0
$\begingroup$

Can anyone provide us with an empirical example (in Python) of market impact visualization, methods of its estimation (in the framework of a parametric model for impact functions) ?

$\endgroup$
3
  • $\begingroup$ this question is a duplicate of a few questions already answered. Please augment your question, citing the other questions and explaining why yours is different. Have a look at quant.stackexchange.com/search?q=%22market+impact%22 $\endgroup$
    – lehalle
    May 8 at 6:13
  • $\begingroup$ @lehalle Thanks for your answer Sir. The thing is that i have LOBSTER data and i am not sure that it contains participation rates. For my case, i did the following : i extracted information of a certain orderID and took only the selling transactions. i have computed the participation rate as : size / ask Volume. then i have plotted the midPrice change against this participation rate. But i didn't notice any particular shape that researchers got in their papers/thesis. Am i computing the participation rate in a wrong manner ? or one cannot compute it using LOBSTER data ?. Thanks in advance. $\endgroup$ May 8 at 14:31
  • $\begingroup$ @lehalle by the way, i am using AMZN data (10-levels LOB) from this link lobsterdata.com/info/DataSamples.php $\endgroup$ May 8 at 14:34

1 Answer 1

1
$\begingroup$

The market impact is the influence of the pressure exerted by the flow of metaorders on price dynamics.

Metaorders are large orders issued in general by asset managers or investment banks; they are then split in small child orders according to an optimal trading strategy.

Nowadays, if you look at orderflows of child orders, you will not see any market impact. Your can see the price impact of one transaction (typically of a market order, ie a liquidity consuming order).

Without a database of metaorders you will not see any market impact. In the past (12 years ago or more), it was possible to infer the presence of a large imbalance of metaorders just looking at the buying or selling pressure at 1min to 5min time scale, but it is not more the case. With fragmentation especially, metaorders are far more protected from detection that in the past.

If you want details, I would recommend either

$\endgroup$
1
  • 1
    $\begingroup$ Thanks for your answer Sir. The thing is, i’ve read most of the theoritical litterature out there. But i’m still learning how to implement stuff in practice. I’ve just began written the book written by you and sophie laruelle, it’s really helping me getting stuff 'in practice’. Now i understand that i need to have acces to private data in order to model and see market impact. $\endgroup$ May 10 at 16:07

Not the answer you're looking for? Browse other questions tagged or ask your own question.