Event study with multiple events on 1 day

[Davey Bakker]

Hi,

I'm writing my Master Thesis on the effect of player transfers on the stock prices of listed football clubs.

However, when I'm trying to make my date variable a time-series variable (command tsset Date) it gives:
"repeated time values in sample".
I know why I get this error, since I got a dataset with multiple events on 1 day (transfers from 15 clubs over 15 years)
but what should I do to fix this?

I'm trying to get abnormal returns over event periods, but these event periods consist of multiple events.

Your help is appreciated!

kind regards,

Davey

[Joseph Coveney]

Probably naive, but would

tsset football_club transfer_date

help? Or does a club have multiple transfers on a single day?

[Nick Cox]

I'd guess the other way. As the aim is to predict stock prices, your transfer dataset needs to be collapsed or contracted to match.

[Davey Bakker]

Joseph Coveney Yes you are right, clubs also got multiple transfers on single days (mostly first of July because of transfer window);
so I will still get the same error, What else can I do?

Is it possible to add another variable to the tsset code like player name, because then every transfer should be unique (since a player is moving only 1 time a day), right?

I don't have access to stata right now, but could that work?

Nick Cox What exactly happens to the dataset when using collapse or contract? Other than creating a new dataset containing the frequency of transfer on a single day?

[Nick Cox]

On the last paragraph: what exactly happens depends on what you ask for. I know that sounds like a silly answer but it's the only one I can imagine here.

More generally, it makes no sense to define panels as (club, player, date) triples as you will end up with panels that are singleton observations.

[Phil Bromiley]

You need to think about exactly what you want substantively with multiple events on the same day. Often, event studies drop events when there is alternative news on the same day but that is usually news of a different type.

Do you want the observation to be the transfer in which case you might want to duplicate the outcomes for each transfer. Or, do you want to assume something different happens when you have multiple transfers in which case you might want to count transfers or create dummies indicating 1, 2, 3, etc. transfers on a given day.? I'm sure there are other alternatives, but you need to be clear on what makes sense - then after you've taken a shot at programming it, we can help you with remaining problems.