Dependecy of personal income on various factors by Egor Chernyi

Introduction

Income inequality and wage disparities are widely studied in economics and social sciences. Understanding the key factors that influence personal income can help individuals make informed career choices and assist policymakers in addressing wage gaps. This study aims to analyze the dependency of personal income on various factors, including age, education level, occupation, gender, marital status, and working hours. Using the Adult Income Dataset, which contains demographic and economic attributes of individuals, we will investigate which variables have the most significant impact on income levels. The key research question we seek to answer is: “How does personal income depend on various factors?” To address this question, I will: 1. Explore the dataset – Understand its structure, clean the data, and identify relevant variables. 2. Analyze statistical relationships – Examine correlations and distributions of income across different demographic and professional groups. 3. Visualize patterns – Use graphs and charts to illustrate trends and dependencies. By the end of this project, I aim to identify the most influential factors in shaping personal income, assess whether common assumptions—such as higher education leading to higher wages—hold true, and evaluate the effectiveness of a predictive model in classifying individuals based on their income level.

import numpy as np # linear algebra import pandas as pd

Run to view results

Data collecting and cleaning

df = pd.read_csv('data.csv') df.head()

Run to view results

df.info()

Run to view results

# Count unique column dtypes in the data frame df.dtypes.value_counts()

Run to view results

# Number of unique values in the columns df.nunique()

Run to view results

df.shape

Run to view results

# Checking unique values for each columns def check_unique_value(): for col in df.columns: unique_values = df[col].unique() print(f"Unique Values in column {col} : \n {unique_values} \n") check_unique_value()

Run to view results

By observing the unique values in the columns, I noticed that there are values that are "?" which represents unknown value. The categorical columns values also have a space in front, so we'll remove them to reduce confusion.

categorical_columns = df.select_dtypes(['object']).columns df[categorical_columns] = df[categorical_columns].map(lambda x: x.strip())

Run to view results

unknown_value_columns = df.columns[df.isin(['?']).any()] unknown_value_columns

Run to view results

(df[unknown_value_columns] == '?').sum()

Run to view results

We will replace the "?" values with the most frequent values in each of the columns.

for col in df[unknown_value_columns]: mode_value = df[col].mode()[0] print(f"Most frequent unique value in {col} column : {mode_value}") df[col].replace("?", mode_value, inplace = True) print(f"\nSum of unknown value (?) rows in the columns : \n{(df[unknown_value_columns] == '?').sum()}")

Run to view results

# Checking Null Values df.isnull().sum()

Run to view results

Data Analysis

import seaborn as sns import matplotlib.pyplot as plt

Run to view results

df_clean = df.copy()

Run to view results

How personal income affected by age

# Age Distribution plt.figure(figsize = (10,6)) sns.histplot(data = df_clean, x = 'age', bins = 20, hue = 'salary', palette = 'dark', kde = True) plt.xlabel('Age') plt.ylabel('Frequency') salary_label = ['More than 50k', 'Less than 50k'] plt.legend(labels = salary_label, title = 'Salary range', ) plt.title('Age Distribution of working adults by salary range') plt.show()

Run to view results

Higher-income individuals tend to be older. The average age for high earners is higher (~44 years) than that of lower earners (~37 years).

average_age = df_clean.groupby('salary')['age'].mean() print(average_age)

Run to view results

# T-statistic and p-value from scipy import stats t_stat_age, p_value_age = stats.ttest_ind(df_clean[df_clean['salary'] == 0]['age'], df_clean[df_clean['salary'] == 1]['age'], equal_var = False) print("T-test is: ", t_stat_age) print("P-value is: ", p_value_age)

Run to view results

T-test also shows significance of age regarding personal income/ Possible explanation: -Older indivuduals have more work experience, leading to higher salaries. -Career growth and promotions often come with age -Younger indivuduals might still be in education, internships, or entry-level jobs Key insights: Age is correlated with income, ut there may be a plateau where salary growth slows down at a certain age

What employment type is highly rewarded

# Workclass type distribution plt.figure(figsize = (12,4)) sns.countplot(data = df_clean, x = 'workclass', hue = 'salary', palette = 'deep') plt.xlabel('workclass') plt.ylabel('Frequency') plt.title('workclass by salary range') plt.legend(labels = salary_label, title = 'Salary Range') plt.show()

Run to view results

Private sector jobs dominate the dataset, but higher salaries are more common in self-employment. Key findings: - Private-sector is the most common work class, but mostly low-icnome earners - Self-employed has a higher proportion of high-earners, possibly due to business ownership and flexible earning potential - Government jobs are more stable, often requiring higher education, leading to relatively etter pay - Unpaid workers and those who never worked are not represented. Possible explanation: - Job type and industry matter: government and self-employed individuals may have specialaozed skills or ownership benefits - Career security: Government jos may offer structured promotions and pay raises over time

Race and income

# Race distribution fig, axes = plt.subplots(1, 2, figsize = (12, 6)) sns.countplot(data = df_clean, y = 'race', hue = 'salary', palette = 'pastel', ax = axes[0]) axes[0].set_xlabel('Frequency') axes[0].set_ylabel('Race') axes[0].legend(loc = 'lower right', title = 'Salary Range', labels = salary_label) axes[1].pie(df_clean.race.value_counts(), labels = df_clean.race.value_counts().index, autopct = '%1.1f%%', colors = sns.color_palette('pastel')) plt.title('Race Distribution') plt.show()

Run to view results

The distribution of income across races is uneven. The White racial group has the highes proportion of high earners Key findings: - White individuals dominate the high-income group - Other racial groups, including Black, Asian, and Native American Individuals, have fewer high-income earners Possible explanation: - Systematic factors such as access to higher education and better job opportunities could contribute to these diffrences. - Occupational disparities and historical inequalities in USA may influence wage gaps

Personal icnome dependency on education level

# Education distribution plt.figure(figsize = (6, 10)) sns.countplot(data = df_clean, y = 'education', hue = 'salary', palette = 'pastel') plt.title('Education level by salary range') plt.ylabel('Education level') plt.xlabel('Frequency') plt.legend(title = 'Salary Range', labels = salary_label) plt.show()

Run to view results

Higher education levels strongly correlate with higher income. Key findings: - Individuals with Doctorate, Masters and Bachelors degrees have the highest-percentage of high-income earners - High-School graduates and below are mostly in the low-income group - Technical and Associate degrees offer moderate improvement in income potential Possible explanation: - Higher education typically leads to specialized skills and qualifications, making individuals eligble for higher-paying jobs - Many high-income careers require advanced degrees

Hard work pays off or personal income and working hours dependency

# Hours per week plt.figure(figsize = (8,6)) sns.histplot(data = df_clean, x = 'hours-per-week', bins = 10, color = 'orange', kde = True) plt.title('Working hours per week by adults') plt.xlabel('Working hours per week') plt.ylabel('Frequency') plt.show()

Run to view results

High-income individuals tend to work more hours per week on average. Key findings: - Most low-income individuals work around 40 hours per week - Higher- income individuals tend to work more than 40 hours per week, often on demanding or specialized roles -- Some individuals working fewer hours ut still earning high income might be on high-paying fields or self-employed Possible explanation: - Working more hiurs can increase total earnings, especially in hourly wage jobs - Some high-income professions require long-hours - Part-time jobs are more common among low-income earners

Final thoughts

This dataset provides valuable insights into income dependency on various factors. Next steps of this research could be: 1. In-depth statisticall analysis - Correlation analysis - Chi-square test etc. 2. Predictive modelling - Logistic Regression: Since income is binary, logistic regression could be used to predict whether an individual earns a high salary based on age, education, work class, and working hours

.css-15w88e5{color:var(--chakra-colors-fg-neutral-primary);font-weight:inherit;letter-spacing:-0.09px;}Dependecy of personal income on various factors by Egor Chernyi

Introduction

Data collecting and cleaning

Data Analysis

How personal income affected by .css-o90n1z{color:var(--chakra-colors-fg-neutral-primary);font-weight:inherit;letter-spacing:-0.09px;-webkit-text-decoration:underline;text-decoration:underline;}age

What employment type is highly rewarded

Race and income

Personal icnome dependency on education level

Hard work pays off or personal income and working hours dependency

Final thoughts

Dependecy of personal income on various factors by Egor Chernyi

How personal income affected by age