Using Datastream or IBES for forecast accuracy

Not long ago a student contacted me with regard to the subject of forecasts. Forecast accuracy is defined as the absolute difference between the consensus analysts’ forecasts and actual earnings per share divided by the firm share price at the beginning of the quarter. Differences between actual and forecast earnings can be considered “surprise data”.
His question concerned the variables that he got through Datastream & IBES and which variables should be used. The answer I came up with (with the help of the Thomson helpdesk) was the following:

Datastream has direct data types for the Surprise earnings but this is only available for companies that have quarterly data coverage. Important in this case is, that only the current values of these data types are held – no history is maintained. It is also very difficult to calculate the Surprise data manually because of Datstream padding function.
Regardless of whether you decide to use IBES or Datastream data types, you should not use data types from both sources in the same analysis! Only use IBES data types or only use Datastream data types.

To answer the questions on using actual EPS and EPS forecasts and prices the analysis could be done using the following IBES data types:
EPSI1YR = Earnings Per Share End Date of Quarterly Period INT1
EPSI1MN = Earnings Per Share Mean INT1
I0IND = EPS Last Rep Int Period Indicator
I0EPS = EPS Reported Interim EPS (INT1)

Using the output from IBES (through Datastream) you have to be careful when comparing EPS data as all data should be for the exact same quarter. See below for an example with colours to indicate the quarter data to be compared:


Bankscope cancelled and replaced

Bureau van Dijk (BvD) publishes the Bankscope database. It was announced recently that Fitch, the main provider of the Bankscope data, has terminated it’s contract with BvD. As a result, the availability of bank data will undergo some important changes, with possibly disruptive consequences.  BvD is currently working on a new Bankscope data product: Orbis Bank Focus. However, the coverage of the new database is limited at the moment: historical coverage in this new database is severely restricted, with up to only three years of data (Bankscope has ten (or more) years of history). BvD are prioritizing cross-sectional coverage, and are working on including recent & historical data for banks in different countries. The new product is obviously a serious restriction for academic research.

The current Bankscope database will remain available until the end of 2016, at which point the contract between Fitch and BvD will end, and BvD will need to take down the data. If this affects your research, I advise you to download data as soon as possible.


Filing size of 10-K reports

No too long ago a student asked me if it was possible to find out what the size was for specific filings of companies with the SEC. Specifically it concerned the 10-K filings which are the annual reports with a comprehensive summary of the companies financial health. As I understand it, the file size of these filings was to be used in an exercise and served as a proxy for financial reporting readability.

To find the original filings it is of course possible to use the EDGAR search option to find the original full-text filings and parts (including the XBRL filings):

Using the Central Index Key (CIK), Ticker, or Company name it is easy to find a specific company. Using the Filing Type search box you can narrow the search down to specific filings, like 10-K. This is often possible going back to 1993/1994 (depends on the company).

A specific 10-K filing overview in Edgar would look as follows:

If you need to do this for several hundreds of companies and multiple years it would take some time to collect the file size data. If you have access to the Audit Analytics database it is possible to get the 10-K file size for a large number of companies at once through the WRDS platform. Audit Analytics has a part database that is called Accelerated Filer.

Through this part database it is possible to get filings data (from 2000 to now) using Ticker lists or CIK lists. In the example output below I have put the following variables:

NAME = Company Name
FORM FKEY = Form name (10-K and 10-Q)
FISCAL YE = Fiscal Year End
FILE SIZE = Filing size

Example output using Excel to filter for just 10-K filings:

Important: If you look at the 2015 10-K form file size according to Audit Analytics for the company ADVANCE AUTO PARTS INC this is 17 MB. This corresponds (roughly) with the size of Complete submission file according to EDGAR: 17.748.770. The Complete submission file in Edgar includes not just text and html codes, but may also include pictures and any other file types (Excel files etc.).