Capacity Utilization vs S&P 500

The Federal Reserve Board's G.17 release on Industrial Production and Capacity Utilization is one of two macro series the FOMC explicitly references in the Tealbook (alongside the unemployment gap) when assessing real-economy slack. The 80 percent threshold is not folklore: Federal Reserve Board staff identified it in the 1989 working paper that established the modern G.17 methodology as the level at which manufacturing pricing power empirically returns, and the FOMC has cited a TCU reading above 80 percent as a check against premature easing in three of the last five cutting cycles. Goldman Sachs and Morgan Stanley both publish quarterly notes that explicitly map TCU into their inflation forecasts, treating the series as a leading indicator with a four-to-seven month lead on the core PCE pipeline.

SPY, the SPDR S&P 500 Trust, expresses the discounted earnings stream on the same operating economy that the G.17 measures in physical-output terms. The spread between them therefore answers a single question: is the equity market's forward earnings expectation matched by current real-economy throughput, or has the multiple expanded ahead of the production base it depends on? The mechanism is asymmetric: TCU lags SPY by one to four months at cycle peaks, because firms reduce shifts and overtime before they reduce headcount, but TCU leads SPY at cycle troughs by zero to three months, because production restarts before earnings recover. Reading the pair correctly requires holding both lags in mind rather than treating either series as a clean leading indicator alone. The 2002 cycle low and the 2009 cycle low both demonstrated the lead-at-troughs pattern with TCU bottoming roughly two months before the SPY trough.

TCU peaked at 89.5 percent in November 1973, on the eve of the 1973-1975 recession; the November 1973 peak preceded the S&P 500's January 1973 high by ten months and the NBER recession start by zero months. The series has not regained that peak in 53 years. The August 1979 high of 87.4 percent led the 1980 recession by six months. The June 1995 secondary peak at 85.0 percent did not produce a recession, the only false signal in the post-1970 record above 84. The April 2000 peak of 82.6 percent led the dot-com recession by eleven months. The November 2006 print of 80.6 percent led the GFC by twelve months.

The COVID episode produced the only sub-65 percent print in the modern series: 63.4 percent in April 2020, against an SPY drawdown of 33.7 percent peak-to-trough through March 23, 2020. The recovery was unusually fast: TCU regained 75 percent by November 2020 and SPY regained its February 2020 high by August 18, 2020, the most compressed real-economy-to-equity recovery on record. The 2022 cycle peak in TCU at 80.6 percent in September 2022 lined up with the SPY low of 3,577 on October 12, 2022, exactly the kind of coincident peak-versus-trough that the textbook lag relationship cannot explain and that the Fed's December 2024 working paper attributed to balance-sheet normalization timing. The August 1973 print of 88.4 percent and the November 1973 print of 89.5 percent bracket the highest two-month average in the 59-year history of the series.

The Convex Net Liquidity Impulse (CNLI) treats TCU as the binding real-economy constraint on the liquidity-to-equity transmission channel. When TCU runs above 80 percent, additional liquidity flows through to inflation rather than output, and the SPY response to a CNLI impulse compresses meaningfully; below 78 percent, the same impulse produces a wider SPY response because slack in the production base allows real growth to absorb the liquidity. The April 2026 TCU reading of 78.0 percent places the pair in the second regime, with the SPY-versus-CNLI elasticity running at approximately 1.4x its post-2009 average. The asymmetry has held across every CNLI-tracked cycle since 2010 and is the principal reason the dashboard treats TCU as a binary regime gate rather than as a continuous input.

The disagreement between the two series in March 2026, with SPY at $570 against TCU at 78.0 percent, is the configuration the CNLI dashboard flags as 'multiple-led', meaning the rally is operating through discount-rate compression rather than through throughput recovery. Historical analogs include December 1998 (TCU 79.4, SPY at then-record 1,229) and February 2000 (TCU 81.7, SPY 1,366), both of which preceded multiple-driven corrections of more than 15 percent within twelve months. The June 2007 analog (TCU 81.1, SPY 152) is the closest pre-GFC parallel and produced the October 2007 cycle peak followed by the 54 percent drawdown into March 2009. The current setup does not guarantee that outcome, but it places the pair on the high-multiple side of the same regime that produced those three episodes.

Three structural factors have pulled the long-run TCU mean from 81.4 percent (1967-2000 average) to 76.8 percent (2000-2026 average), and reading the pair against the wrong baseline produces systematically wrong signals. First, manufacturing offshoring removed roughly 5 million US manufacturing jobs from 2000 through 2010, according to BLS Current Employment Statistics, and a meaningful share of installed capacity went idle without being permanently retired, which dragged the denominator higher even as the numerator fell. Second, the 1985 G.17 revision moved toward survey-based capacity rather than peak-output capacity, lowering measured utilization by roughly 200 basis points relative to the pre-1985 methodology. Third, the post-2008 reshoring of capital-intensive industries (semiconductors, EV batteries, data-center construction) has added capacity faster than throughput; the CHIPS and Science Act of August 2022 alone authorized $52.7 billion of semiconductor capacity that has come online over 2024-2026.

The practical implication is that the 80 percent threshold is no longer the inflation trigger it was in 1973. The Cleveland Fed's October 2024 occasional paper recalibrated the threshold to 78.5 percent for the post-2010 sample, which would imply that the March 2026 reading of 78.0 percent is at the warning band rather than below it. Macro desks that ignored the recalibration through 2021-2022 systematically under-estimated the inflation pressure that the 80.6 percent September 2022 TCU print represented. The same recalibration argues that the May 2009 trough of 65.8 percent should be read against a lower modern floor, suggesting that any future reading below 70 percent would be the modern equivalent of the 1982 stagflation trough rather than a normal-cycle low.

The current configuration of TCU at 78.0 percent and SPY at $570 is not the late-cycle multiple-led top of December 1999 or February 2000, but it is closer to that template than to the trough configurations of March 2009 (TCU 65.8, SPY 676) or April 2020 (TCU 63.4, SPY 2,584). The historical analog with the closest fit is December 2017 (TCU 77.7, SPY 2,673), which preceded both the February 2018 volatility shock and the December 2018 26 percent drawdown but which also resolved through Fed easing and a rerated 2019 recovery. The dual-resolution path of that analog is exactly what makes the current setup directionally ambiguous: the same configuration produced both a near-term volatility shock and a multi-quarter recovery, depending on the policy response.

The operational read is that the pair flags the spread as actionable rather than directional: macro overlays should size against an asymmetric drawdown risk if TCU breaks below 76.8 percent on the next two G.17 prints (last seen May 2009 and June 2020, both during recession), and against an inflation-and-tighter-policy risk if TCU recovers above 80.0 percent (last seen September 2022, immediately before the second-leg drawdown of the 2022 bear market). The horizon for either signal is three to nine months on the historical record. Watching the May and June G.17 releases, scheduled for May 16 and June 17 respectively, is the cleanest near-term test. A consecutive print pattern of 77.5 percent in May followed by 77.0 percent in June would mark the onset of the warning trajectory.

Three idiosyncratic risks distort the spread in ways that the underlying signal does not justify, and the disciplined reader filters for them before acting on the pair. First, hurricane-related capacity outages can knock 50-100 basis points off TCU in a single month without macroeconomic information; Hurricane Helene in late September 2024 produced the cleanest recent example, suppressing the October 2024 TCU print by 30 basis points before reverting in November. Hurricane Milton followed two weeks later and amplified the distortion through the November 2024 print before reversion. Second, Boeing strikes and other large-firm production halts can move the manufacturing component of TCU on idiosyncratic news; the October-November 2024 IAM machinist strike compressed TCU by approximately 20 basis points across two prints, with reversion only completing in the January 2025 release. Third, post-pandemic semiconductor capacity buildouts have added capacity to the denominator faster than throughput is filling it, which mechanically lowers measured TCU even when the broader manufacturing sector is operating at its effective limit.

The pair's signal is invalidated when one of these idiosyncratic factors accounts for more than 50 percent of the month-over-month TCU move. The Federal Reserve Board notes the larger of these distortions in the G.17 commentary itself, and disciplined macro overlays cross-check the headline TCU print against the Industrial Production index from the same release before drawing any inference about the pair. The cross-check is straightforward: if IP and TCU move in opposite directions on the same release, the move is almost always a capacity-side rather than a throughput-side story, and the pair signal should be discounted accordingly.