Case Study 6.2 — When the pipeline is the constraint: the 2020–2021 volume shock

A real, public market event, examined from the angle Chapter 6 cannot reach on its own. The macroeconomic and policy facts here are public record. The file described in "The composite file" section is a clearly labeled composite, built from documented industry patterns, not a real borrower's record. No statistic in this study is invented; where a figure would be needed, this study tells you to look it up.

Background

In March 2020, in response to the economic disruption of the COVID-19 pandemic, the Federal Reserve cut its policy rate to near zero and resumed large-scale purchases of agency mortgage-backed securities. Mortgage rates fell, and over the following months they reached record lows in Freddie Mac's Primary Mortgage Market Survey, a series that has been published weekly since 1971.

What followed was one of the largest refinance waves in the history of American mortgage lending. Tens of millions of households held loans at rates well above the new market, and the arithmetic of refinancing became obvious enough that borrowers did not need to be sold on it. They called.

This is, in a sense, the best problem a mortgage business can have. This case study is about why it was also a very serious one, and about what it teaches that Chapter 6 — which follows one file through a process with ample capacity behind it — structurally cannot.

The issue

Chapter 6 treats turn time as a discipline variable. Order early. Un-merge the condition list. Drive to clear to close. Every one of those levers assumes that the constraint on the file is attention, and that additional attention produces additional speed.

In 2020 and 2021 that assumption broke, industry-wide and simultaneously, for a reason that is worth stating in a single sentence:

When every lender in the country is at capacity at the same time, the constraint on your file is not your effort. It is a queue you are standing in with everyone else.

The bottlenecks were the same third-party stages §6.3 identified, and they compounded:

Appraisers. The number of licensed and certified appraisers is not elastic in the short run; it takes years to credential one. When refinance volume and purchase volume rise together, the appraisal queue lengthens for everyone, and no relationship with an appraisal management company creates an appraiser who does not exist. (Agency appraisal-waiver programs absorbed part of the load on eligible files; eligibility rules for those programs change and must be verified currently.)

Title and settlement. Title examiners and closing agents faced the same volume against fixed staff, in an environment where county recorder offices themselves were operating under pandemic restrictions.

Underwriting. Lenders could and did hire, but an underwriter is not a role you staff in a fortnight; the training and the authority levels take time, and the people you hire in a boom are the people you lay off in the bust, which every operations manager knows while they are hiring.

Lock desks. Longer files require longer locks, and longer locks cost more (Chapter 30 explains why). A lender whose average file had lengthened by three weeks was suddenly buying a materially different product from the secondary desk.

The contested decision

Here is where this becomes a case study rather than a history.

Faced with more applications than they could process, lenders had to choose. Broadly, three responses were available, and the industry used all three:

1. Take everything and let the calendar stretch. Maximum volume, maximum revenue on paper, and a pipeline in which every borrower's experience degrades — including the borrowers who were already in process before the wave hit, who did nothing wrong and now wait.

2. Ration by price. Widen the spread between what the secondary market will pay and what you quote the borrower, so that fewer borrowers say yes. During this period the primary–secondary spread — the gap between the rates borrowers actually paid and the yields on the securities backing those loans — widened noticeably, and market commentators attributed a substantial part of that widening to capacity constraints rather than to credit risk. This is a documented and widely discussed pattern; look up the contemporaneous spread data rather than relying on a recalled figure.

3. Ration by rule. Restrict intake directly: pause a channel, tighten eligibility, stop taking applications from non-relationship borrowers, or lengthen the minimum lock period so that only borrowers willing to wait apply.

Option 2 is the contested one, and it is contested for a reason worth sitting with. A lender that prices worse in order to slow demand is charging borrower A more so that borrower B, already in the pipeline, gets closed on time. That is a defensible operational decision. It is also, described plainly, a decision to make a specific consumer pay more because of a constraint the consumer had no part in creating and no way to see. Both descriptions are accurate.

There was a policy layer on top of this as well. In August 2020 the Federal Housing Finance Agency announced an adverse market refinance fee of 50 basis points on most refinance loans purchased by Fannie Mae and Freddie Mac, initially effective September 1, 2020. After significant industry and congressional objection, the effective date was delayed to December 1, 2020, and the fee was subsequently eliminated for loans delivered on or after August 1, 2021. These are historical policy actions with documented dates; guarantee-fee and delivery-fee policy changes on a schedule and must be verified currently before you quote any of it to a borrower.

The composite file

The following is a composite, constructed from documented patterns of the period to make the mechanism visible. It is not a real borrower's file, and no figure in it should be treated as data.

COMPOSITE — a refinance taken in the 2020-21 wave          [constructed composite]

  A rate-and-term refinance. Borrower is a repeat client, W-2, strong file,
  no derogatory credit, comfortable equity. On any ordinary Tuesday this is a
  30-day loan and a routine one.

  day  1   application taken; disclosures out; AUS run - Approve/Eligible
  day  1   appraisal ordered. AMC accepts the order.
  day  2   title ordered
  day  6   borrower has returned every document requested. File is complete
           except for the appraisal and title.
  day 14   title commitment received. Clean.
  day 15   loan officer calls the AMC. The order has not been assigned to an
           appraiser. Nothing is wrong; there is a queue.
  day 22   AMC assigns an appraiser. Inspection scheduled for day 29.
  day 31   appraisal report received
  day 31   file submitted to underwriting
  day 38   conditional approval - 6 conditions, all minor, all borrower items
  day 41   all conditions cleared and submitted
  day 47   conditions reviewed; CLEAR TO CLOSE
  day 48   Closing Disclosure issued and received; waiting period runs
  day 52   closing and funding

  The loan officer made no mistakes. Every document was in hand on day 6.
  The file took 52 days, and 30 of them were spent waiting for an appraisal
  ordered on day 1.

Compare this to the Linden Street file's decomposition in §6.7. On Linden Street, the largest controllable failure was eleven days of a documented file sitting still — a management failure with a management fix. On this composite, the largest block is thirty days of an order placed on day one. There is no management fix. There is only an honest conversation on day 15 and a longer lock.

That second item is not a throwaway. The Linden Street file was locked for 30 days on day 12 and expired on day 42, and extending it cost \$914.38 on a file whose delay was largely self-inflicted. This composite ran twenty days longer than the shop's normal file, for reasons no one on the file controlled. A lock term is a forecast, and in a constrained market it must be a forecast of the market's calendar rather than of your own diligence. Chapter 30 prices what that costs; the process point is that the lock decision belongs to the day you learn the queue is long, not to the day it expires.

What it shows

Chapter 6's levers are real and they are not sufficient. Ordering on day 1 instead of day 5 is still worth four days. But a lever that produces four days against a thirty-day queue is a lever whose value has changed, and a loan officer who does not notice that will keep making commitments calibrated to a market that no longer exists.

Turn time is a capacity variable before it is a discipline variable. §6.7 teaches you to measure your turn time honestly. This case adds the second half: you must also know whether the number you are measuring is describing your process or describing your market. The same shop, the same people, the same discipline, produced a 30-day file in 2019 and a 52-day file in 2021.

The commitment you make is where the damage lands. No borrower in this period was harmed by a long appraisal queue in itself; they were harmed by being told a number that assumed there was no queue. This is Chapter 6's argument in a different key — the file was not damaged by the delay, it was damaged by the gap between the delay and the promise.

Pipeline capacity is a real constraint on a person, not only on a company. §6.1 said a pipeline has capacity and that the binding constraint is attention. A loan officer who took forty applications in a month during this period had, in a meaningful sense, taken more files than they could communicate about — and the failure mode of an over-capacity pipeline is not visible errors, it is silence.

The limits this teaches

Three honest limits on Chapter 6's material:

1. Compression has a floor set by other people's calendars. Some of a file is irreducible, and it varies with market conditions you do not control.

2. "Drive to clear to close" is excellent advice with ample capacity behind it and thin advice without it. In a constrained market the advice becomes set the expectation to the constraint, in writing, on day one, and update it when the constraint moves.

3. A pipeline report tells you what is stuck. It does not tell you whether you should have taken the file. That question — intake discipline — is the one this case study raises and Chapter 39 answers.

Discussion questions

1. Option 2 — rationing demand by widening price — is defended as protecting borrowers already in the pipeline and criticized as charging a consumer for a constraint they cannot see. Take each position seriously in a paragraph. Then say what, if anything, a loan officer owes a borrower they are quoting during a capacity crunch.

2. In the composite file, the loan officer learns on day 15 that the appraisal has not been assigned. Write the call they make to the borrower that afternoon. What do they say about the closing date, and what do they say about the rate lock?

3. §6.3 argues that ordering third-party work on day 7 rather than day 12 is the biggest lever in the file. Is that claim weakened or strengthened by this case study? Defend your answer with the composite's day numbers.

4. Suppose your branch's average days-to-close rises from 32 to 49 over four months. Name three questions you would ask before concluding anything about your team's performance.

5. Intake discipline: at what point does taking another application become a decision that harms the borrowers already in your pipeline? Give a concrete, checkable rule you could actually apply on a Monday — not a principle.

Sources: Federal Reserve policy actions of March 2020 and subsequent agency MBS purchases; Freddie Mac Primary Mortgage Market Survey; Federal Housing Finance Agency announcements regarding the adverse market refinance fee (announced August 2020; effective date delayed to December 1, 2020; eliminated for loans delivered on or after August 1, 2021); contemporaneous trade reporting on capacity constraints and primary–secondary spreads. The composite file is constructed for teaching and is labeled as such. Verify all current fees, guidelines, and appraisal-waiver eligibility with the source before relying on any of it.