Long-history recession indicators

When short-term yields exceed long-term yields the curve is inverted. An inversion has preceded every US recession since 1960 with one false positive (1966), typically with 6–18 months lead. Zero is the threshold. Pre-1982 portion is monthly GS10 − TB3MS (10y constant maturity minus 3m T-bill secondary market rate); post-1982 is daily T10Y3M. Conceptually the same indicator at slightly different sampling — the 1973 and 1980 inversions both visible.

Policy stance, not a recession indicator per se. Every recession in the last 50 years was preceded by a hiking cycle, then deep cuts during/after. Reads naturally as a pair with the yield curve.

The 3-month average unemployment rate minus its trailing 12-month low. When this exceeds 0.5pp, recession has begun in every postwar US case. Triggered in 2024-08 and reset shortly after — one of the few false signals.

4-week moving average of seasonally-adjusted initial claims, in thousands. Sustained rise from cycle lows is the signal — slight lead-to-coincident. Log scale to keep the COVID 2020 spike from compressing the rest of history.

Coincident, not leading — unemployment rises after a recession has already started. Useful as confirmation for the Sahm and claims signals, and to see how deep each historical recession went.

The classic long-history credit-stress proxy, monthly back to 1953. Wider spreads signal funding stress and risk repricing. Substituted here for the BofA HY OAS series because FRED's free CSV restricts that one to the last three years. Spreads above 3pp coincide with every postwar recession.

The single best real-economy leading indicator in the modern record. Housing peaked mid-2006 — over a year before the 2007–09 recession was officially dated. Rate-sensitive sector turns first.

Coincident-to-lagging — Australian unemployment rises as the recession is underway and keeps rising for months after the trough. Series is the standard ABS Labour Force seasonally-adjusted monthly rate from February 1978 onwards. Grey bands are Australian recessions (two consecutive negative GDP quarters): 1974–75 (oil-shock stagflation), 1981–83 (early-80s), 1990–91 ('the recession we had to have'), and 2020 (COVID). Each of the three pre-COVID recessions took unemployment from ~6% to 10%+; 1992 peaked above 11%. The COVID spike was sharper but recovered far faster thanks to JobKeeper. The current 4.3% is sitting near the 50-year low.

Coincident-to-mildly-leading: equities typically begin falling 3–6 months before the recession is officially dated, and start recovering before the trough. Log scale because the absolute level grew ~70x from 1950.

Nasdaq Composite divided by the S&P 500 at the same point in time. The raw-level ratio has visited 3.5x exactly three times in 55 years: the dot-com peak (3.62x, March 10 2000), the everything-bubble peak (3.58x, February 12 2021), and today (3.47x, April 2026 — third-highest reading on record). Each of the first two peaks was followed by a tech-led drawdown of 50%+. Caveat: tech's share of S&P 500 earnings has roughly doubled since 2000, so a structural baseline shift from ~1.5x to ~2.5x is plausible — see the indexed-from-1971 chart below to interrogate the secular-uptrend question.

The same two indices, both rebased to 100 on 5 February 1971 (the day the NASDAQ Composite went live), plotted on a log scale so equal vertical distance equals equal percentage growth. This answers a question the ratio chart can't: is the post-2000 rise of tech a secular re-rating or a recurring bubble pattern? The two lines move together until ~1995, NASDAQ pulls dramatically away into 2000, fully reverts to S&P by 2002, then steadily outpaces the broad market from ~2010 onwards. The 2020–21 acceleration is visibly above the post-2010 trend channel — your stimulus-distortion read has direct visual support here. Whether the channel slopes are 'permanent' or just the next bubble in slow motion is the real question; either reading fits the data.

What the options market is pricing for the next 30 days of S&P movement. Below 15 = complacency; 15–25 = normal; above 30 = stress; above 40 = panic. Spikes are short-lived but hugely informative about market state.

Flight-to-safety asset; rises in stagflation and during currency-debasement fears. Negatively correlated with real interest rates. Log scale because the absolute level grew ~140x over 80 years. From 1934 to August 1971, the USD was legally pegged to gold at $35/oz under the Bretton Woods system — the London market traded slightly below par at ~$33.85. On 15 August 1971 Nixon suspended gold convertibility ('Nixon shock'), severing the dollar from gold and leading to the modern floating-rate regime. The chart splices monthly LBMA fix data (1940-2000) with daily front-month futures (2000+).

FRED PCU3253113253111 — the only long-history free series for anhydrous ammonia specifically. Index, not USD price (Dec 1984 = 100). BLS discontinued it in Dec 2014 with the move to the new NAICS PPI structure, so post-2014 use the urea chart above for the continuing nitrogen story. What's worth seeing: the 1973-74 oil shock takes ammonia from ~62 to ~205 in one year — by far the largest single-year move in 50 years of data. The 2008 commodity-boom spike (~330 → 405) is the only comparable move. Log scale to keep both spikes legible.

Urea spot price (Black Sea / Middle East FOB), monthly USD/mt from January 1960. Source: World Bank CMO 'Pink Sheet'. Direct anhydrous ammonia daily price data 1960-present is not free (Argus, ICIS, S&P Platts paywall it; FRED's Anhydrous Ammonia PPI ran 1965-2014 monthly but was discontinued). Urea is ammonia + CO2 and tracks ammonia prices closely with a small predictable spread. Two visible spikes: the 1973-74 energy shock (gas is the main feedstock for ammonia synthesis) and the 2021-22 European gas crisis (Russia, the world's largest urea exporter, saw massive disruption). Log scale because the range is $20 → $900+.

ICE Coffee 'C' front-month, the global Arabica benchmark. Quoted in US cents per pound. Not a real recession indicator — coffee is dominated by Brazilian weather (frosts, droughts), Vietnamese Robusta supply, and the BRL/USD exchange rate. The 2011 and 2022 spikes loosely line up with post-recession demand recovery (~24 months after the 2008 trough and the 2020 trough respectively), but the 2014 and 2025 peaks are pure supply shocks with no recession nearby. Included because it directly impacts daily life. Log scale because the range is 40c → 440c.

Behaves as a high-beta risk asset (correlates with Nasdaq) most of the time, with occasional decoupling episodes. Limited history (2014+) — only fully covers the 2020 COVID drawdown so cross-cycle interpretation is thin.

Rising DXY in a global stress event signals tightening dollar liquidity, which transmits stress to emerging markets and dollar-debtor economies. The hidden recession-transmission lever.