How to Fix AttributeError: 'Int64Index' object has no attribute 'month'

AttributeError: 'Int64Index' object has no attribute 'month' looks like a pandas bug but is really a type mismatch: you asked for a calendar property from something that isn't a date. The .month accessor reads a datetime64 value under the hood, and an integer index has no such value to read.

What the error means

.month, .year, .day, .hour, .dayofweek, and friends are datetime accessors. They exist on exactly two things:

• a DatetimeIndex (used as df.index.month), and
• a datetime Series, via the .dt accessor (used as df['date'].dt.month).

If df.index is an integer index — pandas calls it Int64Index before 2.0, or just Index with dtype='int64' after — it has no .month, so Python raises AttributeError. The index might display like dates (20230101) and still be plain integers.

Why it happens

1. The dates were never parsed. pd.read_csv(...) without parse_dates leaves a date column as strings (object) or integers, and setting it as the index gives you a non-datetime index.
2. A transformation reset the type. A groupby, reset_index, or arithmetic step can produce a fresh integer index where you expected the datetime one.
3. .dt is missing (or wrongly present). On a Series you need .dt.month; on an index you need plain .month. Using the wrong one triggers this error or its cousin, "Can only use .dt accessor with datetimelike values."
4. pandas 2.0 removed the class. Int64Index/Float64Index/UInt64Index are gone. Any code — yours or a dependency's — doing isinstance(idx, pd.Int64Index) now fails with AttributeError: module 'pandas' has no attribute 'Int64Index'.

Diagnose it

Print the type before you touch the accessor:

print(df.index.dtype)     # int64  → not datetimes
print(type(df.index))     # <class 'pandas.core.indexes.base.Index'> or Int64Index
print(df.index[:3])       # looks like dates but may be ints

If dtype is int64 or object, that's your answer.

The fixes

Convert the index to real datetimes, then the accessor works:

df.index = pd.to_datetime(df.index)   # now a DatetimeIndex
df["month"] = df.index.month          # works

Working from a column instead? Convert the column and use .dt:

df["month"] = pd.to_datetime(df["date"]).dt.month

Hitting the pandas 2.0 Int64Index removal? Replace isinstance checks with dtype predicates:

# Before (breaks on pandas 2.0)
if isinstance(df.index, pd.Int64Index): ...
# After
if pd.api.types.is_integer_dtype(df.index): ...

If the offending isinstance lives inside a third-party package, upgrade it — older statsmodels, scikit-learn, and xarray releases referenced the removed classes and have since been patched.

Prevent it

• Parse dates at load time: pd.read_csv(path, parse_dates=["date"], index_col="date") hands you a DatetimeIndex immediately.
• Convert once, early, right after loading, so every downstream step already has the right type.
• Assert the type in pipelines that depend on it: assert isinstance(df.index, pd.DatetimeIndex) fails loudly at the source instead of 40 lines later.
• Remember the .dt rule: index → .month; column → .dt.month.

Confirm the dtype, convert to datetimes, pick the right accessor for index vs. column, and the error disappears.

Sources

• pandas user guide