How Cyber Deposits Affect Perceived Competition in Banking Markets

With the proliferation of high-speed internet connections, online banking is becoming more common. In a recent survey, the Conference of State Bank Supervisors (CSBS) found that 91.5% of community banks offered mobile banking.See Conference of State Bank Supervisors. “CSBS 2019 National Survey of Community Banks,” in Community Banking in the 21st Century, 2019. In the same survey, 82.6% of community banks offered electronic bill payment, and 78.9% offered remote deposit capture.

A less common technological development is the use of online-only branches for gathering deposits nationwide. A big advantage of online-only branches is their lower overhead compared with that of a traditional brick-and-mortar branch, which can flow to the bank’s customers, employees and owners.

An important distinction between a “cyber” branch and a traditional branch is the area of service. If a bank sets up a traditional branch in a given market—such as a metropolitan area or a county—it typically serves customers only in that market. A cyber branch, in contrast, can draw depositors and other customers from the entire nation. When the bank reports those deposits at a branch level, the market with a cyber branch gets credit for deposits that clearly come from outside of the market.

Existing Cyber Branches

All Federal Deposit Insurance Corp. (FDIC)-insured institutions report their branch-level deposits as of June 30 each year, and the FDIC publishes the raw data in its annual Summary of Deposits (SOD).See FDIC Summary of Deposits. The SOD designates any full-service cyber branch with a special code (that is, with a branch service type equal to 13). The physical address for a cyber branch is usually the same as for the head office, but the bank can choose to attribute the cyber deposits somewhere else as well.

As shown in Table 1, there were 86,374 bank branches nationwide in 2019. At that time, 191 (0.22%) of all branches were designated as cyber branches, with a service type code of 13.Twenty-one of these branches were designated as cyber branches but did not yet have any deposits attributed to them. These branches were spread across 114 markets. Although small in number, these branches were larger than average, accounting for $407 billion (3.18%) of total deposits.

A total of 141 U.S. banks had one or more cyber branches, and nine of those banks had only cyber deposits. Of the 141 banks, 109 were considered community banks (with total assets less than $10 billion), constituting 2.1% of all community banks. In the CSBS survey, 2.2% of community banks claimed an online-only division; however, an additional 2.2% were actively planning to start one, and another 15.2% had discussed creating one. Thus, we can expect the amount of cyber deposits to rise in future years, given the continuing increase in use of technology in banking.

Effect of Cyber Branches on Market Concentration

The branch-level data in the SOD can be used to determine the level of concentration in any given market.For a background on the relevant laws and regulations, see Meyer. To determine market concentration and identify mergers as potentially anti-competitive, the Department of Justice (DOJ) uses a common measure of market concentration: the Herfindahl-Hirschman Index (HHI).

For banking markets, the HHI is calculated by summing the squares of banks’ shares of deposits in a given market. For example, if there are five banks in a market, and each bank has 20% market share, the resulting HHI would be 2,000 (20² + 20² + 20² + 20² + 20²). The DOJ then uses the measure to categorize the market:

• A market with an HHI of 1,000 or less is considered unconcentrated (competitive).
• A market with an HHI between 1,000 and 1,800 is moderately concentrated (moderately competitive).
• A market with an HHI above 1,800 is highly concentrated (uncompetitive to highly uncompetitive).See U.S. Department of Justice and the Federal Trade Commission, Horizontal Merger Guidelines, Aug. 19, 2010.

This also means that a perfectly competitive market would have an HHI of zero, while a pure monopoly would have an HHI of 10,000 (100²).

A merger that would increase a market’s HHI by 200 points or more and result in the market HHI exceeding 1,800 would generally require that regulators conduct an additional, customized analysis to identify potential mitigants before allowing the merger. A merger that would result in a bank having a share of deposits of 35% or more in a given market triggers a similar analysis.

Not surprisingly, rural markets tend to be much more concentrated than urban markets. The percentage of rural markets considered highly concentrated (an HHI above 1,800) has hovered in the high 80s over the last 15 years, while the analogous percentage of urban banks has hovered in the high 20s.See Meyer.

It is easy to see why the presence of cyber branches in a market can complicate such a measure of market concentration. Regardless of the size of the bank receiving cyber deposits, they clearly distort the measured level of concentration.

We can see from Table 1 that cyber branches do not represent a large proportion of deposits nationwide, either as a percentage of branches or of total deposits. However, the proportion can be quite large in particular markets, as shown in Table 2. For example, 60% of the deposits credited to the Hardy County, W.Va., market come from two cyber branches. In a market like this one, analysts should investigate how many of those deposits are actually local before computing any market concentration ratios. Unfortunately, the data needed to conduct such an analysis are not currently available in an easily accessible form, so regulators may need to make some assumptions.

In the absence of more granular data, one potential assumption is that none of the deposits of a cyber branch were actually gathered from depositors in that market. This assumption would be valid if the vast majority of the deposits credited to a cyber branch were gathered from depositors across the nation. Setting cyber deposits from some positive number to zero in a given market could have one of two effects on the calculated HHI:

• If the cyber deposits belong to one or more of the less dominant banks in the market (with a relatively low market share), then the change would make the market appear more concentrated. That is, the artificial inflation of the deposits of the smaller market-share banks no longer masks the relative dominance of the larger market-share banks.
• Conversely, if the cyber deposits belong to one or more banks with a relatively large market share, the removal of those deposits would reduce the perceived dominance of the higher-share banks and, thus, reduce the measured HHI.

In reality, the former situation is more common than the latter one. That is, banks with cyber branches in a market tend to have a relatively small role in that market (at least on paper). To illustrate this point, setting cyber deposits to zero (i.e., making the assumption that none of the deposits come from consumers or businesses residing within the market) increases the measured HHI in 60 markets and decreases the measured HHI in only 33 markets. The largest measured increase in the HHI is 524, and the largest decrease is 2,602. One must try to understand the specific circumstances of each market before applying a one-size-fits-all methodology.

Conclusion

The upshot of this analysis is that we need more granular data to fully answer the deposit concentration question. Using either extreme of 0% or 100% can distort the picture of competitiveness in individual markets.

In addition, as the cost of technology decreases, the measurement problems associated with cyber deposits are likely to get worse, not better. Even though bankers know the addresses of their cyber depositors, the systematic reporting of such data may impose nontrivial regulatory burdens on the banking system (and especially on smaller community banks), and this article does not necessarily call for such a change. Rather, it serves as a warning about a potential measurement problem that could mislead analysts in both directions.