Overview
- Kurtosis is a statistical measure that describes the peakedness of the curve of a distribution. It is defined as the fourth central moment divided by the standard deviation.
- When the distribution is thin and tall it is called a Leptokurtic distribution.
- When the distribution is Leptokurtic both of the following things happen:
- Large number of small deviations from the mean
- Large number of large deviations from the mean
- In Investment, an asset with Leptokurtic returns can mean higher probability for extremely low or extremely high returns and associated higher value at risk.
- When the distribution is less peaked or flatter than the normal distribution, it is called Platykurtic distribution. In a Platykurtic distribution, the tail of the distribution is extremely thin with outliers less than that of the normal distribution.
- When the distribution is Mesokurtic the curve resembles that of a normal distribution curve.
Finding Kurtosis for a pandas Series
- The class Series from the Python library pandas implements a one-dimensional collection with several statistical and mathematical functions for Data Analysis.
- Series.kurtosis() function computes the Fisher’s kurtosis or Excess Kurtosis for the data present in the series. As per Fisher’s kurtosis - A leptokurtic distribution has a Kurtosis value greater than 0, a normal distribution or a mesokurtic distribution has a Kurtosis value of 0 and a Platykurtic distribution has a Kurtosis value smaller than 0. Kurtosis can be calculated for pandas.DataFrame columns and rows as well.
Example 1
The Fisher’s Kurtosis value found for the pandas.Series instance in this example is greater than 0 and hence the distribution present in the Series is Leptokurtic.
# Python example program to compute kurtosis of
# the distribution represented by a pandas.Series
import pandas as pds
# Percentage returns from the investment on an asset
returns = [3,3,10,3,5,4,5,10,4]
# Create pandas.Series instance
series = pds.Series(returns)
print("Kurtosis:")
print(round(series.kurtosis(), 2))Output
Kurtosis:
0.18Example 2
# Example Python program that creates a normal curve
# for the gold price per ounce for 100 days between 23Oct2025
# and 10Jun2025. Gold prices had a tremendous bull market
# in this time period which is reflected in the normal curve
# which is positively skewed and having a leptocurtic kurtosis.
import matplotlib.pyplot as plt
import numpy as np
from scipy.stats import norm
import pandas as pds
# 100 days gold prices per ounce
goldPrices =[4154.86, 4104.61, 4109.10, 4359.40, 4249.69, 4213.30, 4304.60, 4201.60,
4163.40, 4133.00, 4000.40, 3972.60, 4070.50, 4004.40, 3976.30, 3908.90,
3868.10, 3897.50, 3873.20, 3855.20, 3793.90, 3756.50, 3753.40, 3801.80,
3761.60, 3705.80, 3678.30, 3717.80, 3725.10, 3719.00, 3686.40, 3673.60,
3682.00, 3682.20, 3677.40, 3653.30, 3606.70, 3635.50, 3592.20, 3551.82,
3512.00, 3516.10, 3474.30, 3435.70, 3420.10, 3404.90, 3418.50, 3381.60,
3388.50, 3358.70, 3378.00, 3382.60, 3383.20, 3408.30, 3399.00, 3404.70,
3491.30, 3453.70, 3433.40, 3434.70, 3426.40, 3399.80, 3348.60, 3352.80,
3368.30, 3354.00, 3392.50, 3431.10, 3455.10, 3501.80, 3462.90, 3358.30,
3345.30, 3359.10, 3336.70, 3359.10, 3364.00, 3325.70, 3321.00, 3316.90,
3342.80, 3346.50, 3342.90, 3359.70, 3349.80, 3307.70, 3287.60, 3333.90,
3329.00, 3320.00, 3380.60, 3385.70, 3384.97, 3408.10, 3406.90, 3417.30,
3452.80, 3402.40, 3343.70, 3343.40]
series = pds.Series(goldPrices)
print("Kurtosis:")
print(round(series.kurtosis(), 2))
goldPrices = np.sort(goldPrices)
mean = np.mean(goldPrices)
std_dev = np.std(goldPrices)
# Compute the probability distribution function of the normal distribution
y = norm.pdf(goldPrices, loc = mean, scale = std_dev)
# Plot the normal curve
plt.plot(goldPrices, y, color = 'blue')
plt.axvline(mean, color='red', linestyle='solid',
label=f'Mean (μ) = {mean:.2f}')
# Draw vertical on both sides of the mean
# defining 1 standard deviation away from the mean
plt.axvline(mean + std_dev, color='green', linestyle='--',
label=f'1SD (±σ), Std Dev={std_dev:.2f}')
plt.axvline(mean - std_dev, color='green', linestyle='--')
# Draw vertical on both sides of the mean
# defining 2 standard deviation away from the mean
plt.axvline(mean + 2 * std_dev, color='violet', linestyle='--',
label=f'2SD (±2σ)')
plt.axvline(mean - 2 * std_dev, color='violet', linestyle='--')
# Give a title
plt.title('Gold prices - 23Oct2025 to 10Jun2025')
plt.xlabel('Price per ounce of gold in US Dollars')
plt.ylabel('Probability Density')
# Display the legends
plt.legend()
# Display the plot
plt.show()Output
Kurtosis:
0.07

Example 3
The Fisher’s Kurtosis value found for the pandas.Series instance in this example is less than 0 and hence the distribution present in the Series is Platykurtic.
# Python example program to compute kurtosis of
# the distribution represented by a pandas.Series
import pandas as pds
# Values of a distribution
vals = [2,2.2,2.3,2.1,1.9,2.3]
# Create pandas.Series instance
series = pds.Series(vals)
print("Kurtosis:")
print(round(series.kurtosis(), 2))Output
Kurtosis:
-1.48