Who turns first in the US business cycle?

Each peak

9 business-cycle peaks: who turned first each time

This section answers a different question from the clock, so the months differ:

• The clock: averaged over many cycles, where its peak falls relative to our reference index’s peak (one turn counts as 65 months).
• This section: at each of the 9 NBER business-cycle peaks since 1960, the month within 30 months of the peak when the indicator reached its own high (negative = before the peak). NBER peaks come 5 months after the reference index’s on average.
• Only indicators that follow the cycle and have data for at least 6 peaks are shown; Expected inflation, Taiwan Buffett indicator, Taiwan stocks, Euro-area stocks, Lending standards and OFR stress have fewer and appear only on the clock.

One peakMiddle halfMedianEarlier and later halves disagree

1. Median (vs NBER)Before the NBER peak
2. Leading indexEach peak: 10 mo. earlier, 9 mo. earlier, 11 mo. earlier, 2 mo. earlier, 15 mo. earlier, 12 mo. earlier, 2 mo. earlier and 12 mo. earlierMedian (vs NBER)10.5 mo. earlierBefore the NBER peak8/8Median: 10.5 mo. earlier · 8 of 8 before the NBER peak
3. US stock valuationEach peak: 10 mo. earlier, 10 mo. earlier, 10 mo. earlier, 12 mo. later, 6 mo. earlier, 10 mo. earlier, 12 mo. earlier, 3 mo. earlier and 18 mo. laterMedian (vs NBER)10 mo. earlierBefore the NBER peak7/9Median: 10 mo. earlier · 7 of 9 before the NBER peak
4. UK stockshalves differEach peak: 14 mo. earlier, 18 mo. earlier, 9 mo. earlier, 5 mo. earlier, 10 mo. earlier, 10 mo. earlier, 2 mo. earlier and 8 mo. earlierMedian (vs NBER)9.5 mo. earlierBefore the NBER peak8/8Median: 9.5 mo. earlier · 8 of 8 before the NBER peak
5. US stockshalves differEach peak: 10 mo. earlier, 9 mo. earlier, 9 mo. earlier, 13 mo. later, 5 mo. earlier, 10 mo. earlier, 11 mo. earlier, 1 mo. earlier and 18 mo. laterMedian (vs NBER)9 mo. earlierBefore the NBER peak7/9Median: 9 mo. earlier · 7 of 9 before the NBER peak
6. US Buffett indicatorhalves differEach peak: 13 mo. earlier, 10 mo. earlier, 9 mo. earlier, 12 mo. later, 6 mo. earlier, 11 mo. earlier, 12 mo. earlier, 3 mo. earlier and 16 mo. laterMedian (vs NBER)9 mo. earlierBefore the NBER peak7/9Median: 9 mo. earlier · 7 of 9 before the NBER peak
7. GDP growthEach peak: 8 mo. earlier, 8 mo. earlier, 9 mo. earlier, at the same time, 4 mo. earlier, 11 mo. earlier, 1 mo. earlier and 10 mo. earlierMedian (vs NBER)8 mo. earlierBefore the NBER peak7/8Median: 8 mo. earlier · 7 of 8 before the NBER peak
8. Nonfarm payrollsEach peak: 8 mo. earlier, 6 mo. earlier, 9 mo. earlier, 19 mo. later, 1 mo. later, 21 mo. earlier, 11 mo. earlier, at the same time and 8 mo. earlierMedian (vs NBER)8 mo. earlierBefore the NBER peak6/9Median: 8 mo. earlier · 6 of 9 before the NBER peak
9. Japan stocksEach peak: 3 mo. earlier, 9 mo. earlier, 12 mo. earlier, at the same time, 9 mo. earlier, 12 mo. earlier, 4 mo. earlier and 17 mo. laterMedian (vs NBER)6.5 mo. earlierBefore the NBER peak6/8Median: 6.5 mo. earlier · 6 of 8 before the NBER peak
10. Business confidenceEach peak: 10 mo. earlier, 28 mo. earlier, 19 mo. later, 10 mo. earlier, 1 mo. earlier and 17 mo. laterMedian (vs NBER)5.5 mo. earlierBefore the NBER peak4/6Median: 5.5 mo. earlier · 4 of 6 before the NBER peak
11. Fed funds rateEach peak: 5 mo. earlier, 1 mo. earlier, 5 mo. later, 3 mo. earlier, at the same time, 14 mo. earlier, 6 mo. earlier, 9 mo. earlier and 10 mo. earlierMedian (vs NBER)5 mo. earlierBefore the NBER peak7/9Median: 5 mo. earlier · 7 of 9 before the NBER peak
12. Philly FedEach peak: 11 mo. earlier, 15 mo. later, 3 mo. earlier, 24 mo. earlier, 14 mo. earlier, 4 mo. earlier and 16 mo. laterMedian (vs NBER)4 mo. earlierBefore the NBER peak5/7Median: 4 mo. earlier · 5 of 7 before the NBER peak
13. World stockshalves differEach peak: 16 mo. later, 9 mo. earlier, 12 mo. earlier, 14 mo. later, 4 mo. earlier, 10 mo. earlier, 11 mo. earlier, 1 mo. earlier and 18 mo. laterMedian (vs NBER)4 mo. earlierBefore the NBER peak6/9Median: 4 mo. earlier · 6 of 9 before the NBER peak
14. German stockshalves differEach peak: 4 mo. earlier, 11 mo. earlier, 15 mo. earlier, 27 mo. later, 3 mo. earlier, 9 mo. earlier, 2 mo. earlier and 17 mo. laterMedian (vs NBER)3.5 mo. earlierBefore the NBER peak6/8Median: 3.5 mo. earlier · 6 of 8 before the NBER peak
15. Net share of central banks hikingEach peak: 2 mo. earlier, 2 mo. earlier, 1 mo. earlier, 1 mo. earlier, 3 mo. later, 16 mo. earlier, 8 mo. earlier, 4 mo. later and 14 mo. earlierMedian (vs NBER)2 mo. earlierBefore the NBER peak7/9Median: 2 mo. earlier · 7 of 9 before the NBER peak
16. Wageshalves differEach peak: 12 mo. earlier, 12 mo. later, 14 mo. earlier, 2 mo. earlier, 2 mo. earlier, at the same time, 9 mo. later and 4 mo. laterMedian (vs NBER)1 mo. earlierBefore the NBER peak4/8Median: 1 mo. earlier · 4 of 8 before the NBER peak
17. US margin loanshalves differEach peak: 14 mo. earlier, 15 mo. earlier, 14 mo. earlier, 14 mo. later, 4 mo. earlier, 20 mo. later, 5 mo. later, 2 mo. later and 18 mo. laterMedian (vs NBER)2 mo. laterBefore the NBER peak4/9Median: 2 mo. later · 4 of 9 before the NBER peak
18. US inflationhalves differEach peak: 3 mo. later, 3 mo. later, 10 mo. later, 3 mo. later, 15 mo. earlier, 5 mo. later, 3 mo. earlier, 4 mo. later and 23 mo. earlierMedian (vs NBER)3 mo. laterBefore the NBER peak3/9Median: 3 mo. later · 3 of 9 before the NBER peak
19. US 10-⁠yr yieldhalves differEach peak: 7 mo. earlier, 2 mo. later, 18 mo. later, 22 mo. later, 4 mo. later, 4 mo. later, 10 mo. later, 15 mo. earlier and 16 mo. earlierMedian (vs NBER)4 mo. laterBefore the NBER peak3/9Median: 4 mo. later · 3 of 9 before the NBER peak
20. Bond volatilityhalves differEach peak: 2 mo. later, 17 mo. later, 6 mo. later, 3 mo. later, 3 mo. later, 8 mo. later, 12 mo. later and 1 mo. earlierMedian (vs NBER)4.5 mo. laterBefore the NBER peak1/8Median: 4.5 mo. later · 1 of 8 before the NBER peak
21. Core inflationhalves differEach peak: 9 mo. later, 14 mo. later, 5 mo. later, 5 mo. later, 6 mo. later, 5 mo. later, 5 mo. later and 20 mo. earlierMedian (vs NBER)5 mo. laterBefore the NBER peak1/8Median: 5 mo. later · 1 of 8 before the NBER peak
22. Credit spread (widest)Each peak: 6 mo. later, 11 mo. later, 11 mo. later, 3 mo. earlier, 12 mo. later and 3 mo. laterMedian (vs NBER)8.5 mo. laterBefore the NBER peak1/6Median: 8.5 mo. later · 1 of 6 before the NBER peak
23. Yield curve (steepest)Each peak: 11 mo. later, 14 mo. later, 18 mo. earlier, 19 mo. earlier, 11 mo. later, 25 mo. later, 11 mo. later, 22 mo. later and 23 mo. laterMedian (vs NBER)11 mo. laterBefore the NBER peak2/9Median: 11 mo. later · 2 of 9 before the NBER peak
24. Unemployment (highest)Each peak: 14 mo. later, 15 mo. later, 20 mo. later, 8 mo. later, 10 mo. earlier, 24 mo. later, 14 mo. later, 22 mo. later and 6 mo. laterMedian (vs NBER)14 mo. laterBefore the NBER peak1/9Median: 14 mo. later · 1 of 9 before the NBER peak

Back to the clock↑

Not on the clock

Which ones are not on the clock

The clock shows only the 18 indicators that follow the cycle steadily. The other two groups are listed here in full: those whose earlier and later halves (each indicator’s own record split in two) differ too much in position or sync too weakly, and those that do not follow swings of 1.5 to 8 years. The cut-offs were fixed before looking at the results.

They follow it, but the earlier and later halves disagree (15)

Each row is one indicator: in the first and second halves of its record, how many months its peak came, on average, before or after the reference peak (measured as on the clock; zero is the reference index’s peak, not the NBER peak). To count as steady, an indicator must pass both tests: halves no more than 8.15 months (45°) apart, and sync strength of at least 0.4 in both halves. Every row here fails at least one; failures are marked ×.

First half (before the split year)Second half (from the split year)Dashed = sync strength below 0.4 in that halfWithin 8.15 months (45°) of the first half×fails this test

• SplitPosition gaplimit 8.15 mo.Sync, 1st / 2ndlimit 0.4
• Taiwan cycle signalfirst half: 13 mo. earlier; second half: 19 mo. earlier (split in 2005); first-half sync strength 0.37 (below 0.4)20056 mo.0.37× / 0.401Split 2005 · gap 6 mo. · sync 0.37× / 0.401
• US Buffett indicatorfirst half: 21 mo. earlier; second half: 8 mo. earlier (split in 1989); the halves are 13 months apart (limit 8.15 months)198913 mo.×0.52 / 0.43Split 1989 · gap 13 mo.× · sync 0.52 / 0.43
• Industrial outputfirst half: 9 mo. earlier; second half: 17 mo. earlier (split in 2009); second-half sync strength 0.39 (below 0.4)20098.1 mo.0.55 / 0.39×Split 2009 · gap 8.1 mo. · sync 0.55 / 0.39×
• US margin loansfirst half: 17 mo. earlier; second half: 7 mo. earlier (split in 1989); the halves are 9 months apart (limit 8.15 months)19899 mo.×0.47 / 0.47Split 1989 · gap 9 mo.× · sync 0.47 / 0.47
• US stocksfirst half: 16 mo. earlier; second half: 5 mo. earlier (split in 1987); the halves are 11 months apart (limit 8.15 months)198711 mo.×0.51 / 0.45Split 1987 · gap 11 mo.× · sync 0.51 / 0.45
• World stocksfirst half: 19 mo. earlier; second half: 5 mo. earlier (split in 1991); the halves are 14 months apart (limit 8.15 months); first-half sync strength 0.32 (below 0.4)199114 mo.×0.32× / 0.54Split 1991 · gap 14 mo.× · sync 0.32× / 0.54
• UK stocksfirst half: 15 mo. earlier; second half: 5 mo. earlier (split in 1992); the halves are 10 months apart (limit 8.15 months); first-half sync strength 0.396 (below 0.4)199210 mo.×0.396× / 0.65Split 1992 · gap 10 mo.× · sync 0.396× / 0.65
• German stocksfirst half: 8 mo. earlier; second half: 4 mo. earlier (split in 1993); first-half sync strength 0.19 (below 0.4)19934 mo.0.19× / 0.58Split 1993 · gap 4 mo. · sync 0.19× / 0.58
• US 10-⁠yr yieldfirst half: 10 mo. later; second half: at the same time (split in 1990); the halves are 10 months apart (limit 8.15 months)199010 mo.×0.47 / 0.42Split 1990 · gap 10 mo.× · sync 0.47 / 0.42
• US inflationfirst half: 10 mo. later; second half: 2 mo. earlier (split in 1987); the halves are 12 months apart (limit 8.15 months); second-half sync strength 0.35 (below 0.4)198712 mo.×0.65 / 0.35×Split 1987 · gap 12 mo.× · sync 0.65 / 0.35×
• Wagesfirst half: 5 mo. later; second half: 8 mo. later (split in 1995); first-half sync strength 0.399 (below 0.4); second-half sync strength 0.35 (below 0.4)19953 mo.0.399× / 0.35×Split 1995 · gap 3 mo. · sync 0.399× / 0.35×
• Core inflationfirst half: 14 mo. later; second half: 11 mo. later (split in 1992); second-half sync strength 0.34 (below 0.4)19923 mo.0.73 / 0.34×Split 1992 · gap 3 mo. · sync 0.73 / 0.34×
• US dollarfirst half: 25 mo. later; second half: 17 mo. later (split in 2010); the halves are 8.19 months apart (limit 8.15 months); first-half sync strength 0.35 (below 0.4)20108.19 mo.×0.35× / 0.63Split 2010 · gap 8.19 mo.× · sync 0.35× / 0.63
• Financial stressfirst half: 23 mo. later; second half: 20 mo. later (split in 2003); first-half sync strength 0.05 (below 0.4)20034 mo.0.05× / 0.71Split 2003 · gap 4 mo. · sync 0.05× / 0.71
• Bond volatilityfirst half: 19 mo. later; second half: 23 mo. later (split in 1994); first-half sync strength 0.39 (below 0.4)19945 mo.0.39× / 0.46Split 1994 · gap 5 mo. · sync 0.39× / 0.46

The cut-offs were fixed before looking at the results; cases right at the edge are judged by the cut-off too, with no adjustment.

They do not follow swings of 1.5 to 8 years (25)

Policy rates, Fed assets, Real rate, German 10-⁠yr yield, Japan 10-⁠yr yield, USD/JPY, EUR/USD, USD/CNY, USD/TWD, World trade, Taiwan export orders, Supply pressure, Oil, Natural gas, Copper, Food prices, Commodities, Gold, Silver, China stocks, Emerging markets, India stocks, Korea stocks, Real house prices (8 economies) and US house prices

This looks only at swings 1.5 to 8 years long; longer cycles in house prices and commodities are outside this analysis.

Too short to calculate: Taiwan house prices.

How long is a cycle

Not fixed. Our reference index had 20 peaks from 1951 to 2023, 14–89 months apart (median 38); the 9 NBER peaks were 18–146 months apart (median 94.5). One turn of the clock = 65 months: the reference index’s average monthly progress around the cycle, turned into the length of one turn. It converts angles into months and is not the length of any actual cycle. The data do not support timing anything with a fixed cycle length.

Back to the clock↑

Yield-curve inversion

What happened after the yield curve inverted

Yield curve = the US 10-year Treasury yield minus the 3-month bill rate; when short rates are above long rates, the curve is inverted. Below are all inversions since 1953 (those less than 6 months apart are merged) and the NBER business-cycle peaks. On the clock, the yield curve’s position marks its steepest point (averaged over many cycles); this section measures how far each inversion (the start of a below-zero stretch) came before or after an NBER peak, which is a different thing.

• 8 inversions: 6 were followed by an NBER peak within 36 months of the start, 5–16 months later; 2 have not been so far (Sep 1966⁠–Jan 1967 and Nov 2022⁠–Nov 2024).
• Looking back from the peaks: 7 of the 9 NBER peaks had an inversion in the 36 months before (median 11 months earlier); the peaks of 1960 and 1990 did not.
• As of Oct 2026, the most recent NBER peak is Feb 2020. Peaks are dated only after the fact (the February 2020 peak, for example, was announced in June 2020), so after the latest inversion we can only say that no peak has been announced so far.
• Inversions come only every few years, so this section counts events, not month-by-month forecasts; whether the curve helps a month-by-month forecast is on the Network page.

Back to the clock↑

Method and limits

How these numbers are calculated, and what they cannot prove

• In short: strip each series of its long-run trend and short-term noise, keep only the swings lasting 1.5 to 8 years, and see where its peaks fall on average.
• Data: the raw series of all 59 indicators (not the positions shown on this site); prices are taken as logs, and quarterly data are placed in the middle month of the quarter.
• Reference index: the average of the cycles in US nonfarm payrolls, industrial production (levels) and unemployment (inverted). Its peaks catch 8 of the 9 NBER peaks since 1960, on average 5 months earlier.
• Cycle: only swings of 1.5 to 8 years are kept (Baxter–King band-pass filter, dropping 3 years at each end), then each month is converted into an angle on the cycle (Hilbert transform). A dot’s place on the clock is its average angle against the reference index; one turn is about 65 months.
• Sync: the sync strength (0 to 1) must beat 95% of 500 random shuffles of the swings for an indicator to count as following the cycle; to count as steady, the angles of the earlier and later halves must be within 45 degrees and both halves must have a sync strength of at least 0.4.
• Each peak: the indicator’s own high closest to each NBER peak, within 30 months; yield-curve inversions less than 6 months apart are merged.
• Methods and thresholds were fixed before looking at the results, and a second program recalculated everything independently with the same results. Data to Oct 2026; recalculated once a year in October.

What this cannot prove

• There have been only 9 NBER business-cycle peaks since 1960.
• The 15 listed above sit in different places in the first and second halves: a past position is no guarantee it will stay.
• Only swings of 1.5 to 8 years are covered; longer cycles in house prices and commodities are not.
• The reference index and the NBER peaks are both for the United States.
• The filter needs 3 years of data on each side, so the last 3 years cannot be placed yet; that is why the clock does not mark “where we are now”.
• The data are the latest vintages; GDP and leading indicators are revised later.
• The unemployment rate is a component of the reference index; nonfarm payrolls and industrial production appear on this site as year-on-year rates, while the index uses their levels — so part of their sync with it is built in (marked * in the list).

Full method (section 9 of “How hot and cold are measured”)→