The Market You Need to Measure Before You Judge a Merger
For completed U.S. bank mergers, the decisive measurement problem is not how many institutions disappear from a national list. It is how many credible mortgage competitors remain in the place and product where a household seeks credit. This chapter separates the national banking sector from the local mortgage market, then explains what lender counts, market shares, and the Herfindahl–Hirschman Index can—and cannot—tell us about mortgage power.
The distinction is more than a technical preference. A bank is an organization with a balance sheet, branches, funding sources, and multiple lines of business. A mortgage lender is a possible source of a particular loan. Those categories overlap, but they are not identical. A bank can operate in many counties, originate mortgages in only some of them, and compete with nonbank firms that do not appear in a count of banks. If the question is whether a merger changes a household’s mortgage options, the relevant comparison must follow the mortgage product and the market in which alternatives are actually available.
Three layers that the word “bank” hides
The first layer is the national banking sector. This is where the familiar consolidation story is measured: the number of banks and thrifts, the assets controlled by the largest institutions, and the legal identities that merge or disappear. These measures are important because they describe scale, funding capacity, and the potential reach of large institutions. They are also the measures most likely to appear in a headline about concentration.
The second layer is the local mortgage product market. The Federal Reserve study constructs county-quarter observations of mortgage originations and counts active lenders at the parent-company level. It includes both banks and nonbanks. That choice asks a more relevant question for a mortgage applicant: which independent organizations originated this product in this local market during this period? It does not assume that a bank’s presence in a county means it was an active mortgage competitor there, or that a nonbank mortgage company should be ignored because it is not an insured depository institution.
The third layer is the borrower outcome. A lender count or a concentration index describes the competitive environment; it does not directly tell us what rate a borrower received, whether an application was approved, or whether a loan later became delinquent. Those outcomes are connected to market structure through pricing, underwriting, product choice, funding, and entry. But the connection has to be tested. A low HHI is not itself evidence that every borrower received a good rate, just as a high national asset share is not itself evidence that a local mortgage applicant faced a monopoly.
This layered view also clarifies why market definition is a substantive judgment. A relevant market has at least two dimensions: the product being compared and the geography in which customers can reasonably turn to alternatives. The FDIC’s merger policy describes product markets broadly enough to include functionally equivalent services from other kinds of competitors. It gives mortgage bankers as an example of firms that may compete directly with depository institutions for real-estate loans. For a mortgage question, leaving those firms out would make the measured market narrower than the set of alternatives a borrower may actually face.
Geography is equally consequential. The study uses counties as local market units because mortgage lending is highly localized and because counties provide a consistent way to compare lender activity over time. That does not mean a county is a perfect map of every borrower’s choice set. Some lenders originate across county lines, and a borrower may consider a product from an institution headquartered elsewhere. The county is a useful approximation for the study’s question, not a natural law of mortgage competition.
The data show why it is risky to substitute national bank counts for this local layer. In the study’s bank–county–quarter observations, the average number of active mortgage lenders is 164 and the median is 132. These observations are implicitly weighted toward counties where more banks are active, which makes them representative of the competitive environment faced by an average bank-linked origination. At the unweighted county-quarter level, the mean is 75 lenders and the median is 50. Both descriptions point to substantial competition, but they answer different questions. The first gives more influence to markets with more active banks; the second gives each county-quarter equal weight.
That weighting distinction is not a minor footnote. Imagine two counties: one hosts many lenders and produces a large share of the observations associated with active banks; the other is rural and has fewer lenders. A bank-weighted average will lean toward the first county. A county-weighted average will give both counties equal standing. Neither is automatically the “true” number of choices. The statistic must be labeled by its unit and weighting rule before a reader can know what it represents.
Parent-company aggregation solves another hidden problem. A single banking organization may operate through several charters, subsidiaries, or legal entities. Counting each one as an independent competitor could inflate the apparent number of firms. The study aggregates mortgage origination at the bank-parent level so that the lender count is closer to the number of independent organizations making strategic decisions. At the same time, it retains nonbank mortgage companies in the local market panel. The result is not a count of every brand a consumer might see; it is an effort to count independent lenders that actually originated mortgages in the observed county and quarter.
Counts, shares, and HHI answer different questions
A lender count is the simplest structural measure. It asks how many organizations were active in a county-quarter. Its strength is intuitive: a market with 100 active originators looks less dependent on one institution than a market with three. Its weakness is that it treats a lender with a tiny share and a lender with half the market as one competitor each. A count therefore measures the breadth of participation, not how evenly business is distributed or how aggressively firms compete.
Market shares add that distributional information. In the Federal Reserve study, the median lender captures only 0.39 percent of its county market. The mean is 1.53 percent because a small number of larger lenders pull the average upward, and even the 90th-percentile lender controls only 3.6 percent. The skew is important: the typical lender is a marginal player, while a few firms have larger positions. Banks hold somewhat higher shares than the full bank-and-nonbank population, consistent with advantages from branches, deposit relationships, and regulatory scale, but the paper describes the difference as modest. Nonbank mortgage companies provide additional competitive pressure.
The Herfindahl–Hirschman Index, or HHI, compresses the entire share distribution into one number. It is calculated by squaring each competitor’s market share and adding the squared shares. Squaring makes large shares count disproportionately more. If five firms each have 20 percent, the HHI is 2,000: five contributions of 20 squared. If 100 firms each have one percent, the HHI is 100. The second market has many more participants and a much more even distribution, so its HHI is lower even though both examples contain a finite number of competitors.
The HHI is useful because it distinguishes a market with many small firms from one in which a few firms dominate. But it remains a proxy. It does not measure a lender’s markup, the quality of service, approval standards, switching costs, or the ease with which a new entrant could take business. Two markets with the same HHI can have different borrower experiences if their products, customer segments, underwriting rules, or entry barriers differ. A concentration statistic can raise or lower the prior probability of market power; it cannot replace an outcome analysis.
The study reports a median HHI of approximately 400 for its bank–county–quarter observations. The 75th percentile remains below 1,000, and only a small fraction of county-quarters exceed 1,800, primarily in rural areas with limited lending activity. These figures indicate that most observed mortgage originations occur in markets that look unconcentrated under common antitrust reference points. They do not establish that every market is equally competitive, nor do they show that the merger itself caused the measured HHI.
The comparison with merger policy needs care. The reinstated FDIC Bank Merger Statement of Policy uses HHI as a primary measure when market shares are not clearly insignificant and describes a normal initial screen based on deposit-market concentration: absent a Department of Justice objection, the FDIC normally will not deny on antitrust grounds when post-merger HHI is 1,800 or less, or when a post-merger HHI above 1,800 rises by less than 200 points. The policy also considers other participants, functional equivalents, entry, outside providers, and the possibility that a merger creates a stronger competitor.
That regulatory benchmark and the study’s mortgage HHI are related but not interchangeable. The policy’s example is principally about shares of total deposits in a relevant banking market, while the study’s measure is based on mortgage originations and includes nonbank lenders. Deposits and mortgages can have different geographies, products, customer relationships, and competitors. A market can look concentrated in deposits while mortgage origination remains contested by nonbanks; the reverse can also be possible. Comparing the numbers without comparing what they measure would create a false precision.
Geographic scope supplies another check on the national-versus-local story. The median bank in the study originates mortgages in only five counties per year-quarter. The mean is 29, driven by a small number of national institutions operating across hundreds of counties, and even the 90th percentile is 35 counties—about one percent of U.S. counties. Most banks therefore remain predominantly local in mortgage activity despite decades of consolidation. The institutions may be larger in assets or belong to larger holding companies, but their mortgage decisions still meet borrowers in specific markets.
This localization changes the mechanism by which a merger could affect households. If an acquirer and target overlap in one county, the loss of one independent lender may matter there. If they serve largely different counties, the national corporate combination may have little immediate effect on a borrower’s local mortgage alternatives. If the merged institution faces dozens of other lenders, including nonbanks, it may have limited ability to raise rates or tighten approvals without losing applications. If entry is difficult and the merging firms hold large local shares, the same legal transaction deserves more concern. The market structure must be observed at the point where the borrower’s alternative is formed.
There is also a timing issue. A county-quarter snapshot describes who was active during a period, but mergers can change the market dynamically. A competitor may enter, expand, or withdraw after the transaction. A lender count can rise while individual shares fall, or a count can stay constant while one firm becomes dominant. That is why the study later examines event-time changes in lender counts and HHI rather than relying only on a pre-merger cross-section. Structure is not just a starting inventory; it is a process.
The practical rule is straightforward. Before using concentration to predict mortgage harm, identify the product, the geography, the independent lender, the time period, and the weighting convention. Then ask what the structural measure is being used to explain. If the claim concerns mortgage rates, the relevant structure should include mortgage competitors and should be connected to observed pricing. If it concerns deposits or branch access, mortgage-originations HHI may not answer it. A measure becomes informative only when its market and outcome match.
That alignment is the foundation for interpreting the merger study. Once the local market is defined, the next problem is causal: a post-merger average may move because the acquirer and target brought different portfolios into one reporting entity. The evidence must therefore compare the merged institution with what comparable local lenders were doing at the same time, rather than treating any movement after the legal completion date as proof of a new market strategy.