Another Stats Question: high frequency

CFA Level III
#1

If we assume data set 1 is unchanged, say monthly, and for data set 2 use daily instead of monthly, of coz we are going to see lower correlations. However, what if I use daily for both data set 1 and 2? Any one? Should I call my list again? :))))) Thanks guys, you are really helpful!!!

#2

what asset class are we talking about ?

#3

abs(correlation) will be higher if you use daily for both IMHO

#4

>abs(correlation) will be higher if you use daily for both IMHO It depends on asset class. If one of the asset classes has illiquid securities, they are not priced too often - you might actually end up seeing lower correlations.

#5

right - like Real estate for example …daily returns will be quite useless

#6

i thought the question was if daily return exists, what is going to happen…

#7

all of you guys are correct! And I am with comp_kid 1) if we use monthly for 1, daily for 2, correlation will be low. 2) if we have, and do use daily for both 1 and 2, correlation should be high.

#8

what if you have negatively correlated assets, what is going to happen? i am not a stat guy, that was just a guess

#9

The more I though about it: if use both daily data, outcome can be either way. It is only when we have asynchronism, then correlation is low…

#10

a bit confused about this as well, high frequency data itself won’t necessarily lead to low corr, it only happens when we also have asynchronism. right?

#11

totally lost…someone please explain.

#12

This is confusing. But the word “asynchronism” already tells us that correlation is lower. If two things are synchronous, they must be high correlated. The opposite should be true for asynchronous. I am an engineering guy. :slight_smile:

#13

But do they correlated because asynchronous data is skewed? or simply because one set of data is synchronous and the other is not?

#14

if you compare monthly returns of s&p500 and nikkei, correlation will be high if you compare daily returns, correlation can be (should check) theoretically lower, because of the asynchronism (one market opens when the other is closed) I think you have to separate that with “others”, mainly stale pricing of real estate and alternative investments. For those, you just have stale pricing = underestimated volatility = underestimated correlations. You will not have asynchronism with alternative investments