The July issue of Dimensions featured an interview with Emil Efthimides, the regulatory monitor for the Global Data division at Bloomberg LP. He oversees the collection and delivery of financial-statement data for use by investors and financial analysts. Efthimedes works on the front line of XBRL-tagged financial data with Bloomberg’s standardization team, global market representatives, and fundamentals analysts.
In an interview with Dimensions, Mr. Efthimides explains how Bloomberg consumes XBRL data to fast-track company financials to analysts and investors. He emphasizes that filers must pay attention to XBRL data quality to tell their financial story to investors with accuracy, speed, and competitive excellence.
NOTE: The views expressed here are entirely Mr. Efthimides’s, and they do not necessarily reflect those of Bloomberg LP or any other organization.
How long has Bloomberg been using XBRL data from companies’ SEC filings?
We started using XBRL data from US filers in 2014, so this is actually the fourth fiscal year in which we are using that data. We have increased the usage every year. For the most recent annual filing period, the 2016 annual, we processed SEC filings from over 3,000 companies. We extract more content each time. Our usage has been increasing, and we have been getting more out of [XBRL data] every year.
How are you using XBRL data in your products?
We have a number of different steps that we go through when inputting company financials into the system. XBRL is the very first input, normally. It is a machine-readable format, so that can come up faster in front of [our] analysts, before they have a chance to look at the paper filings. It is the first cut of the financial statement that goes up in front of them. The paper filing, the 10-Q or the 10-K, follows after that. We scrape the data from the XBRL, and we put that in front of our analysts, who then complete the financial statement. When the financial statement is completed, partly from the XBRL data and partly from the paper statement, then it will go onto the system and be available live to users of the Bloomberg service.
So Bloomberg analysts see the XBRL data before they see the “paper filing” (which is now actually in electronic form)?
Right. It is scraped by the machines, and it goes on first. Then the analyst will supplement that with data from the paper filing.
Are these internal or outside analysts?
These are internal [Bloomberg] analysts who process the SEC filing and put it live on the system so that our users, many of whom are financial analysts and investors, will be able to see the company’s financial statement on the system.
What are the advantages of XBRL data in your experience?
The main advantage that we see is speed to market, which is really crucial when you are in this business. You want to put financial data out as soon as it is released. XBRL does give you a chance to do that. Of course, you always have to balance speed with accuracy and completeness. Those are other things that we need.
As I said, the financial statement is usually completed from the paper statement [filed on EDGAR]. If we were, theoretically, able to pull 100% of the financial data from XBRL, then we would be able to put it out there quickly, accurately, and completely. Currently, with US filings, we are not able to do 100% of the statement with straight-through processing, but ultimately we think that is the main advantage of XBRL. It will populate data automatically that potentially our analysts could miss. In the US XBRL filings, every number that appears in the filing should be tagged with XBRL tags, even things that are in text, which may be hard for an analyst to locate. So another big strength is that [XBRL] can deliver things which an analyst might miss if he or she were just looking through a paper statement.
Finally, another big thing is that everybody files these statements, the big companies and the small ones too, and this allows the small filers to get their data out on a par with the big filers. If many companies are reporting on a given day, normally our staff is going to be drawn to the bigger filers first, but XBRL allows data from the small filers to get into the system faster. Otherwise it might be delayed—dealt with after all of the big companies have filed.
Do you also find that you are getting more out of the footnotes than you were in the past?
Yes, it is possible that there could be data in the XBRL that we would not have caught in the past. Now we do go through all of the footnotes anyway—we are supposed to extract data from footnotes—but that is harder to find, so there could be cases where the XBRL does make it easier for us to locate that data.
What about the level of data quality that you see in XBRL filings? How much cleanup is needed after you extract the XBRL data? Have you seen an improvement?
There has definitely been an improvement over time. As I said, we are extracting more data from XBRL with each filing period. Part of that is due to machine learning on our side. Part of it is also better filings. As we process more and more documents, we obviously build up a storehouse of what each particular company has filed before. That is what I mean by “machine learning.” We are able to go back, and if it is filed the same way as it was the previous time, we will be able to extract it this time. But there has also been improvement in the way companies have been filing. We do still need cleanup. That is not to say we can straight-through process the data, because right now we cannot. That would be the ultimate goal. It does always have to be mediated by an analyst.
Work is being done by industry groups we belong to, such as the XBRL US Data Quality Committee. Certainly, a lot of their initiatives are aimed at improving the quality of filers’ information, and we are seeing that it is improving over time. It is a bigger problem for the small companies. What happens is that at a small company, there may be only one person dedicated to doing the XBRL filing—there may not be a lot of resources. The bigger companies just have more resources. But it is all important for us—the small companies as much as the big ones.
Even before XBRL, how much cleanup did you have to do with paper filings?
We have different levels of quality control. There is the initial data entry; then there are automated quality checks on the data; and then we have human quality checks on that data. There is always cleanup at every stage of the process, whether we are working from XBRL or not. Financial statements are very complicated, with a lot of information, and there are multiple levels of quality control and cleanup.
When Inline XBRL becomes mandatory, will your process change at all?
We will find out. We expect it is going to be easier to extract data. With Inline XBRL, the XBRL is embedded in the statement, so you do not have to deal with two separate instances of the document. We expect it will be easier. We have started testing it. So far, very few companies have issued Inline XBRL, but we expect, if anything, that it definitely should be an improvement.
Are there specific issues with the quality of XBRL data that you see frequently?
We have seen a variety of quality in different US filings. Generally, we think that XBRL is a real leap forward. We are at the forefront of using it. But there are areas that need to improve.
One area that we focus on in our work with the XBRL US Data Quality Committee is the use of extensions. That is when companies are using ad hoc line items as opposed to items that are in the US GAAP Taxonomy. By definition, when they do that, they use a line item that is not in the taxonomy, so the machine is not able to recognize it and read it.
Extensions are a major problem. When companies use extensions, data is going to be dropped. That is one area where there is a real focus in our quality efforts. We want to see companies use fewer extensions. As I said, the XBRL US Data Quality Committee is working on that. I believe that the Financial Accounting Standards Board is focused on it. I believe that the SEC is aware of the issue. I believe we are going to be making progress on that front.
Do you have any suggestions for filers about the creation of their XBRL data?
Yes, the main suggestion is to use the taxonomy. If you create good filings, using the tags the way they are designed, you are going to get your story out to investors fast. You want to do that. You want to get your data out in front of the people who will be investing.
Create good filings. There are plenty of tags for every imaginable situation that you want to report. The taxonomy is very complete. Get familiar with it. Use it. It will work for you if you get your story out faster than anybody else.
Beyond XBRL-tagged data in financial statements of US companies, are there other types of XBRL-formatted data that you use or would like to be using?
There is a lot we would like to use. Currently, we are using only financial-statement data, but we are open to any other product. We definitely do believe in XBRL. We are in talks with different parties about XBRL—for instance, for municipal bond issuers’ financial statements, for corporate actions, and for a number of other areas. We would be in favor of any of that. It is just a matter of creating a taxonomy and having filers follow that taxonomy.
Would you like to see XBRL in earnings releases?
Oh, absolutely. The earnings release is an area where speed is key, and also where the number of data items you are reporting is pretty limited. It should be an excellent area for XBRL. That is the information which really moves the market. When you report your earnings, that is going to move the price of your stock. XBRL would be perfect for getting that out as fast as possible.
What about the end users, the financial analysts, investors, and others in the market? Are they aware that XBRL is behind your product now?
They should be aware. On the Bloomberg system, when you look at financial statements for a company, you can click through on any data. If I am looking at the income statement and I see a number for sales, I can click on that number, and it will pull up the document where that number appears. In other words, we have transparency back to the original statement where we extracted that data. In the case of XBRL, they will see the XBRL data where we extracted it from XBRL filings. It is rendered that way on the Bloomberg system. We use the SEC viewer to display XBRL data.
You started using XBRL financial data from companies outside the United States before you began to use XBRL data from US filings. What are the foreign countries from which you get corporate XBRL data, and how is their XBRL tagging different?
We have been using XBRL from companies outside the United States for longer than we have been using XBRL from companies in the United States. Currently, we take XBRL data from companies in seven foreign markets: Japan, Korea, Taiwan, China, Israel, Mexico, and Brazil. There are a few others on the horizon.
It is different overseas. It is more limited XBRL. Usually, it is just the face of the statements, without the notes. But it works better. Those XBRL taxonomies are more limited, but they also discourage the use of extensions in a way that is very rigid but that certainly works well for us, in that it is very easy for us to process the entire statement straight through to production without anybody even looking at it. In the case of Japan, it is actually a fairly complicated taxonomy, but it is so structured and they follow it so closely that it works really well for us.
Now, US XBRL is more ambitious. You are required to tag everything, whether it is in the face of the statement or in the notes. It tries to accomplish more, but at the same time that creates more complexity.
We are able to straight-through process XBRL data from most of those other countries. That is where we would like to get to with the United States, but we are not there yet. We can, for instance, straight-through process all of the financial statements from every filer in the Korean market in one minute. That is really the strength and the promise of XBRL. We are able to do that in a few other markets too.
But the footnotes in the foreign filings are not tagged at all?
No, not at all. That is the tradeoff. They give us a lot less data, and we can process it quickly, but it is not the complete statement. We have to handle the footnotes the old-fashioned way.
Why did it take longer to begin using XBRL data from the US market?
The foreign markets are more limited and structured. We first looked at XBRL in the United States when it was launched in 2009, and at that point we thought it was more complicated. We looked at the other markets probably because they were easier and we could get more bang-for-the-buck right away—they were the low-hanging fruit. Then in 2014, we revisited the US [setting] and started extracting XBRL data from US filings. But that is still a work in progress. We are not at the stage with the United States as we are with the other markets, where we process all of the XBRL data immediately.
Do you receive requests for XBRL data in financials of IFRS companies that file with the SEC?
Yes, we do. We process that data when they make it available. All seven of the foreign countries that I mentioned are IFRS jurisdictions, so we are already processing XBRL for IFRS. The big thing for IFRS is going to be in 2020, when the European Union mandates XBRL issuance for all of the companies in the member nations. That will be the big bang for IFRS reporting. We are looking forward to that and will be ready to process that.