The question
In a DCF, the discount rate arrives as an input: a number you choose, justify in a footnote, and move on from. That framing bothered me. The risk-free rate at the base of every cost of equity is not a constant of nature; it is a policy variable, sitting somewhere on a cycle, and it moves together with the cash flows it is discounting.
So: when the Fed moves, where does the effect actually land, and in what order?
The transmission channels
A rate change reaches equity prices by more than one route, and the routes work at different speeds and sometimes in opposite directions. Separating them is most of the analytical work.
| Channel | Mechanism | Speed | Hits hardest |
|---|---|---|---|
| Discount rate | Risk-free rate feeds the cost of equity, which compresses the multiple | Immediate, often on the announcement | Long-duration assets: growth, unprofitable tech |
| Cash flow | Tighter policy slows demand, so forward earnings estimates fall | Lagged: quarters, not days | Cyclicals, consumer discretionary |
| Credit | Spreads widen, refinancing gets dearer and covenants bite | Fast in spreads, slow in defaults | Leveraged balance sheets, real estate |
| Bank margin | Steeper curve widens net interest margin | Follows the curve, not the level | Banks, and it runs the other way |
| Exchange rate | A stronger dollar erodes translated overseas earnings | Fast | Large exporters, high-foreign-revenue firms |
The bank-margin channel is worth dwelling on, because it is the counterexample that keeps the analysis honest: the claim "higher rates are bad for equities" is not a claim about equities, it's a claim about long-duration equities. Sign conventions that hold for one sector reverse in another.
Why duration is the organising idea
That single relationship explains most of the cross-sectional pattern. A company whose cash flows sit far in the future has a small r − g denominator, and a small denominator makes the price arithmetically hypersensitive to the discount rate. Utilities and staples, whose cash is nearer and steadier, move less on the same shock.
Method
Define the cycles
Identify discrete tightening and easing episodes from the historical federal funds target, each one dated from the first move in a direction to the last. Working in episodes rather than in monthly changes avoids treating a pause within a cycle as a reversal of it.
Set event windows
Measure over a window around each episode rather than on announcement days alone. Announcement-day moves capture surprise, not policy: if a hike was fully priced in advance, the day itself shows almost nothing while the cycle shows a great deal.
Measure multiples, not prices
Track the change in forward P/E and EV/EBITDA by sector rather than raw returns. This is what isolates re-rating from earnings growth: a sector whose price held up because earnings rose is telling a completely different story from one whose multiple expanded.
Overlay credit
Put investment-grade and high-yield spreads on the same timeline. Credit generally reprices before equity concedes anything, which makes spreads the more informative early series.
Rank sectors by sensitivity, then check the survivors
Compare the sensitivity ordering across cycles. Any sector that ranks consistently is showing something structural; anything that ranks differently each time was reflecting that particular cycle's circumstances and should not be generalised.
Cycles examined
| Episode | Direction | Cumulative move | Terminal rate | Character of the cycle |
|---|---|---|---|---|
| 2004 – 2006 | Tightening | +425 bps | 5.25% | Seventeen consecutive 25bp moves at a pre-announced “measured pace” |
| 2007 – 2008 | Easing | −500 bps | 0 – 0.25% | Crisis response, heavily front-loaded, ending at the zero bound |
| 2015 – 2018 | Tightening | +225 bps | 2.25 – 2.50% | Gradual lift-off from zero, roughly one hike per quarter at its fastest |
| 2019 – 2020 | Easing | −225 bps | 0 – 0.25% | Three mid-cycle “insurance” cuts, then an emergency collapse to zero |
| 2022 – 2023 | Tightening | +525 bps | 5.25 – 5.50% | Fastest since the early 1980s, including four consecutive 75bp moves |
Policy rates and credit spreads are taken from the Federal Reserve Bank of St. Louis FRED database. The pace column matters as much as the size: the 2022–23 cycle delivered only a hundred basis points more than 2004–06 but did it in about a third of the time, and it is the speed, not the destination, that determines whether balance sheets get a chance to adjust.
Sector sensitivity ordering
Ranked 1 (most sensitive) to 6. The ordering is a classification rather than a measurement: it is the output of the duration logic above applied to each sector's cash flow profile, and it is the ranking that stayed stable across cycles rather than any single cycle's numbers.
| Sector | Duration profile | Sensitivity rank | Dominant channel | Direction under tightening |
|---|---|---|---|---|
| Information technology | Long | 1 | Discount rate | Negative, immediate |
| Real estate | Long, levered | 2 | Credit | Negative, builds with refinancing |
| Utilities | Short, bond-like | 3 | Discount rate | Negative; competes with bonds for yield buyers |
| Consumer discretionary | Medium | 4 | Cash flow | Negative, lagged by quarters |
| Consumer staples | Short | 5 | Cash flow | Mildly negative |
| Financials | Short | 6 | Bank margin | Positive while the curve steepens, then negative on credit losses |
Utilities and information technology look nothing alike as businesses, yet both sit near the top of the sensitivity ranking, because duration, not sector, is what rates actually price. That is also why a WACC is not an input you select and defend: it is a position on a cycle. Valuing a long-duration business at the bottom of a rate cycle and holding that rate flat through the forecast bakes in a monetary policy assumption that nobody ever writes down.
Limitations
- Correlation, not causation. Rate cycles do not occur in isolation. The Fed tightens because the economy is running hot, and a hot economy moves multiples on its own.
- Expectations do the work, not levels. Markets price the anticipated path. A measured hike that came in below expectations can loosen conditions, which makes the realised rate a poor explanatory variable on its own.
- Sector composition drifts. The technology sector of 2004 and of today are not the same collection of businesses, so a sensitivity comparison across two decades is comparing labels as much as companies.
- Small sample. A handful of cycles is not enough observations to separate a structural pattern from a run of coincidences with any statistical confidence.
- Regime dependence. Quantitative easing and balance-sheet policy mean the funds rate stopped being a complete description of monetary stance.
Next iteration
I'd use market-implied expected rates rather than realised ones, since it's the surprise that moves prices. I'd also control for the earnings cycle explicitly, because without it the analysis keeps attributing to monetary policy an effect that partly belongs to the economy that provoked it.