US Inflation Is Cooling Off — Here’s What the Numbers Really Say

economics
data analysis
inflation
Published

February 16, 2026

US consumer prices rose 2.8% year-over-year in January 2026 — down from 3.0% in December, the lowest reading since July 2025. Core CPI ticked up to 2.9%. But the real story isn’t in the headline number. It’s in the momentum, the decomposition, and the structural dynamics underneath.

The Python code used to generate each chart is included in this post. Click on any code block to see the full implementation. The complete data pipeline includes:

  • 01_fetch_data.py — Fetches data from FRED (requires free API key)
  • 02_clean_data.py — Cleans and computes derived series
  • 03_visualizations.py — Generates all charts
  • 04_compute_stats.py — Computes summary statistics

All data sourced from FRED (Federal Reserve Economic Data).

Key Metrics

2.8%
Headline CPI
2.9%
Core CPI
3.5%
Supercore
2.9%
3-Month Annualized

Data as of January 2026 | Source: Bureau of Labor Statistics via FRED

The Big Picture

Headline CPI fell to 2.8% in January 2026, down from 3.0% in December 2025. That’s the lowest reading since July 2025.

Show code 
# =============================================================================
# Exhibit 1: The Headline Story
# =============================================================================

df = df_main.dropna(subset=["headline_yoy"])

fig, ax = plt.subplots(figsize=(10, 5))

# COVID recession shading
ax.axvspan(pd.Timestamp("2020-02-01"), pd.Timestamp("2020-04-01"),
           alpha=0.1, color="#888888", zorder=0)
ax.annotate('COVID\nrecession', xy=(pd.Timestamp("2020-03-01"), 0.5),
            fontsize=8, color='#888888', ha='center', va='bottom')

# Fed 2% target
ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.7)
ax.text(df.index[-1], 2.15, 'Fed 2% target', fontsize=8, color=COLORS['fed_target'],
        ha='right', va='bottom')

# CPI YoY
ax.plot(df.index, df['headline_yoy'], color=COLORS['primary'], linewidth=2)
ax.fill_between(df.index, df['headline_yoy'], alpha=0.1, color=COLORS['primary'])

# Peak annotation
peak_idx = df['headline_yoy'].idxmax()
peak_val = df['headline_yoy'].max()
ax.annotate(f'Peak: {peak_val:.1f}%', xy=(peak_idx, peak_val),
            xytext=(20, 10), textcoords='offset points',
            fontsize=9, fontweight='bold', color=COLORS['accent'],
            arrowprops=dict(arrowstyle='->', color=COLORS['accent'], lw=0.8))

# Latest value
latest_val = df['headline_yoy'].iloc[-1]
ax.annotate(f'Jan 2026: {latest_val:.1f}%', xy=(df.index[-1], latest_val),
            xytext=(-60, 20), textcoords='offset points',
            fontsize=9, fontweight='bold', color=COLORS['primary'],
            arrowprops=dict(arrowstyle='->', color=COLORS['primary'], lw=0.8))

ax.set_ylabel('Year-over-Year Change (%)')
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y'))
ax.xaxis.set_major_locator(mdates.YearLocator())
plt.tight_layout()
plt.show()
Figure 1: CPI year-over-year change since 2019, with the Fed’s 2% target for reference. The COVID-era spike peaked at 9.0% in mid-2022 and has been grinding lower since.

The trend is clear: inflation has been on a slow, grinding path lower since the post-pandemic peak. But the headline number can be misleading — it’s a trailing 12-month average that smooths over what’s actually happening right now.

Beyond the Headline

Two measures matter more than headline CPI for understanding where inflation is actually headed: core CPI (excluding volatile food and energy) and supercore (services excluding energy, which the Fed closely watches for underlying price pressure).

Show code 
# =============================================================================
# Exhibit 2: Headline vs Core vs Supercore
# =============================================================================

df = df_main.dropna(subset=['headline_yoy', 'core_yoy', 'supercore_yoy'])

fig, ax = plt.subplots(figsize=(10, 5))

ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.5)

ax.plot(df.index, df['headline_yoy'], color=COLORS['primary'], linewidth=2,
        label='Headline CPI')
ax.plot(df.index, df['core_yoy'], color=COLORS['accent'], linewidth=2,
        label='Core CPI (ex food & energy)')
ax.plot(df.index, df['supercore_yoy'], color=COLORS['secondary'], linewidth=2,
        label='Supercore (services ex energy)')

# Direct labels at end of each line
for col, color, label in [
    ('headline_yoy', COLORS['primary'], 'Headline'),
    ('core_yoy', COLORS['accent'], 'Core'),
    ('supercore_yoy', COLORS['secondary'], 'Supercore'),
]:
    last_val = df[col].iloc[-1]
    ax.text(df.index[-1], last_val, f'  {label}: {last_val:.1f}%',
            fontsize=9, color=color, va='center', fontweight='bold')

ax.set_ylabel('Year-over-Year Change (%)')
ax.legend(loc='upper right', fontsize=8)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y'))
ax.set_xlim(right=df.index[-1] + pd.Timedelta(days=180))
plt.tight_layout()
plt.show()
Figure 2: Headline, core, and supercore inflation. Supercore — services excluding energy — is the Fed’s preferred gauge of underlying price pressure.

Core CPI at 2.9% is still above the Fed’s target. This matters because core strips out the noise of gas prices and grocery bills, revealing the underlying inflation trend.

Momentum: What’s Happening Right Now

The most important chart most people never see: inflation momentum. Instead of looking at trailing 12-month averages, we can annualize shorter windows — 1-month, 3-month, 6-month — to see whether inflation is accelerating or decelerating right now.

Show code 
# =============================================================================
# Exhibit 3: Inflation Momentum
# =============================================================================

df = df_main.dropna(subset=['headline_yoy', 'headline_3m_ann']).copy()
df = df[df.index >= '2022-01-01']

fig, ax = plt.subplots(figsize=(10, 5))

ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.5)

ax.plot(df.index, df['headline_yoy'], color=COLORS['primary'], linewidth=2.5,
        label='12-month YoY', zorder=3)
ax.plot(df.index, df['headline_6m_ann'], color=COLORS['secondary'], linewidth=1.5,
        alpha=0.8, label='6-month annualized')
ax.plot(df.index, df['headline_3m_ann'], color=COLORS['accent'], linewidth=1.5,
        alpha=0.8, label='3-month annualized')
ax.plot(df.index, df['headline_1m_ann'], color=COLORS['light'], linewidth=1,
        alpha=0.6, label='1-month annualized')

for col, color, label in [
    ('headline_yoy', COLORS['primary'], '12m'),
    ('headline_6m_ann', COLORS['secondary'], '6m'),
    ('headline_3m_ann', COLORS['accent'], '3m'),
]:
    last_val = df[col].iloc[-1]
    ax.text(df.index[-1], last_val, f'  {label}: {last_val:.1f}%',
            fontsize=8, color=color, va='center', fontweight='bold')

ax.set_ylabel('Annualized Rate (%)')
ax.legend(loc='upper right', fontsize=8)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%b\n%Y'))
ax.xaxis.set_major_locator(mdates.MonthLocator(interval=3))
ax.set_xlim(right=df.index[-1] + pd.Timedelta(days=90))
plt.tight_layout()
plt.show()
Figure 3: Inflation momentum — short-run annualized rates vs the trailing 12-month average. When the 3-month rate falls below the 12-month rate, it signals genuine cooling.

The 3-month annualized rate is 2.9%, which is 0.1pp above the 12-month rate. This suggests near-term price pressures are running hotter than the trailing average.

What’s Driving the Drop

Not all prices move together. Breaking down inflation into its major components reveals where the cooling is coming from — and where stickiness remains.

Show code 
# =============================================================================
# Exhibit 4: Component Radar Chart
# =============================================================================

from matplotlib.patches import FancyBboxPatch

component_names = ['Shelter', 'Food', 'Energy', 'Medical', 'Transport.', 'Apparel']
component_cols = ['shelter_yoy', 'food_yoy', 'energy_yoy', 'medical_yoy',
                  'transportation_yoy', 'apparel_yoy']

# Pick 4 time snapshots (January of each year where available)
snapshots = {}
for year in [2023, 2024, 2025, 2026]:
    target = f'{year}-01'
    matches = df_comp[df_comp.index.strftime('%Y-%m') == target]
    if len(matches) > 0:
        snapshots[f'Jan {year}'] = matches.iloc[0]

# Radar chart setup
N = len(component_names)
angles = np.linspace(0, 2 * np.pi, N, endpoint=False).tolist()
angles += angles[:1]  # Close the polygon

fig, ax = plt.subplots(figsize=(8, 8), subplot_kw=dict(polar=True))

snapshot_styles = {
    'Jan 2023': {'color': COLORS['light'], 'ls': '--', 'lw': 1.5, 'alpha': 0.6},
    'Jan 2024': {'color': COLORS['secondary'], 'ls': '--', 'lw': 1.5, 'alpha': 0.7},
    'Jan 2025': {'color': COLORS['primary'], 'ls': '-', 'lw': 2, 'alpha': 0.8},
    'Jan 2026': {'color': COLORS['accent'], 'ls': '-', 'lw': 2.5, 'alpha': 1.0},
}

for label, row in snapshots.items():
    values = [row[col] for col in component_cols]
    values += values[:1]  # Close
    style = snapshot_styles.get(label, {'color': 'gray', 'ls': '-', 'lw': 1, 'alpha': 0.5})
    ax.plot(angles, values, label=label, **style)
    if label == 'Jan 2026':
        ax.fill(angles, values, alpha=0.08, color=style['color'])

# Highlight energy axis
energy_idx = component_names.index('Energy')
energy_angle = angles[energy_idx]
ax.plot([energy_angle, energy_angle], [0, ax.get_ylim()[1]], color=COLORS['secondary'],
        linewidth=2, alpha=0.3, zorder=0)

ax.set_xticks(angles[:-1])
ax.set_xticklabels(component_names, fontsize=10)
ax.set_title('Inflation by Component: How the Shape Has Changed', pad=20)
ax.legend(loc='upper right', bbox_to_anchor=(1.3, 1.1), fontsize=9)

plt.tight_layout()
plt.show()
Figure 4: Radar chart showing year-over-year inflation by component across four time periods. The web shrinks toward the center as inflation cools — except shelter, which remains elevated. Energy (highlighted) has collapsed from its 2022 peak.

The radar chart shows how the inflation “footprint” has shrunk over the past three years. Energy (highlighted axis) has collapsed from double-digit increases to near zero. Shelter remains the stubborn outlier — still elevated while other categories have normalized. Transportation has turned negative, reflecting falling used car prices.

The story is clear: shelter remains the single largest contributor to above-target inflation. At 3.4% YoY, it accounts for roughly 1.2pp of the headline number.

The Shelter Question

Shelter is the key to the “last mile” of disinflation — and there’s reason to believe it will keep falling.

Show code 
# =============================================================================
# Exhibit 5a: Shelter vs Headline — Showing the Lag
# =============================================================================

df_sh = pd.read_csv("data/clean/shelter_detail.csv", index_col=0, parse_dates=True)
df_sh = df_sh.dropna(subset=['shelter_yoy', 'headline_yoy'])

fig, ax = plt.subplots(figsize=(10, 5))

ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.5)

ax.plot(df_sh.index, df_sh['headline_yoy'], color=COLORS['primary'], linewidth=2,
        label='Headline CPI')
ax.plot(df_sh.index, df_sh['shelter_yoy'], color=COLORS['accent'], linewidth=2.5,
        label='Shelter CPI')

# Annotate the lag with arrows between peaks
headline_peak = df_sh['headline_yoy'].idxmax()
shelter_peak = df_sh['shelter_yoy'].idxmax()
ax.annotate('', xy=(shelter_peak, df_sh['shelter_yoy'].max() + 0.3),
            xytext=(headline_peak, df_sh['shelter_yoy'].max() + 0.3),
            arrowprops=dict(arrowstyle='<->', color=COLORS['neutral'], lw=1.5))
lag_months = (shelter_peak - headline_peak).days // 30
ax.text(headline_peak + (shelter_peak - headline_peak) / 2,
        df_sh['shelter_yoy'].max() + 0.6,
        f'~{lag_months}-month lag', ha='center', fontsize=9,
        color=COLORS['neutral'], fontweight='bold')

# Direct labels
for col, color, label in [
    ('headline_yoy', COLORS['primary'], 'Headline'),
    ('shelter_yoy', COLORS['accent'], 'Shelter'),
]:
    last_val = df_sh[col].iloc[-1]
    ax.text(df_sh.index[-1], last_val, f'  {label}: {last_val:.1f}%',
            fontsize=9, color=color, va='center', fontweight='bold')

ax.set_ylabel('Year-over-Year Change (%)')
ax.legend(loc='upper left', fontsize=8)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y'))
ax.set_xlim(right=df_sh.index[-1] + pd.Timedelta(days=120))
plt.tight_layout()
plt.show()
Figure 5: Shelter CPI vs headline CPI (year-over-year). Note how shelter peaks and troughs lag headline by roughly 12-18 months — the ‘last mile’ of disinflation is already baked in.

The chart above makes the lag crystal clear: headline CPI peaked months before shelter CPI did. The BLS shelter measure reflects the average rent across all existing leases, not just new ones — so it takes 12-18 months for market-rate changes to fully flow through.

Show code 
# =============================================================================
# Exhibit 5b: Shelter Subcategories
# =============================================================================

fig, ax = plt.subplots(figsize=(10, 5))

ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.5)

ax.plot(df_sh.index, df_sh['shelter_yoy'], color=COLORS['accent'], linewidth=2.5,
        label='Shelter (total)', alpha=0.5)
ax.plot(df_sh.index, df_sh['oer_yoy'], color=COLORS['primary'], linewidth=2,
        label="Owners' Equiv. Rent (OER)")
ax.plot(df_sh.index, df_sh['rent_yoy'], color=COLORS['secondary'], linewidth=2,
        label='Rent of Primary Residence')

# Direct labels
for col, color, label in [
    ('oer_yoy', COLORS['primary'], 'OER'),
    ('rent_yoy', COLORS['secondary'], 'Rent'),
    ('shelter_yoy', COLORS['accent'], 'Shelter'),
]:
    last_val = df_sh[col].iloc[-1]
    ax.text(df_sh.index[-1], last_val, f'  {label}: {last_val:.1f}%',
            fontsize=9, color=color, va='center', fontweight='bold')

ax.set_ylabel('Year-over-Year Change (%)')
ax.legend(loc='upper right', fontsize=8)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y'))
ax.set_xlim(right=df_sh.index[-1] + pd.Timedelta(days=120))
plt.tight_layout()
plt.show()
Figure 6: Shelter decomposed into its two main components: Owners’ Equivalent Rent (OER, ~26% of CPI) and Rent of Primary Residence (~8% of CPI). Both are declining from their 2023 peaks.

Shelter peaked at 8.2% in March 2023 and has been falling steadily since. At 3.4% now, both OER and primary rent are on a clear downward trajectory. Private-sector data on new tenant rents (from Zillow, Apartment List, and others) has been running near or below the pre-pandemic trend for months — meaning further shelter disinflation is already baked into the pipeline.

Broad-Based or Narrow?

Is the cooling concentrated in a few volatile categories, or is it broad-based? Two sets of measures help answer this question.

Sticky vs Flexible CPI

The Atlanta Fed splits the CPI basket into sticky items (things that change price infrequently, like rent, insurance, and education) and flexible items (things that change frequently, like gas, food, and airfares).

Show code 
# =============================================================================
# Exhibit 6: Sticky vs Flexible CPI
# =============================================================================

df = df_alt.dropna(subset=['sticky_cpi', 'flexible_cpi']).copy()

fig, ax = plt.subplots(figsize=(10, 5))

ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.5)

ax.plot(df.index, df['sticky_cpi'], color=COLORS['accent'], linewidth=2.5,
        label='Sticky CPI (rent, insurance, education)')
ax.plot(df.index, df['flexible_cpi'], color=COLORS['primary'], linewidth=2,
        label='Flexible CPI (gas, food, airfares)')

for col, color, label in [
    ('sticky_cpi', COLORS['accent'], 'Sticky'),
    ('flexible_cpi', COLORS['primary'], 'Flexible'),
]:
    last_val = df[col].iloc[-1]
    ax.text(df.index[-1], last_val, f'  {label}: {last_val:.1f}%',
            fontsize=9, color=color, va='center', fontweight='bold')

ax.set_ylabel('Year-over-Year Change (%)')
ax.legend(loc='upper right', fontsize=8)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y'))
ax.set_xlim(right=df.index[-1] + pd.Timedelta(days=180))
plt.tight_layout()
plt.show()
Figure 7: Atlanta Fed Sticky vs Flexible CPI. Flexible prices normalized quickly after the pandemic shock. Sticky prices — the holdout — are now declining too.

Flexible CPI at 0.7% has essentially normalized. The remaining inflation is concentrated in sticky prices at 3.0% — but these too are on a clear downtrend.

Median & Trimmed-Mean CPI

The Cleveland Fed’s median CPI and 16% trimmed-mean CPI strip out the tails of the price-change distribution. When these measures fall, it means disinflation is broad-based, not driven by a few outlier categories collapsing.

Show code 
# =============================================================================
# Exhibit 7: Median & Trimmed-Mean CPI
# =============================================================================

df = df_alt.dropna(subset=['median_cpi', 'trimmed_mean_cpi']).copy()

fig, ax = plt.subplots(figsize=(10, 5))

ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.5)

# Headline as light reference
ax.plot(df_main.index, df_main['headline_yoy'], color=COLORS['light'], linewidth=1,
        alpha=0.6, label='Headline CPI (reference)')

ax.plot(df.index, df['median_cpi'], color=COLORS['accent'], linewidth=2.5,
        label='Median CPI')
ax.plot(df.index, df['trimmed_mean_cpi'], color=COLORS['primary'], linewidth=2,
        label='16% Trimmed Mean CPI')

for col, color, label in [
    ('median_cpi', COLORS['accent'], 'Median'),
    ('trimmed_mean_cpi', COLORS['primary'], 'Trimmed Mean'),
]:
    last_val = df[col].iloc[-1]
    ax.text(df.index[-1], last_val, f'  {label}: {last_val:.1f}%',
            fontsize=9, color=color, va='center', fontweight='bold')

ax.set_ylabel('Year-over-Year Change (%)')
ax.legend(loc='upper right', fontsize=8)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y'))
ax.set_xlim(right=df.index[-1] + pd.Timedelta(days=180))
plt.tight_layout()
plt.show()
Figure 8: Cleveland Fed Median and Trimmed-Mean CPI (with headline for reference). These outlier-robust measures confirm that disinflation is broad-based.

Median CPI is at 3.0% and trimmed-mean at 2.7%. Both are declining, confirming this isn’t just energy prices doing the heavy lifting.

Historical Context

Finally, where does 2.8% sit in the long arc of American inflation?

Show code 
# =============================================================================
# Exhibit 8: The Long View
# =============================================================================

df = df_long.dropna(subset=['headline_yoy'])

fig, ax = plt.subplots(figsize=(10, 5))

ax.axhline(y=2.0, color=COLORS['fed_target'], linestyle='--', linewidth=1, alpha=0.4)

# Recession shading
long_recessions = [
    ('1969-12-01', '1970-11-01'), ('1973-11-01', '1975-03-01'),
    ('1980-01-01', '1980-07-01'), ('1981-07-01', '1982-11-01'),
    ('1990-07-01', '1991-03-01'), ('2001-03-01', '2001-11-01'),
    ('2007-12-01', '2009-06-01'), ('2020-02-01', '2020-04-01'),
]
for start, end in long_recessions:
    ax.axvspan(pd.Timestamp(start), pd.Timestamp(end), alpha=0.08, color='#888888')

ax.plot(df.index, df['headline_yoy'], color=COLORS['primary'], linewidth=1.5)
ax.fill_between(df.index, df['headline_yoy'], alpha=0.1, color=COLORS['primary'])

# Historical annotations
annotations = [
    (pd.Timestamp('1974-12-01'), 12.3, 'Oil Shock I'),
    (pd.Timestamp('1980-03-01'), 14.8, 'Volcker Era Peak'),
    (pd.Timestamp('2008-07-01'), 5.6, 'GFC'),
    (pd.Timestamp('2022-06-01'), 9.1, 'Post-COVID Peak'),
]
for date, yval, label in annotations:
    ax.annotate(label, xy=(date, yval),
                xytext=(0, 12), textcoords='offset points',
                fontsize=8, ha='center', color=COLORS['neutral'], fontweight='bold',
                bbox=dict(boxstyle='round,pad=0.2', facecolor='white',
                         edgecolor='none', alpha=0.8))

latest_val = df['headline_yoy'].iloc[-1]
ax.annotate(f'Jan 2026:\n{latest_val:.1f}%', xy=(df.index[-1], latest_val),
            xytext=(-50, 30), textcoords='offset points',
            fontsize=9, fontweight='bold', color=COLORS['primary'],
            arrowprops=dict(arrowstyle='->', color=COLORS['primary'], lw=0.8))

ax.set_ylabel('Year-over-Year Change (%)')
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y'))
ax.xaxis.set_major_locator(mdates.YearLocator(10))
plt.tight_layout()
plt.show()
Figure 9: CPI year-over-year from 1960 to present. Gray bars indicate NBER recessions. The current reading is unremarkable by historical standards — well within the ‘normal’ range of the post-Volcker era.

In historical context, 2.8% is utterly unremarkable. The long-run average since 1960 is 3.7%, and the post-COVID spike of 14.6% in March 1980 now looks like an acute episode, not a regime change. The “Great Moderation” of inflation that characterized the 1990s-2010s appears to be reasserting itself.

What It Means

For the Fed: The data gives the Federal Open Market Committee (FOMC) more room to cut rates, but the sticky-price holdouts (shelter, insurance) mean they won’t rush. Markets are pricing in further easing, but the pace will depend on whether the 3-month momentum continues to run below the 12-month average.

For consumers: The relief is real but gradual. Prices aren’t falling — they’re rising more slowly. Shelter costs are still elevated, and services inflation remains above the Fed’s target. The squeeze on household budgets is easing, not ending.

For markets: The disinflation trend supports risk assets and argues against a second inflation wave. The breadth of the cooling (confirmed by median and trimmed-mean measures) makes this more durable than if it were driven by a single volatile category.

Methodology & Data

All data sourced from FRED (Federal Reserve Economic Data):

Key data series used in this analysis
Series Description Source
CPIAUCSL CPI-U All Items BLS
CPILFESL CPI-U Less Food & Energy BLS
CUSR0000SAH1 Shelter BLS
CUUR0000SASLE Services Less Energy BLS
STICKCPIM157SFRBATL Sticky CPI Atlanta Fed
FLEXCPIM157SFRBATL Flexible CPI Atlanta Fed
MEDCPIM158SFRBCLE Median CPI Cleveland Fed
TRMMEANCPIM158SFRBCLE 16% Trimmed Mean CPI Cleveland Fed

Component weights are approximate CPI-U weights as of December 2025. Momentum rates are computed by annualizing the N-month price change: \(\left(\frac{P_t}{P_{t-N}}\right)^{12/N} - 1\).

Data current as of February 16, 2026.