Introduction

This project analyzes the economic effects of the 2018 U.S. aluminum tariffs introduced in March. Through a Difference-in-Differences (DID) framework, we evaluate the impacts of the tariffs on employment, Producer Price Index (PPI), and production within aluminum and non-aluminum industries. The analysis uses normalized datasets for clarity and comparability across variables.

Data Overview

Data

The datasets used in this project include: 1. Employment Data: FRED (Federal Reserve Economic Data) 2. PPI Data: FRED 3. Production Data: FRED 4. Global Aluminum Prices: Trade.gov

The time ranges vary: • Employment and production: 1990–2023 • PPI: 2005–2023

#Load and Prepare Data import pandas as pd import matplotlib.pyplot as plt import os import statsmodels.api as sm from sklearn.preprocessing import MinMaxScaler import numpy as np import seaborn as sns # ============================= # Step 1: Load and Prepare Data # ============================= # Load the datasets employment_alum_path = '/work/Project Datasets/Annual_Aluminum_Employment.csv' employment_non_alum_path = '/work/Project Datasets/Annual_NonAluminum_Emploment.csv' ppi_alum_path = '/work/Project Datasets/Annual_Aluminum_PPI.csv' ppi_non_alum_path = '/work/Project Datasets/Annual_NonAluminum_PPI .csv' prod_alum_path = '/work/Project Datasets/Annual_Aluminum_Production.csv' prod_non_alum_path = '/work/Project Datasets/Annual_NonAlum_Production.csv' global_price_path = '/work/Project Datasets/Annual_GlobalAlum_Price.csv' # Function to load and prepare data def load_and_prepare_data(alum_path, non_alum_path, global_price_path, year_start, year_end): # Load datasets alum_data = pd.read_csv(alum_path) non_alum_data = pd.read_csv(non_alum_path) global_price_data = pd.read_csv(global_price_path) # Assign group labels alum_data['Group'] = 'Aluminum' non_alum_data['Group'] = 'NonAluminum' # Ensure consistent column names alum_data.rename(columns={alum_data.columns[1]: 'Dependent_Var'}, inplace=True) non_alum_data.rename(columns={non_alum_data.columns[1]: 'Dependent_Var'}, inplace=True) # Combine aluminum and non-aluminum data combined_data = pd.concat([alum_data, non_alum_data], ignore_index=True) combined_data = combined_data[(combined_data['Year'] >= year_start) & (combined_data['Year'] <= year_end)] # Add Treatment, Post, and Treatment_Post columns combined_data['Treatment'] = (combined_data['Group'] == 'Aluminum').astype(int) combined_data['Post'] = (combined_data['Year'] >= 2018).astype(int) combined_data['Treatment_Post'] = combined_data['Treatment'] * combined_data['Post'] # Merge with global price data combined_data = pd.merge(combined_data, global_price_data, on='Year', how='left') # Normalize all relevant columns together scaler = MinMaxScaler() combined_data[['Dependent_Var', 'Global_Price']] = scaler.fit_transform(combined_data[['Dependent_Var', 'Global_Price']]) # Drop unnecessary columns combined_data = combined_data[['Year', 'Dependent_Var', 'Treatment', 'Post', 'Treatment_Post', 'Global_Price']] return combined_data # Function to plot regression results def plot_regression_results(data, title): """ Function to plot regression results with borders for improved visuals. """ plt.figure(figsize=(8, 6)) ax = sns.barplot(x=data.index, y=data.values, palette="Blues_d") for container in ax.containers: ax.bar_label(container, fmt="%.2f", label_type="edge", color="black", fontsize=10) plt.title(title, fontsize=14, weight='bold') plt.ylabel("Coefficients", fontsize=12) plt.xlabel("Variables", fontsize=12) plt.grid(axis='y', linestyle='--', linewidth=0.7, color='gray', alpha=0.7) plt.box(False) # Turn off the border box ax.spines['top'].set_visible(False) ax.spines['right'].set_visible(False) ax.spines['left'].set_edgecolor('black') ax.spines['left'].set_linewidth(1) ax.spines['bottom'].set_edgecolor('black') ax.spines['bottom'].set_linewidth(1) plt.show() # Function to run DID regression with customized dependent variable names def run_did_regression(data, dependent_var_name, independent_vars): y = data[dependent_var_name] X = data[independent_vars] X = sm.add_constant(X) model = sm.OLS(y, X).fit() # Set the dependent variable name in the results model.model.data.param_names[0] = dependent_var_name return model

Run to view results

Employment Regression

Objective:

Analyze the effect of the tariffs on employment in aluminum and non-aluminum sectors.

# Employment Regression employment_data = load_and_prepare_data(employment_alum_path, employment_non_alum_path, global_price_path, 1990, 2023) employment_data

Run to view results

# Run DID regression employment_model = run_did_regression(employment_data, 'Dependent_Var', ['Treatment', 'Post', 'Treatment_Post', 'Global_Price']) print(employment_model.summary()) # Plot the regression results for employment #employment_results = {'Treatment': -0.4423, 'Post': -0.1676, 'Treatment_Post': 0.1153, 'Global_Price': -0.3868} # Employment regression plot employment_df = pd.Series(employment_results) plot_regression_results(employment_df, "Regression Results - Employment")

Run to view results

Analysis:

The employment regression analyzes the effects of the 2018 tariffs on employment in the aluminum and non-aluminum sectors. The treatment variable shows a significant negative coefficient, indicating that industries directly affected by the tariffs experienced notable employment declines compared to non-aluminum industries. The post-tariff period variable also has a negative impact, suggesting broader labor market adjustments following the implementation of the tariffs. However, the interaction term (Treatment_Post) is insignificant, indicating limited evidence of a compounded effect on employment specifically due to the tariffs. This suggests that employment reductions might be attributed to factors beyond the tariffs.

Production Regression

Objective:

Evaluate the effects of tariffs on aluminum production.

# Prepare and run regressions for each dataset production_data = load_and_prepare_data(prod_alum_path, prod_non_alum_path, global_price_path, 1990, 2023) production_data

Run to view results

production_model = run_did_regression(production_data, 'Dependent_Var', ['Treatment', 'Post', 'Treatment_Post', 'Global_Price']) print(production_model.summary()) # Define the production regression results and plot #production_results = {'Treatment': -0.1830, 'Post': -0.1700, 'Treatment_Post': 0.1183, 'Global_Price': 0.0335} # Production regression plot production_df = pd.Series(production_results) plot_regression_results(production_df, "Regression Results - Production")

Run to view results

Analysis:

The production regression evaluates the changes in production levels within aluminum and non-aluminum industries post-tariff. The treatment coefficient is negative and statistically significant, reflecting a decline in aluminum production relative to non-aluminum production after the tariff. Interestingly, the post-tariff period also shows a slight decline, indicating possible systemic changes in production trends. The insignificant interaction term (Treatment_Post) suggests that the tariffs’ influence on production, while noticeable, may have been mitigated by external variables like global market dynamics.

PPI Regression

Objective:

Examine the impact of tariffs on Producer Price Index (PPI).

#Prepare and run regression ppi_data = load_and_prepare_data(ppi_alum_path, ppi_non_alum_path, global_price_path, 2005, 2023) ppi_data

Run to view results

ppi_model = run_did_regression(ppi_data, 'Dependent_Var', ['Treatment', 'Post', 'Treatment_Post', 'Global_Price']) print(ppi_model.summary()) # Define the production regression results and plot PPI_results = {'Treatment': -0.4839, 'Post': 0.2153, 'Treatment_Post': -0.1215, 'Global_Price': 0.2609} # PPI regression plot ppi_df = pd.Series(PPI_results) plot_regression_results(ppi_df, "Regression Results - PPI")

Run to view results

Analysis:

The PPI regression focuses on price effects in the aluminum and non-aluminum sectors. The treatment variable is significantly negative, showing a decrease in aluminum PPI compared to non-aluminum PPI during the tariff period. However, the interaction term (Treatment_Post) is again insignificant, indicating that the tariffs alone did not drive a substantial long-term difference in relative pricing trends. The global price variable is significant and positive, emphasizing that global market conditions significantly impacted pricing during the observed period.

Basic Visualizations

We began our exploration with simple visualizations of each variable to identify any observable trends or shifts before and after the tariff implementation in 2018. These plots allow us to assess whether the economic variables exhibit notable changes aligned with the policy’s introduction. These preliminary visualizations serve as a stepping stone for deeper quantitative analysis.

import pandas as pd import matplotlib.pyplot as plt from sklearn.preprocessing import MinMaxScaler # Function to normalize data def normalize_data(df, column_name): scaler = MinMaxScaler() df[column_name] = scaler.fit_transform(df[[column_name]]) return df # Function to plot comparison charts def plot_comparison(data1, column1, data2, column2, title, ylabel, xlim_range): plt.figure(figsize=(10, 6)) plt.plot(data1['Year'], data1[column1], label=f'Aluminum {ylabel}', color='blue') plt.plot(data2['Year'], data2[column2], label=f'Non-Aluminum {ylabel}', color='orange') plt.axvline(x=2018, color='red', linestyle='--', label='Tariff Implementation (2018)') plt.xlim(xlim_range) plt.title(f"{title}", fontsize=14) plt.xlabel("Year", fontsize=12) plt.ylabel(ylabel, fontsize=12) plt.legend() plt.grid(axis='y', linestyle='--', linewidth=0.7, alpha=0.7) plt.tight_layout() plt.show() # Load and normalize data # Employment emp_alum = pd.read_csv('/work/Project Datasets/Annual_Aluminum_Employment.csv') emp_non_alum = pd.read_csv('/work/Project Datasets/Annual_NonAluminum_Emploment.csv') emp_alum = normalize_data(emp_alum, 'Employment') emp_non_alum = normalize_data(emp_non_alum, 'Employment_NonAlum') # Plot charts # Employment Chart plot_comparison(emp_alum, 'Employment', emp_non_alum, 'Employment_NonAlum', "Employment Comparison: Aluminum vs. Non-Aluminum", "Normalized Employment", (2014, 2022))

Run to view results

# Production prod_alum = pd.read_csv('/work/Project Datasets/Annual_Aluminum_Production.csv') prod_non_alum = pd.read_csv('/work/Project Datasets/Annual_NonAlum_Production.csv') prod_alum = normalize_data(prod_alum, 'Production') prod_non_alum = normalize_data(prod_non_alum, 'NonAlum_Production') # Production Chart plot_comparison(prod_alum, 'Production', prod_non_alum, 'NonAlum_Production', "Production Comparison: Aluminum vs. Non-Aluminum", "Normalized Production", (2014, 2022))

Run to view results

# PPI ppi_alum = pd.read_csv('/work/Project Datasets/Annual_Aluminum_PPI.csv') ppi_non_alum = pd.read_csv('/work/Project Datasets/Annual_NonAluminum_PPI .csv') ppi_alum = normalize_data(ppi_alum, 'PPI') ppi_non_alum = normalize_data(ppi_non_alum, 'PPI_NonAlum') # PPI Chart plot_comparison(ppi_alum, 'PPI', ppi_non_alum, 'PPI_NonAlum', "PPI Comparison: Aluminum vs. Non-Aluminum", "Normalized PPI", (2014, 2022))

Run to view results

Conclusion

The analysis reveals that the 2018 tariffs had measurable impacts on employment, production, and prices within the aluminum sector. Employment and production in the aluminum industry declined relative to non-aluminum industries, highlighting the tariffs’ disruptive effects. However, the insignificant interaction terms across all regressions suggest that other external factors, such as global market trends, may have diluted the tariffs’ long-term effects. These findings underscore the complexity of trade policies and their ripple effects on domestic industries.

.css-15w88e5{color:var(--chakra-colors-fg-neutral-primary);font-weight:inherit;letter-spacing:-0.09px;}Introduction

Data Overview

Employment Regression

Production Regression

PPI Regression

Basic Visualizations

Conclusion

Introduction