DataStream - extracting data (Lot of 0 values returned)

[beniamiin]

Hi,

I'm writing my thesis in finance and I have extracted all my figures to calculate the necessary financial ratios. A lot of the values returned are '0' values (not blank). The one extreme case is with 'Long term debt' - workcode 03251. Almost 25% of the values are 0. I know it could be the case that the firms have no long term debt at all. It just seems unlikely to me that this number is 25% of my sample size. Are these values true and can I base my calculations on them?

I have the same issue with Cost of goods Sold - workcode 01051. Where a lot of firms have zero costs related to the products they produce which also seems unlikely.

I figured it might be because of rounding (this would perhaps mean that if a company only had COGS Of say 400 USD it would appear as 0 when I extract the data through excel. Is there any way I can tell the excel add-inn to extract more decimals or extract data in the exact figures so it would return 400 in the above case? At the moment I think they are in '000 thousands.

Any help is greatly appreciated!

Best regards,
Benjamin

 

[daleholborow]

It's late, and I can't be of any specific help, but I'll offer you two pieces of advice regarding processing of financial data, and Thomson Datastream data in general.

1). Read this, and realise that the chances are the financial data you're dealing with is at best incomplete, and at worst.... bad.

Quant Careers Blog by Dominic: Beware of the Python

2) Read this, and understand that I sympathise with your Datastream-related pain

DataStream: Matching bond pricing data with the issuing company financials? - Forum | Quant Network

In summary, it is entirely possible that a massive proportion of the data you have access to is going to be garbage. Best of luck with your project, and sorry if this comes across overly negative, it's meant to be more of a "don't get disheartened, and try to work around the issue".

 

[Ken Abbott]

Pick 10 companies form the bunch.
Download their 10k's.
Spot check.
If you find one difference, throw your dataset out and use Bloomberg.

 

[beniamiin]

Thanks for your replies - I really appreciate it

BR Benjamin