Data AnalysisUpdated 918 words, 4 minutes
Fed funds versus CPI since 1955: the real rate, decade by decade
The effective federal funds rate minus CPI inflation, monthly from 1955 to July 2026, summarised by decade with the longest negative stretches, from FRED.
One chart people ask for is the fed funds rate drawn against inflation. The more useful number is their difference: the policy rate minus realised inflation, an ex-post real rate. This article computes it every month from July 1955 to July 2026 from two FRED series, summarises it by decade, and lists the longest stretches when it sat below zero. The CSV and the script that produced every figure are attached.
The data
FEDFUNDS: the effective federal funds rate, monthly average of daily figures, in percent, from the Federal Reserve's H.15 release. The FRED series starts in July 1954.CPIAUCSL: the Consumer Price Index for All Urban Consumers, all items, seasonally adjusted, index 1982-1984 = 100, from the Bureau of Labor Statistics.
Both were retrieved on 2026-09-05. The last month with both values is July 2026; August 2026 has a fed funds figure (3.63) but the CPI for August is not yet published on that date. One caveat is in the CPI data itself: there is no October 2025 observation in the series as retrieved, so the 12-month change for October 2026 will not be computable from this file when it arrives, and the October 2025 row of the real-rate column is empty.
The method
df["cpi_yoy"] = 100 * (df["CPIAUCSL"] / df["CPIAUCSL"].shift(12) - 1)
df["real"] = df["FEDFUNDS"] - df["cpi_yoy"]
Inflation is the 12-month percent change in the index, which is the convention on the FRED series page. The real rate is the fed funds rate minus that number, in percentage points. This is a backward-looking measure: it subtracts the inflation that happened over the previous year, not the inflation people expected over the next one. A proper ex-ante real rate needs a survey or a market-implied expectation and is a different article.
The shift is twelve rows, and the rows are a complete monthly calendar because FEDFUNDS has a value for every month, so a missing CPI month produces a gap rather than a misaligned comparison. That detail is easy to get wrong when a series has holes.
The table
Output of python code/fed-funds-vs-cpi-real-rates.py on 2026-09-05. The 1950s row covers July 1955 to December 1959 and the 2020s row runs to July 2026.
| decade | months | mean real rate (pp) | min (month) | max (month) | months below zero |
|---|---|---|---|---|---|
| 1950s | 54 | 0.71 | -2.73 (1958-05) | 2.82 (1959-04) | 20 |
| 1960s | 120 | 1.85 | -0.15 (1961-01) | 3.76 (1969-08) | 2 |
| 1970s | 120 | 0.01 | -4.97 (1975-02) | 4.66 (1973-07) | 56 |
| 1980s | 120 | 4.41 | -4.80 (1980-06) | 9.40 (1981-06) | 7 |
| 1990s | 120 | 2.14 | -0.23 (1993-01) | 4.11 (1998-03) | 7 |
| 2000s | 120 | 0.38 | -3.49 (2008-07) | 3.84 (2006-10) | 48 |
| 2010s | 120 | -1.16 | -3.73 (2011-09) | 0.91 (2019-01) | 102 |
| 2020s | 78 | -1.09 | -8.37 (2022-03) | 2.72 (2024-08) | 42 |
The five longest stretches with the real rate below zero:
| from | to | months |
|---|---|---|
| 2009-11 | 2014-12 | 62 |
| 2019-11 | 2023-04 | 42 |
| 1974-09 | 1977-09 | 37 |
| 2015-10 | 2018-10 | 37 |
| 2002-10 | 2005-04 | 31 |
Whole sample, July 1955 to July 2026: mean real rate 1.02 points, median 1.16, and 33.3 percent of months below zero. The latest month with both series, July 2026, has a fed funds rate of 3.63, CPI inflation of 3.30, and a real rate of 0.33.
What it shows
The 1980s stand alone. A mean real rate of 4.41 points and a peak of 9.40 in June 1981 have no counterpart in any other decade; the next-highest decade mean, the 1990s at 2.14, is less than half. The 1970s are the mirror image in a different way: the mean is essentially zero, but that average hides 56 months below zero and a trough of -4.97 in February 1975, when inflation was running near its 1974 to 1975 peak while the funds rate had already been cut.
The 2010s are the only decade with a negative mean, -1.16, and 102 of 120 months below zero. The longest single negative stretch in the record, 62 months from November 2009 to December 2014, sits inside it, and a second 37-month stretch starts ten months after the first ends. The 2020s contain both the deepest negative reading in the sample, -8.37 in March 2022, when the funds rate was still near zero and 12-month inflation was 8.5 percent, and the highest reading since 2007, 2.72 in August 2024. The real rate turned positive in May 2023, fourteen months after the trough.
The peak 12-month CPI inflation in the sample is 14.59 percent in March 1980. The 2020s peak, 8.98 percent in June 2022, is the highest since 1981 but well short of the 1980 figure.
What it does not show
The measure uses realised inflation over the past year, so it lags turning points by construction. In early 1975 the trough of -4.97 is partly a statement about 1974's inflation, not about conditions in the month it is dated. Any month-by-month comparison across regimes has that problem.
It says nothing about whether a given real rate was tight or loose in context. That judgment needs an estimate of the neutral rate, which is unobservable and disputed, and it needs expectations, which are not in these two series. The table describes the arithmetic, and it stops there.
Decade averages are also a blunt cut. The 2000s mean of 0.38 combines a long negative stretch after 2002, a positive run in 2006 and 2007, and the collapse in 2008. Anyone using these summaries should look at the monthly column in the CSV rather than the decade rows.
Finally, CPIAUCSL is revised: seasonal factors are recalculated each year, and the series page notes that FRED carries the current vintage. The figures above are correct for the file retrieved on 2026-09-05 and may differ slightly from a file pulled later.