So

16:24

those are the two components that indeed

16:27

have very intuitive interpretation. And there is

16:29

a four subcomponent of holdings, which is

16:31

exchange evaluation effects. That's why the exchange

16:33

rate is going to enter the story.

16:35

And it captures that you have used

16:37

the or funds, let's say, buying Brazilian

16:40

stocks, and they need to exchange the

16:42

currency in between putting upward pressure potential

16:44

on the currency. Now. If you have

16:46

all the holdings, you can construct those

16:48

components and then you can test various

16:51

theories and kind of discuss what are

16:53

the drivers of equities etc. So the

16:55

revelation that I had in my research

16:57

is actually even if we focus only

16:59

on mutual funds where we have the

17:02

best data, right? We also have very

17:04

good data coming from the U.S. using

17:06

let's say SEC 13F also on not

17:08

just mutual funds but hedge funds to

17:10

essentially everyone that holds equity. It's a

17:12

bit trickier. to use that data to

17:15

decompositions because you don't necessarily see the

17:17

final in close and the net of

17:19

your returns, which you see in the

17:21

mutual funds. So there are the regulatory

17:23

forms and also companies like Morningstar. And

17:26

the mutual funds you just get from

17:28

some database that provides these sorts of

17:30

things. It's Morningstar. So it's morning star.

17:32

So we use Morningstar direct, which reports

17:34

essentially the variables that you need in

17:36

order to construct the following, right? So

17:39

you have a. limited coverage. For the

17:41

US, we have 15% of the coverage,

17:43

right? So what they mean by coverage,

17:45

cauldings relative to the market cup of

17:47

a particular icing. If you go to

17:50

other countries, we have 33 countries and

17:52

20,000 stops, right, icing. So it is

17:54

a global model. You have even more

17:56

limited coverage for China. We have 0.01%

17:58

often. So the idea here is that,

18:00

well, even with this limited coverage and

18:03

even. if we observe only mutual funds

18:05

and again here we focus only on

18:07

a sample starting 2008 which is the

18:09

period when mutual funds growing importance right

18:11

so you're not going to get as

18:14

much of a good fit if you

18:16

kind of do it in the 90s

18:18

of course when mutual does it. management

18:20

industry wasn't so prominent in the equity

18:22

space. So starting in 2008, the observations

18:25

I had is like, well, wait a

18:27

second, we might have limited coverage, but

18:29

if the mutual funds are representative of

18:31

the behavior of the investors we don't

18:33

observe, then you can take their holdings

18:35

and scale them up as if you

18:38

have a representative sample. So what do

18:40

I mean by this? So the good

18:42

news is that we observed the pension

18:44

funds and insurance behavior to mutual funds

18:46

to accommodate. pension funds investment, right? So

18:49

you kind of have also those big

18:51

players indirectly in that data, if you

18:53

wish. And the case, you're going to

18:55

split your funds by type. What do

18:57

you mean by type? So let's take

18:59

large caps, US stocks, active funds, certain

19:02

frequency of rebalancing. So you can have

19:04

a granular bucket, especially for the large

19:06

stocks. When you go to small stocks,

19:08

you don't have that many funds. So

19:10

you can't go that granular. But for

19:13

Apple, you can go really granular. And

19:15

then what you do is say, well,

19:17

let's take each component, let's say the

19:19

portfolio weight changes, right? We can always

19:21

break down for every single funds holding

19:24

the Apple, the portfolio way change of

19:26

fund I with respect to Apple. I

19:28

can decompose it as an average within

19:30

this narrowly defined bucket of funds and

19:32

residual, right? You can always do that

19:34

simple decomposition. Now, this average that I

19:37

can construct based on the sample that

19:39

I observe, as long as it's... fairly

19:41

closely aligned with the population average for

19:43

the funds that belong in that bucket,

19:45

then effectively I can scale up this

19:48

average as if I... observe the universe

19:50

of holding. So does the idea here?

19:52

It turns out that they use incredibly

19:54

balancing, which is very volatile, right? So

19:56

if you look at our plots, you

19:58

see a cloud of dots for the

20:01

portfolio wage change. And the reason why

20:03

it's the average that matters so much

20:05

is because the way to think about

20:07

it, is because the way to think

20:09

about it, is this is the information

20:12

that is contemporaneously priced in by the

20:14

whole market, right? So we all observed

20:16

the end of quarter report, and we're

20:18

going on. forecasting the price of Apple

20:20

going forward. But that information is not

20:22

part of the information set of everyone

20:25

trading today, so it's not going to

20:27

move the prices up all today, right?

20:29

That's why contemporaneously it's going to be

20:31

this portfolio wage change. Now you do

20:33

that for the flows and the net-offier

20:36

returns of components, you carry it up.

20:38

And now it turns out that across

20:40

33 stock markets and 20,000 ices, contemporaneously

20:42

you can count for about 90% of

20:44

the variation of this 20,000 stock prices

20:47

at monthly frequency. 9-0. Okay. Yes, it's

20:49

striking. I've tried to kill it. It's

20:51

actually now when I present this evidence,

20:53

I have to go and show China

20:55

and a stock with a 0.01% coverage

20:57

and show when it doesn't work, right?

21:00

Because no one, I mean, I actually

21:02

want, I can't show you some plots,

21:04

but I don't think we'll have the

21:06

time, but the fit is truly striking,

21:08

especially for the big ice, the fit

21:11

is perfect, right? And again, the idea

21:13

is this, these guys, these guys, these

21:15

guys are representative, the representative, right. Presumably

21:17

the morning star data or this all

21:19

this flow data comes with a lag

21:21

in like is the point in time

21:24

data you're looking at or are you

21:26

looking at contemporaneous decomposition and the way

21:28

I have used it in in order

21:30

to develop investment strategies based on that

21:32

right because the contemporaneous relationship is not

21:35

what you're going to use in order

21:37

to kind of develop investment strategies is

21:39

to essentially exploit the heterogeneity in objectives

21:41

among the different players in the markets.

21:43

What do I mean by this right

21:45

you have some a set of funds

21:48

that are going to be contrary and

21:50

some are going to be momentum. You

21:52

have the index funds versus active. So

21:54

by playing the heterogeneity in objectives, you

21:56

can exploit lacked information in these components.

21:59

And I have developed strategies with a

22:01

sharp ratio of 20% return market neutral

22:03

based on this. That is not going

22:05

to be made publicly available. But it

22:07

is possible to use the sub components.

22:10

And again, to give an example that

22:12

probably most of your viewers are going

22:14

to be familiar with coming from the

22:16

fixed income space, right? So how do

22:18

people explore? the heterogeneity in objectives and

22:20

preferences there. Clearly, when the central banks

22:23

are doing Q&QT, you know, investors are

22:25

front running, for example, the purchase of

22:27

the actual. long-term ground bonds. You're exploiting

22:29

the presence of pension funds that have,

22:31

you know, liability-driven investment and they're going

22:34

to hedge. So this has been exploited

22:36

massively in the fixed income space. This

22:38

is equivalent in the equity space, right?

22:40

Because you do have a lot of

22:42

heterogeneous, you have guys that are doing

22:44

fundamental assets, you have guys that are

22:47

momentum reversal, etc. And that's how you

22:49

can exploit. That's truly positioning from the

22:51

equity space, right, if you wish. And

22:53

then the papers, for those of you

22:55

that might be interested in kind of

22:58

stylized facts coming from this decomposition, because

23:00

then you can take the decomposition and

23:02

ask the question, okay, which one of

23:04

the components drive asset prices, right? So

23:06

if you take IS and level asset

23:08

prices, indeed, we find that it's the

23:11

portfolio way changes that are the main

23:13

driver of essentially individual level asset prices.

23:15

But when you go to the stock

23:17

market level. the S&P versus, you know,

23:19

European indices etc, across 33 stock markets,

23:22

the picture changes dramatically. So what happens

23:24

is all of a sudden for S&P,

23:26

right, or for the US stock market,

23:28

the main driver is no longer the

23:30

portfolio wage changes. It's actually the net

23:33

if he returns and the final inflows,

23:35

the final inflows account. for a small

23:37

fraction of the variation is the net-of-reterners

23:39

that drive most of the duration, but

23:41

they're always significant. So the world that

23:43

we live in for the US stock

23:46

market is you are in a risk-com

23:48

period, money of fleeing fixing funds, going

23:50

into equity markets, which gets amplified by

23:52

the profits generated by equity funds that

23:54

invest in the S&P, right? So this

23:57

is the main driver of- So this

23:59

is the growth of the AOM side

24:01

of things, is driving the- The index.

24:03

The growth of the AOM, but the

24:05

original shock, the exerting shock, comes from

24:07

the final influence into the funds because

24:10

the neatory deterrence is just... function of

24:12

prices itself, right? So it has an

24:14

amplification effect, if you wish, where the

24:16

original source of variation in that model

24:18

is going to be from the final

24:21

inflows into the funds and whatever the

24:23

active guys are doing with the portfolio

24:25

weight changes, right? They can be creating

24:27

one fundamental use, etc. The rest, you

24:29

can map to prices and you can

24:32

kind of solve it out once you

24:34

write the structural model. So, but then

24:36

when you go to Brazil, Turkey... Even

24:38

some other advanced economies, the portfolio changes

24:40

remain an important driver, right? The netafel

24:42

returns disamplification effect is still present. The

24:45

final inflow is always significant, which is

24:47

good because that's where the exosional variation

24:49

comes from. But then the portfolio changes

24:51

is too important. So. Then because we

24:53

have the micro to macro decomposition, so

24:56

that's kind of the first paper to

24:58

truly go across countries and do the

25:00

micro to macro, going from Iceland level

25:02

to stock market, even in, you know,

25:04

in the academic financial literature, we can

25:06

truly decompose why the picture changes, right?

25:09

How come for the year stock market

25:11

at an Iceland level with 60% portfolio

25:13

wage changes and all of a sudden

25:15

the portfolio which changes even slightly negative.

25:17

asset rebalancing. So what do you mean

25:20

by this? When you go to the

25:22

aggregate stock market, what changes is that

25:24

it matters which stock do you buy

25:26

when you sell Apple, right? So for

25:28

the US, investor state within the dollar

25:30

currency market, right? Within the US stock

25:33

market. So if you buy one US

25:35

stock, they're going to sell another US

25:37

stock. As a result, they prefer the

25:39

way changes. Even their, the main driver,

25:41

Iceland will simply cancel out when we

25:44

aggregate all the different stock. because you're

25:46

not moving across. So the cross-air was

25:48

distancing, is if you wish, a negative,

25:50

right? When you go to Brazil and

25:52

Turkey, what happens? You say, well, just

25:55

talk in Brazil and you buy Turkey.

25:57

So you go across currency borders, right?

25:59

And that is kind of the reason.

26:01

Even at the ice and level of

26:03

the stock market level. Just for the

26:05

point for our audience, because I'm just

26:08

making sure that they. they follow this.

26:10

So what you're saying is with US

26:12

markets, in particular, what you find is

26:14

that the single company level, like the

26:16

apples and so on, is the portfolio

26:19

weight changes that make all the difference.

26:21

But when you aggregate up to the

26:23

S&P level, what matters more is the

26:25

growth in the AUN, the fund flow,

26:27

the net profit move between asset classes,

26:29

like bonds to equities, equities to bonds.

26:32

It's what drives, that's what drives, that's

26:34

what drives, that gets. amplified with essentially

26:36

the profits generated by the funds. But

26:38

then with EM countries like Turkey and

26:40

Brazil it's more across countries than suddenly

26:43

the switch happens rather than within US.

26:45

Indeed the portfolio weight changes remain an

26:47

important driver. Yeah. Now What about effects?

26:49

Right? So effects is very interesting in

26:51

that space, especially when you go outside

26:53

of the US stock market. So when

26:56

you draw a picture and then we

26:58

have this cool network plot section the

27:00

follow-up paper, which is on exchange rates

27:02

and currency centrality, where essentially we show

27:04

that the US clearly is the main

27:07

player in town, more so than fixed

27:09

income. So if you look at equity

27:11

markets, it's US dollar funds that pretty

27:13

much buy global equity markets. Right. So

27:15

even if you take UK. their significantly

27:18

smaller amount of assets in the management

27:20

denominated in euros and pounds invested in

27:22

equities. So the US is kind of

27:24

the center, right? So US funds, US

27:26

dollar funds are buying all global stocks.

27:28

That's where the FX role comes in

27:31

and the pressure into. So the next

27:33

thing we've done, and before I'm moving

27:35

to how do you formally map this

27:37

to exchange rates, even these decompositions, here's

27:39

implications for exchange rates due to this

27:42

exchange evaluation effect, I didn't mention, right.

27:44

So what does the exchange rate component

27:46

do in terms of being a driver

27:48

of equities? And the implications here are

27:50

the same at an IC&N and aggregate

27:52

stock market level. So let me discuss

27:55

them at the aggregate stock market level.

27:57

So we find that for every single

27:59

stock market market besides the door, Japanese

28:01

here. Swiss francis mixed back and then

28:03

every currency packed to the dollar, right?

28:06

What happens is that both the currency

28:08

and the stock market moves in the

28:10

same direction. So what do you mean

28:12

by this? When the Brazilian stock market

28:14

appreciates, it's also that the real is

28:16

going to appreciate, right? But that's not

28:19

the case, of course, for the dollar.

28:21

The usual cycle, what is happening now,

28:23

but the usual global financial cycle, now

28:25

it's a US cycle. The usual global

28:27

financial cycle is that all risky assets

28:30

go up and then the door

28:32

depreciates. So what that means is,

28:34

is if you want to... take

28:36

your more macroeconomic hat is so

28:39

as an investor clearly you're benefiting

28:41

both if you're a US dollar

28:43

investor both from the real appreciating

28:45

and the stock market appreciating local

28:48

currency if you're going to buy

28:50

a Brazilian stock. What does it

28:52

mean from a kind of more

28:54

big picture macro point of view?

28:57

It means that the fact that

28:59

the exchange rate is moving, helping

29:01

equilibrium, real goods is it makes sure

29:03

that the volatility of the Brazilian

29:05

stock market in local currency is

29:08

not as high. So it dampens

29:10

the volatility of the local stock

29:12

market. That's why when we calculate

29:14

these scaled covariances when we do

29:16

these various covariance compositions and study

29:18

what fraction of the S&P price

29:20

movement and the Brazilian stock market

29:22

price movement is due to the

29:24

set components. The exchange rate component

29:26

contributes negatively. It dampens the

29:29

volatility of the local stock market

29:31

in local currency, right? kind of the

29:33

exchange rate component and say something more

29:35

concretely about exchange rates because this is

29:37

about, you know, exchange exchange rate component

29:39

is a basket of currency, right? We

29:41

want to study dollar euro, dollar pound,

29:43

we want to study bilateral currencies. Actually,

29:45

you can do that, right? Because you

29:47

have now a system of equations, I

29:49

have 33 stock markets, and then the

29:52

equivalent exchange rates, let's say against the

29:54

door that correspond to those 33 stock markets.

29:56

So you have a system of equations that you

29:58

can kind of flip it. and so forth, all

30:01

the bilateral exchange trade. So now you have

30:03

a model where every single bilateral exchange

30:05

trade. And the interesting thing is that a lot

30:07

of the things that I tell you, essentially everything

30:09

applies not just to door base, right? So

30:11

actually, particularly we study also the euro base

30:13

as well. We can do it for any

30:15

base. You can express any exchange trade

30:17

as a function of what we call net supply

30:20

components that I explain what they are, and loadings

30:22

based on that which you can

30:24

think of it of them as

30:26

elasticisticities. right? That a function of

30:28

the centrality of different stock markets,

30:30

right? So let me just give you

30:32

a concrete example. So what

30:34

is this net supply measure?

30:36

The net supply measure is

30:38

simply the valuation of the

30:40

local stock market. So that's

30:42

kind of the nominal supply

30:45

of equities in Brazilian Rio

30:47

or US dollar, etc. Right. So

30:49

there are stock market specific, which

30:51

is just the price of the

30:53

stock market, minus... the demand for

30:56

that stock market but denominated

30:58

in the investor currency right so granted we

31:00

said that US door funds own everything

31:02

that is going to be primarily

31:04

US door investors demand holdings

31:07

denominated in US doors so to

31:09

the extent the nominal supply denominated in

31:11

Brazil they are real minus these holdings

31:13

that are denominated in dollars if there

31:16

is a gap it's the exchange rate

31:18

that needs to move in order to

31:20

equally pay the market So in that

31:22

example, so the net splice, let's

31:24

say from a US perspective investing

31:27

into Brazil, if the Brazilian rail

31:29

strengthens, then does that mean the

31:31

demand has increased into Brazil? So

31:33

let me just give you a specific

31:36

example to make it a bit clear.

31:38

This is a bit complicated, I think,

31:40

to grasp. But effectively imagine that you

31:42

have, you can measure an exhorgener shock,

31:45

exhorgeness demand coming from the US. you

31:47

know coming from the US for Brazilian

31:49

stocks. The market can clear in two

31:51

ways. One is the Brazilian stock market

31:54

appreciates in local currency, right? And the

31:56

second way is the Brazilian rehab appreciates

31:58

because if the Brazilian have appreciates. Now

32:00

of a sudden the dollar, the dollar

32:03

demand can buy fewer Brazilian stocks,

32:05

so the local stock market doesn't

32:07

have to appreciate. So effectively that's

32:09

what this net supply measures is

32:12

capturing, right? So by zoning into

32:14

the demand nominated in US dollars,

32:16

we are extracting how much the

32:18

currency needs to move in order

32:20

to clear. It's kind of the residual

32:23

claimant. It incorporates the fact that

32:25

the local market in response

32:27

to this exogenous demand shock. But

32:29

then if it's... So either the local stocks could

32:32

increase or the local effects could increase?

32:34

So in a world with fixed exchange rates,

32:36

everything is going to be done by the

32:38

local stock market serving, right? And then actually

32:40

this net supply measure has to be zero or

32:43

close to zero, right? In that world, where you

32:45

have loading exchange rates, actually exchange rates can

32:47

do a pretty big chunk of the job

32:49

in a sense. So then now we have

32:51

a world where every single bilateral exchange trade

32:53

is a function of this net supply

32:56

measures kind of scaled by... particular

32:58

entities that we directly observe, which

33:00

come out of the sub components

33:02

of our decompositions. And then in the

33:04

follow-up paper, which is with Helen,

33:06

we document a lot of novel

33:09

facts around why is the US

33:11

dollar effectively, why does the US

33:13

dollar come off so strongly with

33:15

the global financial cycles, but the

33:17

euro does. Right. So there what

33:19

we are doing is tracing. Why does

33:21

the particular exchange rate move in response

33:23

to the global financial cycle, which we

33:25

measure using this US macro news that

33:27

I mentioned in risk aversion shocks, also

33:30

in that paper we show that the

33:32

global financial cycle, which usually is measured

33:34

as the first principal component of risky

33:36

assets, right? So that's kind of what

33:38

people usually used to calculate these risk-on

33:40

measures. That's where they come from. We

33:42

show that actually can have a much

33:44

more kind of exogenous measure of that.

33:46

We pretty much show that this... Financial

33:49

cycle is driven by contemporaneous and lack

33:51

macro news and risk conversion shops that

33:53

originate from the US due to the

33:55

structure of equity markets, right? The US

33:57

funds own everything so that's why they

33:59

are are very sensitive to these US

34:01

shocks, particularly in flows into equity funds

34:04

that get invested not just in the

34:06

US but also globally. That's kind of

34:08

the big issue. And so what follows

34:10

from that then, so what you're saying

34:12

essentially is that you've established this new

34:14

sort of flow factor that helps explain

34:16

the US centrality. So what follows then

34:18

is if Europe suddenly established ended up

34:20

with the big equity market in the

34:23

world and they were the biggest investors

34:25

in the world, then you end up

34:27

with the euro global cycle, which isn't

34:29

really likely to happen. But what you've

34:31

done is you've added some meat or

34:33

theory to this empirical observation of US

34:35

driving everything. Indeed, and then you

34:37

can use, also we have a lot

34:39

of facts where we show that actually

34:42

focusing on just the local net supply

34:44

and the door is enough to account

34:46

for most of the migration, so you

34:48

can significantly reduce the dimensionality of the

34:51

problem, etc. So it has a lot

34:53

of, essentially... And just going back to

34:55

that Brazil example, you know, in terms

34:57

of what takes the pressures of the

34:59

equity market or the currency, doesn't need

35:01

to move. basically all the pressures gone

35:04

into the empty market which is feels

35:06

kind of counterintuitive because you you kind

35:08

of think that okay if the empty

35:10

market's going to go up then FX

35:12

also has to go up but what you're

35:15

almost saying is actually that doesn't

35:17

have to be the case. Yeah so so

35:19

in a sense what's missing here in fixed

35:21

income so we do need to talk about

35:23

fixed income and let me kind of close

35:25

the the story by telling you what needs

35:27

to happen in fixed income to observe what

35:29

we're happening in equity equity markets. So if

35:31

fixed income is a crucial driver driver And

35:33

again, one needs to do additional work to

35:35

truly nail the causal effect and we are

35:37

going in that direction here. But if the

35:40

fixed income is the main driver and something

35:42

else is happening in the fixed income world,

35:44

it doesn't have to be the case indeed

35:46

that the exchange trade appreciates. It could be

35:48

that just the local stock market

35:50

essentially is appreciating in response to

35:52

exogenous demand for Brazilian stops coming

35:54

from the US. Now the fixed income world

35:57

is very interesting because it's the flip side

35:59

of the equity. So what do we mean

36:01

by this? So in the currency centrality

36:03

paper, what we find is that there

36:05

is a very interesting sorting between how

36:08

much this net supply component responds to

36:10

the cycle shocks, right, which translates into

36:12

how much the exchange rate responds to

36:14

the cycle shocks that we capture, correlates

36:16

very strongly with the openness of the

36:19

stock market. So the more open the

36:21

equity market is, and the Brazilian stock

36:23

market is very open, right? So most

36:25

of it is held by U.S. door

36:27

funds of locally issued equities, the more

36:30

the net supply component's responses is the

36:32

result the exchange rate is loading on

36:34

this global financial cycle, right? Now, let's

36:36

think about, and of course, for the

36:38

U.S. it's kind of the reverse. The

36:41

net supply is kind of the shifting

36:43

in the opposite way. The U.S. is

36:45

close to a close to the market.

36:47

90 percent of U. US dollar denominated

36:49

equity funds. So that's kind of the

36:52

closest to a close if you wish

36:54

equity market there is. That's another unique

36:56

property. The size plus the closeness of

36:58

the two very unique properties that are

37:00

needed in order to make the door

37:03

a good hedge for the global financial

37:05

cycle. Actually, that's explicitly made in the

37:07

paper. What about the fixed income market?

37:09

It's actually quite interesting because the two

37:11

main markets that exchange rates are going

37:14

to clear will be the fixed income

37:16

and the equity market. So the real

37:18

good side that's kind of one year

37:20

and above frequency because we have sticky

37:22

prices. So we don't have to worry

37:25

about it that much. So the fixed

37:27

income market is the flip side in

37:29

terms of openness. So what do you

37:31

mean by this? So the US equity

37:33

market is very close, but the US

37:36

fixed income market is super open. Why?

37:38

Because we know that in normal times,

37:40

US long-term government debt appreciates in bad

37:42

times and the door appreciates. So it's

37:44

the best catch there is. So foreigners

37:47

love to hold US government debt and

37:49

actually they flooding to US government debt

37:51

whenever we have, like to safety episodes,

37:53

right? So Japan is also a fairly

37:55

close stock market, right? Not as close

37:58

as the US, but it's one of

38:00

the more closed. stock markets there is.

38:02

But what about Brazil? What about Turkey,

38:04

right? The market economies. So when you

38:06

look at their fixed market, fixed income

38:09

market, they of course have, in most

38:11

cases, bigger fixed income market than equity

38:13

market, right? The equity markets are not

38:15

that developed. But a big fraction of

38:17

this fixed income market is issued not

38:20

a local currency. It's issued in dollars.

38:22

It's issued in euros. Why? Because Pimko

38:24

doesn't want to hold. local currency. They're

38:26

going to buy euro or door. So

38:28

what is issued in local currency from

38:31

the fixed income market? A lot of

38:33

that is actually internally helped by Brazilian

38:35

investors, if you wish. So actually, if

38:37

you just think about the fraction of

38:39

the fixed income market issued in local

38:42

currency, which is what matters for the

38:44

exchange trade, right, at the end of

38:46

the day. Because we want to think

38:48

about Brazilian reality versus doors. Actually, the

38:50

fixed income market in emerging market economies

38:53

is fairly closed. So it's the flip

38:55

side of equity. So what the exchange

38:57

trade is doing is moving in order

38:59

to equilibrate the stock markets and the

39:01

fixed income markets that are relatively more

39:04

open. So it just so happens that

39:06

in the case of Brazil, the equity

39:08

market is very open, and the fixed

39:10

income market counting only fixing commission in

39:12

Brazilian currency is closed. So it is

39:15

accommodating the equity market. In the case

39:17

of the US. the exchange rate is

39:19

clearing the fixed income market, right? And

39:21

that's why usually you have this negative

39:23

comm movement between long-term government debt and

39:26

stocks in the US. So there is,

39:28

everything has to hold in equilibrium, right?

39:30

All the markets need to clear and

39:32

the exchange rate is entering all this

39:34

market. So again, there's way more work

39:37

that needs to be done and we

39:39

need to do in terms of getting

39:41

truly causal instruments in some for demand

39:43

shocks, etc. This is the picture of

39:45

the world that truly kind of explains

39:48

the data and maps well to the

39:50

data. So fixed income markets are interesting,

39:52

but what we are pushing is that

39:54

actually equity markets are in some ways

39:56

even more interesting. particularly when we want

39:59

to study as well Brazil and effectively

40:01

emerging markets and some small advanced economies

40:03

as well. Another thing to keep in

40:05

mind is fixed income markets, often the

40:07

exchange rate risk is hedged, right? Even

40:10

if we think about euro dollar kind

40:12

of transaction, so US dollar funds buying

40:14

euro bonds and the other way around.

40:16

Usually it's euro funds buying, you know,

40:18

US dollar fixing income assets. So if

40:21

you, for example. putting pressure on the

40:23

door coming from euro, right? European investors

40:25

are buying essentially US diagram and bonds,

40:27

but they're not doing it for carry

40:29

purposes, right? They just want to hold

40:32

their underlying return in dollars. Then they're

40:34

going to hedge the exchange rate risk.

40:36

So you're putting a upper patient on

40:38

the door coming from the fixed income

40:40

market, but then in the derivatives market

40:43

you're putting a pressure on the door

40:45

from the opposite direction, right? So it

40:47

ends up being neutral then because it's

40:49

currently hedged more close, right? Well with

40:51

equity markets usually there's no hedging. It's

40:54

currently unhed. Correct. That's why the causality...

40:56

It's my prior and we do have

40:58

some evidence that even if you have

41:00

like really exogenous calls of instruments that

41:02

you construct, right, of demand shocks, which

41:05

is not trivial to do, but you

41:07

know we've done some work and Helen

41:09

has done some work in the bus,

41:11

actually is going to go through equity

41:13

markets and also we do have another

41:16

paper studying the universe of effects derivatives

41:18

transactions in the UK. We're the first

41:20

to clean that massive data set and

41:22

indeed it is the case. That's what

41:24

we find. Fixing fixing. and not equity

41:27

funds, right? And when you go also

41:29

to pension funds, there's massive hedging, right?

41:31

They are truly hedging the effects risk.

41:33

And they are loaded on fixed income.

41:35

So that's why equity markets are more

41:38

interesting, in my opinion, when we study

41:40

effects, because that's where exogenous demand shocks

41:42

are not hedged and are putting the

41:44

pressure on the effects. And then importantly

41:46

you've come up with this global framework

41:49

to look at flows in a clever

41:51

way to see the transmission mechanism from

41:53

equities flows to FX which is very

41:55

clever and then equally on the on

41:57

the risk premium side. The way you've

42:00

looked at macro expectations and surprises by

42:02

looking at lagged versions of those variables

42:04

also shows, at least in sample, that

42:06

it tends to work quite well. So

42:08

this all fantastic work and I think

42:11

all of it as you say can

42:13

be applied for investors if people think

42:15

about it more carefully and and hopefully

42:17

at some point you'll be able to

42:19

reveal some of your your secret source

42:22

to us all. But before we sort

42:24

of round off I do like to

42:26

ask my guest something on just more

42:28

on the non-academic side you know one

42:30

is we do have some young people

42:33

listening to the podcast as well you

42:35

know many of them out at university

42:37

at the moment and many of them

42:39

will graduate this year or have graduated.

42:41

as they enter the post-university world? Or

42:44

maybe they should stay in university and

42:46

do PhDs. That's a challenging question in

42:48

an AI world. In a sense, to

42:50

be honest, what I'm hearing a lot

42:52

is I don't need to hire a

42:55

research system. I don't need to hire

42:57

a junior. So I mean, again, I

42:59

don't want to scare the young of

43:01

yours. I think one way to... differentiate

43:03

themselves and to kind of have a

43:06

successful career because I do believe it's

43:08

going to get more challenging to get

43:10

the nice jobs where you get to

43:12

work with someone more senior where you

43:14

can learn and be trained etc. The

43:17

business model will have to change. So

43:19

They clearly have to become experts on

43:21

how to utilize AI in their work,

43:23

but what I tell my NBA students

43:25

and in management students that are kind

43:28

of the younger group is that for

43:30

you to use AI effectively, you need

43:32

to be already an expert in something,

43:34

right? So in a sense, if you

43:36

go to child GPT and ask a

43:39

question, you know, to the extent that

43:41

the answer can be useful and, you

43:43

know, can add value to your work,

43:45

you need to understand what on earth.

43:47

actually chirgypity talking about, right? The more

43:50

interesting discussions with chaggypity are the higher

43:52

level discussions. So in a sense, clearly,

43:54

you know, your your viewers are kind

43:56

of the more driven and already, you

43:58

know, they don't need to be motivated

44:01

in this way. But the way I

44:03

would advise them is to learn from

44:05

academia, which is again the word I

44:07

come from, how they can take, exploit

44:09

the knowledge in academia and bring it

44:12

into finance in particular, right, in that

44:14

dimension. I think that in academia there

44:16

is 20 to 30% of the work

44:18

that's very exciting. Not everything is very

44:20

exciting. I'm not going to lie, right?

44:23

academia has its problems and there's a

44:25

lot of inertia but there is 20

44:27

to 30% of the research that I

44:29

would say is very exciting. The most

44:31

exciting research is big data but big

44:34

data what I mean by big data

44:36

in finance and economics is essentially taking

44:38

a theoretical model like what I did

44:40

here is simple accounting identity that is

44:42

coming out of theoretical models but then

44:45

going truly to an Iceland level right,

44:47

a granular level and aggregating from the

44:49

bottom up. What tends to work best

44:51

is aggregating from the bottom up. So

44:53

reading that kind of research, and there

44:56

is a lot of exciting stuff happening

44:58

now, right? So people are using not

45:00

just this holdings data, but you know,

45:02

some people have more, for example, the

45:04

universe of holdings in the fixed income

45:07

space, which is much much harder, because

45:09

clearly need the central banks, you need

45:11

all the market makers because of 40C

45:13

transactions, you can go only so far

45:15

with mutual funds. So that's a very

45:18

excited area of kind of research. The

45:20

propagation of news, right now the AI

45:22

technology allows us to any kind of

45:24

index using annum, right? So the propagation

45:26

of news and here not narrowly restricted

45:29

to macro surprises, which is let's say

45:31

CPI realization minus expectations, but some very

45:33

interesting research by some of my colleagues

45:35

actually in the fixed incomes place is

45:38

let's see how let's say fixed income

45:40

is prices are responding to updates in

45:42

the Congressional Budget Office estimates of the...

45:44

cost to the budget of whatever congressional

45:46

proposal there is. So we can now,

45:49

with the ability to scrape data easily,

45:51

and what I mean is now... even

45:53

young viewers can take advantage of that.

45:55

Like back in the day, I had

45:57

to hire a research assistant to write

46:00

the Python scraping code for me to

46:02

use it, right? Instead of me spending,

46:04

you know, months writing that code. Now

46:06

I can do that with charge you

46:08

between half a day, right? Which is

46:11

what I mean that people don't end

46:13

up hiring research assistance as a result.

46:15

But also that opens the possibility to

46:17

your viewers to go and experiment, right?

46:19

Even if they're not experts on on

46:22

Python coding. I would advise them to

46:24

invest in learning the theory. In particular

46:26

in finance, what's the frontier's intermediation-based asset

46:28

pricing theories? And, you know, they're going

46:30

to find what are the relevant sources

46:33

there, etc. So there's a lot of

46:35

exciting developments there, but if you know

46:37

the theory, then with charge APT that

46:39

is going to help you implement that

46:41

theory in Python, you have much more

46:44

power than someone that has just done

46:46

encoding. and they don't understand how to

46:48

think about financial markets. So they should

46:50

invest in understanding the frontier thinking about

46:52

what drives asset pricing. I think that's

46:55

the most valuable thing. The coding skills,

46:57

yes, you need to still clearly have

46:59

coding experience. Otherwise, even with Judge UPT,

47:01

you know, you can't use it in

47:03

a useful way. But you don't need

47:06

to have fusion computer scientists, right, to

47:08

be able to do incredibly meaningful big

47:10

data. kind of analysis in terms of

47:12

and it's going to be about big

47:14

data right yeah so now the way

47:17

the way I think about it anything

47:19

that is in a textual format can

47:21

be scraped so any if you think

47:23

about discretionary trading so we know that

47:25

inequity already it's about 30% is systematic,

47:28

right? Fixed income is less than that.

47:30

So inequity, the systematic to fundamental based

47:32

kind of, you go download the balance

47:34

sheet and read it, is kind of

47:36

what I think about fundamental based research.

47:39

Systematic is going to keep growing because

47:41

whatever the fundamental analysts are doing, you

47:43

can now automate a lot of that

47:45

analysis. So systematic is going to keep

47:47

growing and important. So being able to.

47:50

do big data, being able to think

47:52

about what is the right model, how

47:54

do you implement it, that is going

47:56

to be crucial going forward. Actually, my

47:58

mee forecast, who knows what I'm right

48:01

or not, but in 10 to 15

48:03

years, especially in the equity space, it's

48:05

going to be mostly systematic plus index

48:07

funds. Of course, ETFs and ETFs can

48:09

be on systematic as well as strategies.

48:12

But even with systematic, there will be

48:14

an overlay always of discretionary, right? How

48:16

much way do I put across systematic

48:18

strategies based on the signal? I think

48:20

that's where we're headed. So not knowing

48:23

big data, not knowing proper theory and

48:25

how to exploit it, because with big

48:27

data you can also introduce a lot

48:29

of noise if you don't know what

48:31

you're doing, right? So it is important

48:34

to start with the theory, the frontier

48:36

theory, and then bring the data. And

48:38

yeah, no, that's great advice. And then

48:40

finally, I just wanted to ask you,

48:42

what's the best way for people to

48:45

reach out to you or follow your

48:47

work, see your work? So my website

48:49

has essentially a lot of the papers

48:51

that we discussed and also there I

48:53

have my CV that is listing my

48:56

Gmail as well. Okay great. I'll include

48:58

a link to that on the show

49:00

notes as well. I'll link it in

49:02

and I'm easy to find. Great. No,

49:04

that's excellent. Yes, with that, you know,

49:07

thanks a lot for this fantastic discussion

49:09

and good luck with all the work

49:11

and especially good luck to the work

49:13

that you're doing on applying it to

49:15

investment strategies directly. You know, I'm quite

49:18

intrigued to see how that goes. So

49:20

yeah, once again, thanks a lot for

49:22

all of this discussion. Thank you so

49:24

much for having me. It's a pleasure.

49:26

Thanks for listening to the episode. Please

49:29

subscribe to the podcast show in Apple,

49:31

Spotify, or have you listened to podcast.

49:33

Leave a five-star rating, a nice comment,

49:35

and let other people know about the

49:37

show. We'll be very, very grateful. Finally,

49:40

sign up for a free newsletter at

49:42

macawive.com. We'll be back soon, so tune

49:44

in then.