Hey guys,

I have a list of firms, for which I want to figure out the amount of business segments at a specific point in time. I found the Compustat business segment file, but I couldn’t find any specific function/variable that counts the segments.

Do I have to define a business segment on my own? For example, one business segment = at least 10% of the combined sales or assets of the company?

Another idea I had: Would it make sense to assume that the number of different SIC codes equals the number of different business segments?

Thanks

hi,

To count the number of segments you need to realize that:
- the company may disclose ‘industry’ segments and/or geographical segments, so you probably want to filter out the geographical ones
- some studies aggregate industries if their sic/naics code is indentical (some don’t)
- some segments may be ‘corporate’ or some eliminations; you probably want to drop segments with negative sales, and probably drop segments without a missing industry code
- segment information is usually repeated; for example, the 10-Ks for fiscal years 2010, 2011 and 2012 will contain segment info for the year 2010, so you need to set a filter on ‘sourcedate’ (for example, set sourcedate equal to datadate)

Digging up a few sample firms and their original 10-K filing on Edgar will probably help to see what is going on.

best regards,

Joost

Hi Joost,

thanks for your answer. I am still struggling with the segments. I have downloaded the Compustat Business Segment file, but have difficulties identifying the number of segments of one firm in a particular year.

In particular, I wonder about the following:
- What is sid (segment identifier from Compustat)? Does one unique sid per permno and fiscal year identify a business segment?
- What should I do with segments that have negative assets and sales (I count only segments that account for at least 10% of sales or assets). Just drop them?
- sourcedate and datadate are always the same in my file. So what do you mean by filter? I agree that some entries look like duplicates (but are not entirely the same though).
- What role play the SIC codes in identifying a business segments. I am also confused about the different types of SIC codes: primary, secondary, standard, etc. Primary seems to be the one with least missing values.

I haven’t found any information about identifying business segments in Compustat on the internet :(

Thanks

hi,

The segment id (sid) is a number to identify segments for a firm so to track (match) them over time; they have no meaning and can’t be compared/matched between firms.

Segments with negative sales or assets are likely to be ‘overhead’ or ‘eliminations’. Often, such segments do not have an industry code.

I did a quick check on srcdate vs datadate. If they are always the same, there is no need to filter. In earlier versions/other sources, they are not always the same. (The distinction has to do with the fact that segment report include multiple years of data. So, segment information over 2008 appears in the annual report of 2008, and 2009, and 2010. Setting a filter such that srcdate equals datadate would give the segment data once (in this example, only in 2008).

Using primary SIC is an accepted (the best?) way of identifying a segment. Primary NAICS would work as well (will give very similar results).

best regards,

Joost