ACMTP10: NY Fed 10-Year Treasury Term Premium Monthly Estimates (1962–2026)

This dataset provides the NY Fed’s ACM model estimate of ACMTP10 monthly since 1962, alongside the companion 2-year term premium series — the data anchor used in most academic and policy work on long-end Treasury pricing.

Dataset: US 10-Year Term Premium — ACM Model (1962–2026) · Updated 2026-06-30

Source: Federal Reserve Bank of New York — ACM Term Premium Model (Research Data)

Macro Takeaway

Dataset Overview

Dataset Variables

The CSV and Excel files contain the following columns.

Column names match the CSV headers exactly.

Download the Complete Dataset

The full ACMTP10 dataset is available in CSV and Excel formats.

Direct CSV Access — Eco3min Structured Dataset

https://eco3min.fr/dataset/us-term-premium.csv

This URL returns the complete ACMTP10 dataset in CSV format with stable column names. It can be used directly in pandas, R, curl, or any data tool.

Using the Dataset in Python

import pandas as pd

url = "https://eco3min.fr/dataset/us-term-premium.csv"
df = pd.read_csv(url, parse_dates=["date"])

print(df.head())
print(df["term_premium_10y"].describe())

Using the Dataset in R

library(readr)

url <- "https://eco3min.fr/dataset/us-term-premium.csv"
df <- read_csv(url)

head(df)
summary(df$term_premium_10y)

Both examples load the dataset directly from the URL — no download or API key required.

Methodology

ACMTP10 is the 10-year component of the term-structure model originally published in Tobias Adrian, Richard Crump and Emanuel Moench, “Pricing the Term Structure with Linear Regressions” (Journal of Financial Economics, 2013). The model is a linear-Gaussian affine specification estimated from Treasury yields across the maturity spectrum, decomposing each yield into a risk-neutral component (expected future short rates) and a term-premium residual.

The NY Fed updates the series monthly with end-of-month observations and publishes the results on its Research Data page. The model is re-estimated periodically, which can lead to small revisions in historical values when new sample periods are added — though the broad shape of the series is stable across vintages.

This Eco3min dataset is updated monthly via automated pull from the NY Fed publication, with end-of-month timestamps.

Data Quality & Provider Notes

ACMTP10 is a model output rather than a direct market observation, which shapes its quality profile: the series is internally consistent and methodologically transparent, but vintage-sensitive in a way that direct yield series like DGS10 are not.

• Release latency. The Federal Reserve Bank of New York updates the ACM term premium series monthly, typically within the first week of the following month. Eco3min mirrors the NY Fed publication with a monthly pull, so latency is bounded by the NY Fed release schedule rather than by the Eco3min pipeline.
• Revisions policy. Unlike direct yield series, ACMTP10 can be revised retroactively when the NY Fed re-estimates the model parameters with additional sample data. Revisions are usually small but can shift individual monthly readings by 10–20 basis points, particularly near sample boundaries.
• Alternative sources. The Kim-Wright (2005) and D’Amico-Kim-Wei (2014) term-premium models — both available from the Federal Reserve Board — provide alternative estimates. These series broadly co-move with ACMTP10 but can diverge by 50–100 basis points in any given month, especially during regime changes.
• Known gaps. Monthly cadence only — no daily granularity. The model assumes stationarity of the underlying factor process, which is an active subject of academic debate for the full 1962-onwards sample.

For research that relies on vintage consistency, always cite the publication date alongside the data and check the NY Fed’s revision notes when extending earlier work.

Common Pitfalls When Using ACMTP10

ACMTP10 is widely cited but several recurring interpretation errors distort the signal — most stem from treating a model output as if it were a directly observed market price.

1. Treating ACMTP10 as a single “true” measure. The term premium is a model residual; different specifications (ACM, Kim-Wright, D’Amico-Kim-Wei) can disagree by 50–100 basis points on the same date. A robust reading uses at least two models and focuses on the direction of change rather than the absolute level.
2. Confusing the term premium with the yield itself. ACMTP10 is not the 10-year yield. It is what remains of the 10-year yield once the expected-rate path is subtracted. A change in ACMTP10 with no change in the underlying 10-year yield means the model has reallocated the same yield between expectations and premium — sometimes a meaningful signal, sometimes a model artefact.
3. Reading negative ACMTP10 as anomalous. Negative term premium readings were common from 2012 to 2022, especially during episodes of large-scale asset purchases. They are not necessarily a market irregularity — they reflect strong duration demand and central bank balance-sheet pressure that compresses long-end yields below the expected policy path.
4. Confusing 10-year ACMTP10 with shorter-maturity term premium estimates. The dataset includes a 2-year term premium column. The 10-year and 2-year series have different magnitudes and frequently move in different directions — using the wrong column produces misleading conclusions about long-end versus short-end risk pricing.

Historical Regimes

1962–1979 — Rising premium with the Great Inflation buildup. ACMTP10 moved from roughly 1% in the early 1960s to peaks above 4% by the late 1970s, reflecting the steady increase in long-run inflation uncertainty as the Bretton Woods system collapsed and oil shocks accumulated.

1980–1985 — Volcker-era peak. ACMTP10 reached its modern highs (often above 4%) as investors demanded extreme compensation for duration in the face of double-digit inflation and aggressive Federal Reserve tightening. The 10-year yield itself crossed 15% during this window. Common misreadings of the term premium are addressed in what investors often get wrong about bonds and interest rates.

1985–2000 — Long disinflation, gradual compression. ACMTP10 declined steadily toward 1–2% as inflation expectations re-anchored and the Federal Reserve’s credibility consolidated. The compression of the term premium accounted for a large share of the decline in 10-year yields during this period.

2000–2008 — Greenspan conundrum. ACMTP10 trended toward zero even as the policy rate cycled between 1% and 5.25%. Academic work attributed the compression to foreign central bank Treasury accumulation; the ACM model documented the resulting term premium collapse explicitly.

2008–2015 — QE-driven negative territory. ACMTP10 turned negative for the first time in the modern sample as the Federal Reserve balance sheet expanded through three rounds of asset purchases. Negative readings became a recurring feature rather than an anomaly.

2015–2021 — Variable but predominantly negative. ACMTP10 oscillated around zero during the normalisation attempt of 2015–2018, then dropped sharply into deeply negative territory through 2020–2021 as the pandemic shock triggered emergency Federal Reserve interventions and another round of large-scale asset purchases.

2022–2026 — Renormalisation into positive territory. As the Federal Reserve shifted from QE to QT and the inflation regime broke, ACMTP10 climbed back to clearly positive readings for the first time in roughly a decade. The trajectory of ACMTP10 alongside the Federal Funds Rate documents the simultaneous unwind of the post-2008 term-premium compression.

Related Macroeconomic Datasets

ACMTP10 is most informative when read alongside the underlying Treasury yields it decomposes and the Federal Reserve balance sheet that has shaped its post-2008 dynamics.

Macroeconomic Dataset Hub

This dataset is part of the Eco3min macro-financial data repository.

Sources

• Federal Reserve Bank of New York — ACM Term Premium Model (Research Data)
• Adrian, Crump & Moench (2013), “Pricing the Term Structure with Linear Regressions”, Journal of Financial Economics

Dataset Reference