The +0.5 Rule: What the NFCI’s Conventional Stress Threshold Actually Caught (1990–2026)

Executive summary

• 0 sustained leads, 1 coincident hit, 1 brief retraced signal, 2 outright misses. Strict out-of-sample audit on the four US recessions since 1990 (post-VIX, post-real-time-NFCI window): GFC = coincident hit (Dec 7, 2007); 1990 S&L/Gulf = brief signal in spring 1989 that retraced sixteen months before the recession; 2001 and 2020 = NFCI never crossed +0.5.
• The “+0.5 rule” is a backcast artifact. NFCI’s first FRED vintage is May 25, 2011. The pre-2011 record — including the 1973, 1980, 1981 recessions where the rule appears to work — is reconstructed using a dynamic factor model calibrated on the full 1971–2010 sample.
• VIX panics and NFCI stress measure different things. Of the 62 weeks since 1990 when weekly-max VIX reached 40 or higher, 34 (55%) coincided with NFCI below +0.5 — equity panics with no systemic-stress confirmation (1998 LTCM, 2001 9/11, 2002 corporate scandals, 2010 Flash Crash, 2011 eurozone, 2015 China, April 2025 tariffs).
• COVID is the largest divergence on record. Weekly-max VIX peaked at 82.7 in March 2020 — the highest in the series. NFCI peaked at 0.31, never reaching the +0.5 threshold.
• The Fed put has compressed NFCI. Every NFCI reading above +1.0 since 1990 belongs to the 2008–2009 GFC. No other event in the post-1990 window has produced a comparable systemic-stress signal — including 2020.

What the conventional reading claims

The Chicago Fed’s National Financial Conditions Index (NFCI) is a weighted average of 105 indicators across money markets, debt and equity markets, and the traditional and “shadow” banking systems. Constructed by Brave and Butters (Chicago Fed, 2011) and updated weekly, it is normalized to mean zero and unit standard deviation over the 1971+ sample. Positive values indicate financial conditions tighter than the historical average; negative values, looser than average.

In the years following its publication, NFCI was adopted by practitioners, financial journalists, and a generation of macro newsletters as a leading recession indicator. The conventional reading, repeated across blog posts and research notes, runs roughly as follows: sustained NFCI readings above +0.5 have preceded every US recession since 1971; the median lead time is around five months; the indicator has zero false positives.

Backtests over the full 1971–2010 sample appear to support this. The series crossed +0.5 in 1971, 1978, 1979 — ahead of the 1973–75, 1980, and 1981–82 recessions. The 2007–09 financial crisis produced the largest NFCI spike in the series, peaking at 3.06 in November 2008.

What 35 years of out-of-sample data show

Four US recessions have occurred since 1990: the 1990–91 S&L / Gulf War recession, the 2001 dot-com recession, the 2007–09 Great Financial Crisis, and the 2020 COVID-19 recession. The table below reports NFCI behavior in the 24 months preceding each recession’s NBER-dated start.

Two recessions (2001, 2020) occurred without NFCI ever crossing +0.5 in the two years preceding them. The 2008 crossing arrived three weeks before the official recession start, qualifying as coincident rather than leading. The 1990 case is intermediate: NFCI did cross +0.5 in March 1989, but only for nine weeks, after which it retraced to negative territory and did not return to stress levels before the recession arrived sixteen months later. By the strictest reading — sustained NFCI above +0.5 in the months preceding a recession — the post-1990 record is zero hits in four attempts.

The same picture emerges when looking at NFCI’s distribution since 1990: the index has spent 81 weeks above +0.5 (4.3% of all weeks) and 31 weeks above +1.5 (1.6%). All of the +1.5 readings, and 78 of the 81 readings above +0.5, occurred during the December 2007 – July 2009 GFC window. Outside of that single episode, NFCI has crossed +0.5 in only two brief stretches: three weeks at the 1990–91 turn (already during the 1990 recession), and the four weeks immediately preceding the GFC.

Two legitimate defenses of the index

The empirical record is what it is, but two defenses of NFCI’s design deserve to be presented in full.

First, NFCI was never intended as a strict recession predictor. Brave and Butters (2011) describe the index as a “comprehensive weekly update” on US financial conditions — a state descriptor aggregating ~105 contemporaneous indicators. The interpretation of NFCI as a leading indicator with a fixed +0.5 threshold is a practitioner convention that grew up around the series, not a property the Chicago Fed has ever claimed. Reading NFCI as describing the current state of financial conditions, the post-1990 record is consistent with what the index is designed to do: the GFC produced extreme readings, the 2020 stress was real but brief and aggressively counteracted by Fed action, and the periods between have shown looser-than-average conditions.

Second, the post-GFC compression reflects genuine structural change. The financial system of 2010–2026 is materially different from the 1970–2010 system the index was calibrated on. Post-Dodd-Frank capital and liquidity requirements, central bank standing facilities, the Fed’s enlarged balance sheet, and an order-of-magnitude expansion of pre-emptive Fed action have all reduced the type and persistence of stress events that drove NFCI’s historical signal. The 2020 COVID episode is the cleanest example: NFCI was rising fast and reached 0.31 by the first week of April 2020 — but the Fed’s announcement of unlimited Treasury and MBS purchases on March 23, the standup of the PMCCF, SMCCF, MLF, and TALF facilities, and the dollar swap line expansion compressed funding stress within weeks. The index is not failing to detect stress; it is correctly describing a system in which stress is dampened faster than it would have been pre-2008.

The implication is not that NFCI is useless — it remains the single best aggregate descriptor of contemporaneous US financial conditions available at weekly frequency. The implication is that the +0.5 rule, as widely cited, describes a regime that no longer exists.

The VIX comparison: equity panic is not systemic stress

The VIX, available daily since January 1990, measures the 30-day implied volatility of S&P 500 options. By construction it tracks the equity market’s pricing of near-term downside risk, not the underlying condition of the financial system. The 1990–2026 record shows that VIX panics (weekly-max readings of 40 or higher) and NFCI stress (sustained readings above +0.5) overlap less than commonly assumed.

Of the 62 weeks since 1990 when weekly-max VIX reached or exceeded 40, only 28 (45%) coincided with NFCI above +0.5. The same divergence, extended to high-yield credit spreads, is set out in our side-by-side of NFCI, the VIX and HY OAS as stress gauges. The other 34 weeks — the majority of VIX panic weeks — occurred while NFCI remained in territory the Chicago Fed would describe as looser than the historical average.

The pattern is consistent: the only window in which both VIX and NFCI sustained extreme readings simultaneously was the October 2008 – April 2009 stretch of the GFC. The broader catalogue of comparable equity-volatility spikes is mapped in our history of major VIX spikes from 1990 to 2026. Every other panic in the modern record was an equity-market event that did not develop into a systemic-funding-stress event. In several cases — 2002 corporate scandals, 2015 China devaluation, April 2025 tariffs — NFCI was outright negative while VIX printed above 40.

In the 2008 GFC, the VIX peak preceded the NFCI peak by about one week: VIX hit its first blow-off on October 24, 2008 (79.1), then its all-time high for that episode on November 21 (80.9); NFCI continued rising into the November 28 print of 3.06, then began to roll over. The two indices were rising together but on different clocks — equity volatility prices forward expectations while NFCI aggregates contemporaneous funding metrics that lag by days to weeks.

What NFCI does capture, when used carefully

The audit is hostile to one specific reading of NFCI — the “+0.5 rule as recession predictor” — not to the index itself. Used as a coincident state descriptor of US funding conditions, NFCI continues to do useful work.

The index moved sharply higher in March 2020, peaking at 0.31 in early April before the Fed’s intervention compressed it. That movement, even though it did not reach the +0.5 threshold, was the largest single-month NFCI swing outside the GFC. A reader watching NFCI in real time during March 2020 saw funding stress building in real time — the threshold was the wrong reading, not the index.

The same applies to the August 2007 quant-fund crisis. NFCI rose from -0.53 in late June 2007 to +0.22 in early September — an eleven-week, 0.75-point swing that was the earliest aggregate signal of the subprime credit dislocation. The level did not reach +0.5 until December 2007, but the direction and pace were informative three months earlier.

Treating NFCI as a state variable, the most useful operational questions are about direction and acceleration, not threshold-crossing. The August 2007 quant-fund swing described above — and, in a different direction, the March 2020 rise that fell short of +0.5 but compressed only after the Fed’s intervention — illustrates the point: meaningful information about US funding stress was visible in NFCI’s slope and acceleration before either episode reached the conventional +0.5 cutoff. The threshold itself, in both cases, was the wrong reading.

Key levels to watch

• Invalidation of the “no systemic stress” reading: a 13-week rise in NFCI of 0.5 points or more. Latest reading (week of 2026-03-13) is -0.486; a move to +0.014 over 13 weeks would meet this threshold. The index would shift from “looser than average” to coincident-stress territory.
• Confirmation of the current loose regime: NFCI continuing to print in the -0.40 to -0.65 range. The post-COVID record (Sep 2020 onward) shows roughly half of weekly observations in this band, with the median at -0.44, and brief excursions above -0.40 typically reverting within one to two months. A sustained move above -0.40 lasting more than four months would mark a regime change away from the current loose configuration.
• Calendar catalyst: Chicago Fed publishes NFCI every Wednesday at 8:30 a.m. CT, covering the previous Friday’s close. The 105 underlying indicators include components updated monthly; revisions to historic readings tend to be larger near the start of each month.

Past distributions are not predictive of future outcomes. The patterns described here are descriptive of the 1990–2026 historical record.

Dataset and methodology

The CSV accompanying this study (download link in PHP-managed shortcode) contains 2,880 weekly Friday-ending observations from January 8, 1971 to March 13, 2026, with the following columns:

• date — Friday-ending week
• nfci — Chicago Fed National Financial Conditions Index (FRED series NFCI). Pre-2011 readings are backcasted by construction.
• vix_week_friday — VIX close on the Friday (FRED series VIXCLS, daily). NaN before January 1990 (VIX inception).
• vix_week_avg, vix_week_max — within-week mean and maximum of daily VIX closes.
• recession, recession_label — NBER US recession flags. Labels: Stagflation I (1973–75), Volcker I (1980), Volcker II (1981–82), S&L/Gulf (1990–91), Dot-com (2001), GFC (2007–09), COVID (2020).
• nfci_regime — five-state classification: very_loose (<-0.5), loose (-0.5 to 0), tight (0 to 0.5), stress (0.5 to 1.5), crisis (≥1.5).
• vix_regime — four-state classification based on weekly max: calm (<15), normal (15–25), elevated (25–40), panic (≥40).
• divergence — pairwise state: aligned, vix_panic_alone, nfci_stress_alone, both_crisis, vix_elevated_nfci_loose, nfci_tight_vix_calm.

Sources: Chicago Fed (NFCI), CBOE (VIXCLS), NBER (recession dating). License: CC BY 4.0. All statistics in this page were computed from this CSV; a numerical audit script verifying every annotated value is available on request.

Frequently asked questions

Citation

If you cite this study, please attribute as: Eco3min Research, “The +0.5 Rule: NFCI Stress Threshold Audit, 1990–2026,” eco3min.fr/en/nfci-vs-vix-stress-audit/, 2026. CC BY 4.0.