How to Calculate Percentage Change in Pandas

1. Use pct_change() to Calculate Percentage Change in Pandas
2. Fill Missing Values Before Calculating the Percentage Change in Pandas
3. Calculate Percentage Change for Multi-Index DataFrame in Pandas

The pct_change() is a function in Pandas that calculates the percentage change between the elements from its previous row by default. In the case of time series data, this function is frequently used.

The output of this function is a data frame consisting of percentage change values from the previous row. We can specify other rows to compare as arguments when we call this function.

It works on the following formula.

Use pct_change() to Calculate Percentage Change in Pandas

This method accepts four optional arguments, which are below.

1. periods - having default value 1. It specifies the periods to shift to calculate the percent change.
2. fill_method - Specifies how to handle NAs before calculating the percentage change.
3. limit - Specifies the amount of consecutive NAs to fill before stopping.
4. freq - Increment to use from time series API (e.g. 'M' or BDay()).

The following is a simple code to calculate the percentage change between two rows. We will call the pct_change() method with the data frame object without passing any arguments.

# Python 3.x
import pandas as pd

df = pd.DataFrame([[2, 4, 6], [1, 2, 3], [5, 7, 9]])
print(df.pct_change())

Output:

Fill Missing Values Before Calculating the Percentage Change in Pandas

For example, we have missing or None values in the data frame. When calculating the percentage change, the missing data will be filled by the corresponding value in the previous row.

Here, ffill means forward fill.

# Python 3.x
import pandas as pd

df = pd.DataFrame([[2, 4, 6], [1, None, 3], [None, 7, 9]])
print(df.pct_change(fill_method="ffill"))

Calculate Percentage Change for Multi-Index DataFrame in Pandas

We can also calculate percentage change for multi-index data frames. We can split the data into groups according to some criteria using the groupby() method then apply the pct_change().

# Python 3.x
df = pd.DataFrame(
    index=pd.MultiIndex.from_product(
        [
            ["Jhon", "Alia"],
            ["CS"],
            ["Python", "Java", "Dart"],
            ["Mid Term", "Final Term"],
        ],
        names=["Student", "Department", "Course", "Exam"],
    ),
    data={"Marks": [50, 40, 30, 60, 40, 40, 30, 70, 40, 50, 20, 30]},
)
print(df)
print(df.groupby(level=[1, 2, 3]).pct_change())

Output: