2010 mock q 27: high frequency data across, frequency, understand, 2010

sabina

What does it mean by high frequency data? I don't understand why having data more frequently would lead to lower correlation and be more sensitive to asynchronism across variables

yuoska

Grrr, I know it's too late, but I wanted to post a note on this before the site went into read-only mode...

I think this is what they're saying:
"Increasing the frequency", means you're increasing the number of samples by measuring it more often. Suppose you had been measuring returns for some investment only on a monthly basis (N=12). One day you decide to measure it on a daily basis (N=365), so you're increasing the number of samples for the same time period. That's what they mean by "high(er) frequency data."

Now, an example of asynchronous data is: Let's say you have been measuring daily returns on two indices, one index follows US stock index, and another a LN stock index. After you collect enough data, your plan is to calculate the correlation between those indices. Since holidays in UK and US are not in sync, however, some UK business days could be a holiday in US and vice-versa. Since you need the same number of observations to calculate the correlation, what many people end up doing is "backfilling" the missing returns on US index by taking the previous-day's close, and you do the same with the UK index. That's the best example I could think of...