GARCH sample size requirement and suitability

Hello Tino,

I went through your YouTube channel and really appreciate the content you are providing to us.
I am a PhD student in finance area. I got a idea while i was reading a business newspaper. The concept is related to REITs and Infrastructure investment trusts (InvITs). These are pretty new to India (Introduced during 2017-2020 period). We have 5-6 such trusts traded on exchanges. They have fared well during covid-19 and i want to use GARCH methodology to test diversifier, hedge, and safe haven properties of REITs/InvITs against traditional assets such as equity, gold, commodities like Oil etc. However we have only 5-6 companies with data ranging from 1 to 5 years. Can GARCH models are useful for shorter time period. Or do you think i have to apply other methodologies?

Sorry for creating a new thread. I could not post my question in the existing thread.

Regards,
Karthik N

Hi Carthy,

Glad you like the channel.

I would usually say you need about 1000 days of observations in order to get accurate GARCH estimates. That translates to about 4 years of trading data (250 biz days a year).

GARCH is not a magic bullet. Maybe GARCH effects aren’t present…

What are you trying to achieve exactly? GARCH is a tool to measure volatility that is heteroskedastic (i.e. it moves about) and is autoregressive (yesterday will influence today). If those properties aren’t present, then…it’s the wrong tool.

Look forward to hearing back from you.

Tino

Hi Tino,

I am looking at hedge and safe-haven properties of Indian REITS. But since these are launched very recently, we have daily data < 1000 days.

Regards,
Carthy

I mean, give it a go. See if the parameters look right and are statistically significant.

Yeah. I will see.

Thank You,
Carthy

1 Like