Understanding the FDIC Share of Deposit Data
My followers know I write about many aspects of retail financial services, most often focused on the business implications of some industry or economic trend. Occasionally, I like to let my inner data geek out and write for the analysts and strategists out there who work behind the scenes performing the market analyses and creating market plans. This article focuses on one of the most valuable datasets we can use in those analyses.
Every year, bankers, credit union leaders, industry analysts, and consultants get excited about the release of the FDIC’s annual Share of Deposit database. In a world filled with data, in an industry awash with facts, this is the one database that provides local data on branch performance. A singular piece of data that drives many thousands of analyses. But do you understand what’s in the database and what its limitations are? Let me explain.
Anyone who has ever undertaken to conduct market analysis for retail banks or credit unions knows about the lack of competitor ‘performance’ data. The FDIC reports branch locations updated weekly, and the NCUA reports credit union sites quarterly. These reports provide the basic ‘where’ data for all ‘branches’ (i.e., address information). While that information is useful, it doesn’t help you understand whether one site is doing better than another.
Fortunately, the FDIC also releases its annual Share of Deposits (SoD) report, and the latest version was just released a few weeks ago -- several weeks earlier than expected. This singular release provides the only piece of data that can help analysts understand one branch site’s performance relative to its competition. The SoD provides total branch deposit levels as of June 30th for the current year. The data has been available for many years, and the current FDIC website shows data back to 1994.
I’ve learned over the years, though, that you can’t just download the data and start working. The database needs to be ‘scrubbed’. I thought I’d share some of those ‘scrubbing’ techniques. The list isn’t exhaustive, and I’m expecting many other practitioners of this dark art will likely chime in with comments about additional useful techniques.
The Raw File
The original 2021 file you download from the FDIC website this year contains 81,819 records and 79 data fields. The 81,819 number represents the number of sites with branch charters, not necessarily what we all think about as retail branches.
According to the FDIC, a site must have a branch charter if “deposits are received, or checks paid or money lent." (Source: Federal Deposit Insurance Act). This language is identical to the definition of "branch" in the National Bank Act. Generally, the statutory definition of "branch" includes any bank facility at which one or more of the following activities is carried on: receiving deposits, paying checks, or lending money. These activities have been collectively referred to as the "core banking functions." More simply said, if it has on-site human tellers or funds loans, it needs a branch charter.
These rules are important and must be understood to correctly analyze the dataset. I specifically mentioned on-site “human” tellers, because ITMs don’t count. We tested this idea at Bank of America years ago and regulators confirmed at that time that ITMs didn’t qualify as tellers.
The second aspect of this definition is that certain sites being marketed as branches aren’t technically branches. Here are two examples I’m sure you’ve heard of or seen. Capital One Cafés aren’t branches, as they offer no onsite teller services nor fund loans. Bank of America’s Advanced centers aren’t branches either for the same reasons. In BofA’s case, they recently added a person to assist customers at the originally full automated sites, but unless that person is a teller, it’s still not a branch. Basically, these types of facilities are remote ATM sites.
I mentioned that there are lots of sites with branch charters that aren’t what you would call a retail branch. Branch types 11 and 12 are the most important ones as they represent “traditional” full-service branches (type 11) and supermarket branches (type 12). In total, there are 79,013 branches, as we’ve come to think of them. The one other branch type that is important to consider is 23, which represents manned drive-up facilities. While this “branch type” is generally associated with Texas, it can be found in nearly every state.
Branch service type 13 represents “digital” branches, generally where the servers are housed for e-branches. The remaining types are limited-service facilities—typically not fully open to the public or very limited in services.
The second important field to understand is the BRCENM code. While the code name seems odd, it's actually straightforward. C stands for closed branch, and is commonly paired with the next field CONSOLD, which tells you which remaining branch the deposits were reassigned to. A note of caution— these fields are not always completed.
The BRCENM “E” code indicates that deposits were estimated. Some banks estimate all their individual branch deposit levels. The “N” code represents non-depository branches. They don’t open accounts there nor do they accept deposits. For example, an established FDIC-insured bank may have a branch or office that only handles commercial lending transactions.
Analysts can use these two fields in combination to determine the true count of what we consider “open retail branches”. That final count would include Branch Service Types 11 and 12, that are not BRCENM codes C and N. That number is 78,455 branches. If you add in drive-ups that take deposits, the number increases to 79,003.
Building A Historical Connection
Analysts know that having more data is usually better. One year’s deposit number is great but having 4-5 years of historical deposit levels to create trend lines is even better. The FDIC provides one data field to help make those historical connections—the Unique Branch Number (UNINUMBR). According to the FDIC, “the unique number [is] associated with a specific physical location of a branch. This number remains with this location over time, regardless of ownership”. This field can be used to connect multiple years of deposit data to a physical site, even after an acquisition or merger. Well, at least in theory.
While the rules are pretty straightforward for closures, relocations, and mergers, they don’t appear to always be followed by every bank. Take for example branch closures. There were 1,589 listed in the latest FDIC raw file. Yet, comparing year-over-year branch counts indicates the number is closer to 4,000.
For branch relocations, the rule states you are technically closing one physical site and opening a new physical site. In those cases, the new FDIC file should show both the old and new sites. The old site should have a BRCENM code of C with the CONSOLD field showing the new branch number where the deposits went. Important to note: banks use their internal branch number field (BRNUM) to make the connection. The new branch in this case should have a new “SIMS_ESTABLISHED_DATE” too, representing the date the new branch opened. This doesn’t always happen either. If it is a short distance relocation, often the “numbers'' move to the new branch site.
Since I mentioned the established date field, I want to note the other date field “SIMS_ACQUIRED_DATE”, which indicates the date the current bank acquired the branch, and its customer base, from another bank. This field is used in mergers and acquisitions. It does not represent cases where one bank leases or buys another bank’s closed branch.
Geography Data Fields
There are numerous Census geography-related fields in the FDIC data that are helpful in analyses. There are fields for:
Address (ADDRESSBR)
City (CITYBR)
State (STALPBR)
ZIP Code (ZIPBR)
County (CNTYNAMB)
Core Based Statistical Area (CBSA_DIV_NAMB)
MSA (CSANAMB)
Latitude and Longitude (SIMS_LATITUDE, SIMS_LONGITUDE)
When using the Lat/Long fields, it’s important to check the level of geocoding provided as not all addresses are point-specific.
Recap
For all the data out there about consumers, small businesses, segments, traffic, product ownership, and spending patterns, there is only one source for national bank branch performance—the FDIC. Credit unions offer quarterly updates on all branches but don’t offer any singular branch performance data fields. If you are creative and have some technical skills, you can add value to that database by combining historical databases to get growth trends, merger impacts, consolidation impacts, relocation impacts, new branch performance, and many other analyses. Just remember to double-check and verify as the FDIC SoD database is good, but not perfect.
