How to Change Data Type of Column in Pandas

To change the datatype (data type) of a column in pandas, use astype(): df["price"] = df["price"].astype(float). Pass a dictionary to change several column types at once, use pd.to_numeric() or pd.to_datetime() when the data is messy, and astype("category") to save memory. This guide shows each way to change a pandas column type, the errors you will meet (and how to fix them), and how missing values affect integer columns.

All examples were run with Python 3.12.5 and pandas 3.0.6 in the Windows Command Prompt. Reference: DataFrame.astype and pandas.to_numeric in the pandas documentation.

Change a column type with astype()

Check the current types with df.dtypes, then assign the converted column back:

import pandas as pd

df = pd.DataFrame({
    "order_id": ["1001", "1002", "1003"],
    "price": ["19.99", "5.50", "12.00"],
    "qty": [2.0, 1.0, 3.0],
    "order_date": ["2025-03-14", "2025-03-15", "2025-03-16"],
})
print(df.dtypes, "\n")

df["order_id"] = df["order_id"].astype(int)      # text -> integer
df["price"] = df["price"].astype(float)          # text -> float
df["qty"] = df["qty"].astype("int64")            # float -> integer
print(df.dtypes)
print(round(df["price"].sum(), 2))               # numbers now add up

Output:

order_id          str
price             str
qty           float64
order_date        str
dtype: object 

order_id        int64
price         float64
qty             int64
order_date        str
dtype: object
37.49

In pandas 3, text columns show the dtype str (earlier versions showed object). astype() returns a new Series; nothing changes until you assign it.

Change multiple column types at once

import pandas as pd

df = pd.DataFrame({
    "order_id": ["1001", "1002", "1003"],
    "price": ["19.99", "5.50", "12.00"],
    "qty": [2.0, 1.0, 3.0],
    "order_date": ["2025-03-14", "2025-03-15", "2025-03-16"],
})

df = df.astype({"order_id": "int64", "price": "float64", "qty": "int32"})   # several columns at once
print(df.dtypes, "\n")

df2 = df.astype(str)                                                         # every column
print(df2.dtypes)

Output:

order_id        int64
price         float64
qty             int32
order_date        str
dtype: object 

order_id      str
price         str
qty           str
order_date    str
dtype: object

Convert messy text to numbers with pd.to_numeric()

astype(float) fails on the first value it cannot parse. pd.to_numeric(..., errors="coerce") turns bad values into NaN instead, and a little cleaning first rescues values like "1,200":

import pandas as pd

df = pd.DataFrame({"amount": ["120", "85.5", "N/A", "1,200", " 42 "]})

try:
    df["amount"].astype(float)
except ValueError as e:
    print("astype(float) -> ValueError:", e)

df["coerced"] = pd.to_numeric(df["amount"], errors="coerce")                   # bad values -> NaN
df["cleaned"] = pd.to_numeric(df["amount"].str.replace(",", "").str.strip(), errors="coerce")
print(df)
print(df.dtypes)

Output:

astype(float) -> ValueError: could not convert string to float: 'N/A'
  amount  coerced  cleaned
0    120    120.0    120.0
1   85.5     85.5     85.5
2    N/A      NaN      NaN
3  1,200      NaN   1200.0
4    42      42.0     42.0
amount         str
coerced    float64
cleaned    float64
dtype: object

Integers with missing values: use Int64

A normal integer column cannot hold NaN. Use pandas’ nullable "Int64" type (capital I), or fill the gaps first:

import pandas as pd

s = pd.Series([3.0, None, 7.0])

try:
    s.astype(int)
except Exception as e:
    print(type(e).__name__ + ":", e)

print(s.astype("Int64"))                 # nullable integer keeps the missing value as <NA>
print(s.fillna(0).astype(int).tolist())  # or fill first

Output:

IntCastingNaNError: Cannot convert non-finite values (NA or inf) to integer.Replace or remove non-finite values or cast to an integer typethat supports these values (e.g. 'Int64')
0       3
1    <NA>
2       7
dtype: Int64
[3, 0, 7]

Convert a column to datetime

Use pd.to_datetime() for dates stored as text or numbers. Give the format when you know it, and errors="coerce" for rows that do not match:

import pandas as pd

df = pd.DataFrame({"shipped": ["2025-03-14", "14/03/2025", "not shipped"],
                   "epoch_s": [1741910400, 1741996800, 1742083200]})

df["shipped_dt"] = pd.to_datetime(df["shipped"], format="%Y-%m-%d", errors="coerce")
df["from_epoch"] = pd.to_datetime(df["epoch_s"], unit="s")
print(df)
print(df.dtypes)
print(df["from_epoch"].dt.day_name().tolist())

Output:

       shipped     epoch_s shipped_dt from_epoch
0   2025-03-14  1741910400 2025-03-14 2025-03-14
1   14/03/2025  1741996800        NaT 2025-03-15
2  not shipped  1742083200        NaT 2025-03-16
shipped                  str
epoch_s                int64
shipped_dt    datetime64[us]
from_epoch     datetime64[s]
dtype: object
['Friday', 'Saturday', 'Sunday']

For formatting dates back to text, see convert a date and time to a string.

Save memory with the category type

A column with a few repeated values (states, departments, status codes) takes far less memory as category:

import pandas as pd
import numpy as np

rng = np.random.default_rng(0)
df = pd.DataFrame({"state": rng.choice(["TX", "CA", "NY", "FL"], size=100_000)})

before = df["state"].memory_usage(deep=True)
df["state"] = df["state"].astype("category")
after = df["state"].memory_usage(deep=True)

print(df["state"].dtype, list(df["state"].cat.categories))
print(f"memory: {before:,} -> {after:,} bytes ({before / after:.0f}x smaller)")

Output:

category ['CA', 'FL', 'NY', 'TX']
memory: 5,100,132 -> 100,336 bytes (51x smaller)

Convert a column to boolean safely

astype(bool) treats every non-empty string as True, including "False". Map the values explicitly:

import pandas as pd

s = pd.Series(["True", "False", "false", "yes", ""])

print(s.astype(bool).tolist())                            # any non-empty string is True!
mapping = {"true": True, "false": False, "yes": True, "no": False}
print(s.str.lower().map(mapping).tolist())                # explicit mapping

Output:

[True, True, True, True, False]
[True, False, False, True, nan]

Let pandas choose: convert_dtypes()

convert_dtypes() picks the best nullable type for every column, which is handy right after loading a file:

import pandas as pd

df = pd.DataFrame({"id": [1, 2, None], "name": ["Ana", "Ben", None], "score": [9.5, None, 7.0], "active": [True, None, False]})
print(df.dtypes, "\n")
print(df.convert_dtypes().dtypes)          # best nullable dtype for every column

Output:

id        float64
name          str
score     float64
active     object
dtype: object 

id          Int64
name       string
score     Float64
active    boolean
dtype: object

Goal	Code
Text to integer / float	df["col"].astype(int) / .astype(float)
Several columns	df.astype({"a": "int64", "b": "float64"})
Messy numbers	pd.to_numeric(df["col"], errors="coerce")
Integers with NaN	df["col"].astype("Int64")
Dates	pd.to_datetime(df["col"], format="%Y-%m-%d")
Repeated labels	df["col"].astype("category")
Automatic	df.convert_dtypes()

Related pandas tutorials:

• Convert floats to integers in pandas
• Remove all non-numeric characters in pandas
• Read a CSV file with pandas
• Convert a pandas DataFrame to JSON

Frequently asked questions