🌍 World Happiness Report — Data Cleaning, EDA & Analysis

HappyMaxx Consulting helps governments understand what makes people happier and which policies improve quality of life. Using global happiness data, we identify key factors — such as healthcare, economic stability, and social support — linked to higher wellbeing. Our goal is to help countries make data-driven decisions that create happier and healthier societies.

Maximising Happiness, One Data Point at a Time.

---

📌 Project Overview

This project performs a full end-to-end data pipeline on the World Happiness Report dataset (2015–2019), sourced from Kaggle, enriched with UN World Population data. It covers data cleaning and preparation, exploratory data analysis (EDA) in both Python and SQL, and a final presentation of insights for a consulting context.

The analysis spans 782 observations, 156 countries, and 10 world regions across five years.

---

🗂️ Repository Structure — How to Navigate This Repo

happiness_miniproject/
│
├── README.md                          ← You are here
│
├── data/
│   ├── HAPPINESS.csv                  ← Raw happiness data (2015–2019, concatenated from Kaggle)
│   ├── population_df.csv              ← UN population data (total pop, density, median age)
│   └── HAPPY_POPULATION.csv           ← ✅ Final clean dataset — start here for analysis
│
├── notebooks/
│   ├── readme_happiness.ipynb         ← Project overview and pipeline documentation
│   ├── FINAL_clean.ipynb              ← Data cleaning pipeline: merges happiness + population data
│   └── FINAL_clean_with_graphs.ipynb  ← EDA notebook: all final visualizations and statistical analysis
│
├── sql/
│   └── MiniProject_happiness.sql      ← MySQL schema + exploratory SQL queries
│
├── HappyMaxx_EDA_Presentation.pptx.pdf             ← 📊 Full project presentation (PDF)
├── MiniProject_happiness_2026-05-21T13_51_57.png   ← MySQL schema diagram

New to the repo? Start with HAPPY_POPULATION.csv if you just want the clean data, or FINAL_clean_with_graphs.ipynb if you want to explore the full analysis interactively.

---

📁 File Descriptions

Data Files

File	Description
HAPPINESS.csv	Raw concatenated happiness data from Kaggle's World Happiness Report CSVs (2015–2019). Contains one row per country per year with happiness score, GDP, life expectancy, freedom, generosity, and corruption scores. Column names are inconsistent across years — see cleaning notebook.
population_df.csv	Population data sourced from UN World Population Prospects. Contains total population, population density (per km²), and median age per country per year.
HAPPY_POPULATION.csv	The final analysis-ready dataset. Result of merging the two sources above after cleaning. 782 rows × 14 columns, zero nulls. This is the file used for all EDA and visualizations.

Notebooks

File	Description
readme_happiness.ipynb	Project documentation notebook. Describes the full data pipeline, column standardization logic, region reconstruction steps, and the final DataFrame schema. Good entry point to understand what was done before touching any code.
FINAL_clean.ipynb	Data cleaning and merging pipeline. Loads HAPPINESS.csv and population_df.csv, applies country name aliases (e.g. North Cyprus → Cyprus, Somaliland Region → Somalia), merges on country_key + year, handles the single remaining null (UAE 2018 corruption filled with country median), and exports HAPPY_POPULATION.csv.
FINAL_clean_with_graphs.ipynb	Full EDA and visualization notebook. Contains all final graphs produced for the analysis: correlation heatmap, GDP vs happiness scatter, life expectancy vs happiness scatter, freedom vs happiness scatter, corruption vs happiness scatter, happiness distribution by region (bar chart), happiness score evolution by region (line chart, 2015–2019), and top 10 happiest and saddest countries (horizontal bar charts). All charts are colored by region and include trend lines.

SQL File

File

Description

MiniProject_happiness.sql

MySQL schema definition and exploratory SQL queries. Creates the normalized relational schema (region → country → year → year_country → happiness) with foreign keys and cascade rules. Includes queries for: average happiness by region, top country per region (using ROW_NUMBER window function), top 10 happiest countries overall, countries above their regional average, happiness–density cross-classification (using CASE buckets), global happiness trend by year, and the relationship between GDP wealth level and life expectancy.

📊 Presentation

The full project presentation is also available on google slides via this link: https://docs.google.com/presentation/d/1rRlBj_aE-nJ3TuXSggVruYRB2JgS3_Yv/edit?usp=sharing&ouid=106830695445682190716&rtpof=true&sd=true

---

📦 Requirements

pip install pandas numpy matplotlib seaborn pycountry

For the SQL component:

• MySQL Workbench or any MySQL-compatible client
• Run MiniProject_happiness.sql to create the schema and execute the exploratory queries

---

🔄 Pipeline Overview

Step 1 — Load the raw data

Five annual Kaggle CSVs (2015–2019) are loaded individually. Each year has between 155 and 158 country records.

Step 2 — Standardize column names

A clean_column_names() function strips whitespace, converts to lowercase, and replaces spaces/dots with underscores. A rename_columns() function unifies field names across years using keyword matching:

Normalized name	Examples of raw names
happiness_score	Score, Happiness Score
happiness_rank	Rank, Happiness Rank
gdp_percapita	Economy (GDP per Capita), GDP per capita
life_expectancy	Health (Life Expectancy)
family	Social support, Family
government_corruption	Trust (Government Corruption)
freedom	Freedom to make life choices

Step 3 — Data quality checks

• Duplicates: checked across all five DataFrames — none found.
• Null values: only one null detected — government_corruption for UAE in 2018, filled with the country's own median across all years (preferred over mean due to an outlier 2019 value).

Step 4 — Clean and standardize country names

The pycountry library resolves country names to their official ISO forms. Edge cases are handled with a country_name_fixes dictionary. A country_clean column is created alongside the original.

Step 5 — Reconstruct the region column

The region column only exists in 2015 and 2016 data. A country_region_dict is built from those years and mapped onto 2017–2019. Countries still missing a region are assigned one via manual_region_fixes. A final null check confirms all rows have a region.

Step 6 — Merge with population data

HAPPINESS.csv is merged with population_df.csv on [country_key, year] using a left join. Country name mismatches between the two sources are resolved with aliases:

country_aliases = {
    "North Cyprus":      "Cyprus",
    "Macedonia":         "North Macedonia",
    "Somaliland Region": "Somalia",
    "Somaliland region": "Somalia",
}

Step 7 — Export final dataset

The cleaned, merged DataFrame (happy_population_df) is saved as HAPPY_POPULATION.csv. Final shape: 782 rows × 14 columns, zero nulls.

---

📊 Final Dataset: HAPPY_POPULATION.csv

Column	Type	Description
happiness_rank	int	Country ranking by happiness score for that year
year	int	Survey year (2015–2019)
country	str	Standardized country name (ISO via pycountry)
region	str	World region (10 categories)
happiness_score	float	Overall happiness score (scale 0–10)
gdp_percapita	float	GDP per capita contribution to happiness score
life_expectancy	float	Healthy life expectancy contribution
family	float	Social support / family contribution
freedom	float	Freedom to make life choices contribution
generosity	float	Generosity contribution
government_corruption	float	Perceived absence of corruption contribution
total_pop	float	Total population
pop_density	float	Population density (people per km²)
age_pop	float	Median age of the population

Key stats:

Metric	Value
Global mean happiness	5.41
Global median happiness	5.37
Std deviation	1.11
Highest score	7.63 — Finland
Lowest score	2.69 — South Sudan
Countries	156
Regions	10
Years	2015–2019

---

🔬 EDA — Exploratory Data Analysis

The EDA was conducted in parallel through three tools: Python (notebook + scripts) and MySQL.

Python EDA (FINAL_clean_with_graphs.ipynb)

Descriptive statistics were computed for all 10 numeric variables, including mean, median, mode, standard deviation, min, max, skewness, and kurtosis.

Correlation analysis (Pearson r with happiness score):

Variable	r	Strength
GDP per capita	0.79	Strong
Life expectancy	0.74	Strong
Median age	0.68	Moderate–Strong
Family / social support	0.65	Moderate
Freedom	0.55	Moderate
Government trust	0.40	Moderate
Generosity	0.14	Weak
Population density	0.08	Negligible
Total population	−0.04	Negligible

Visualizations produced:

1. Correlation heatmap — all happiness variables (lower-triangle, RdYlGn palette)
2. GDP per capita vs happiness score — scatter colored by region with trend line
3. Life expectancy vs happiness score — scatter by region with trend line
4. Freedom vs happiness score — scatter by region with trend line
5. Corruption vs happiness score — scatter by region with trend line
6. Average happiness score by region — bar chart sorted descending
7. Happiness score evolution by region — line chart 2015–2019
8. Top 10 happiest countries — horizontal bar chart
9. Top 10 saddest countries — horizontal bar chart

SQL EDA (MiniProject_happiness.sql)

SQL queries explored the same data from a relational perspective:

• Average happiness by region — simple aggregate ranking confirming Python results
• Top country per region — window function (ROW_NUMBER OVER PARTITION BY region) to identify each region's best performer
• Top 10 happiest countries overall — aggregated across all years
• Countries above their regional average — correlated subquery to find overperformers within each region
• Happiness × density cross-classification — CASE bucketing of both happiness level (Very Low / Low / Medium / High) and population density (Sparse / Medium / Dense / Very Dense) to look for patterns
• Global happiness trend by year — yearly average to check whether happiness improved 2015–2019
• GDP wealth level vs life expectancy — CASE buckets on GDP show a striking 3× gap in life expectancy scores between the wealthiest and poorest country groups

---

🔬 Hypotheses & Conclusions

Hypothesis 1 — Economic and Social Development & Happiness ✅ Confirmed

Hypothesis: Countries with higher GDP per capita, stronger social support systems, and higher life expectancy will report higher happiness scores, suggesting that both economic prosperity and social wellbeing contribute positively to national happiness levels.

Result: Confirmed. GDP per capita (r = 0.79) is the single strongest predictor of happiness in the dataset, followed closely by life expectancy (r = 0.74) and family support (r = 0.65). Countries in the highest GDP tier average a score of 6.81 versus 4.00 for the lowest tier — a 1.7× gap. Western Europe, North America, and Australia/New Zealand dominate the top rankings. One notable outlier: Latin American countries consistently score above their GDP level, driven by strong family and social support scores.

---

Hypothesis 2 — Demographics & Happiness ⚠️ Partially Rejected

Hypothesis: Population characteristics such as total population size, population density, and median age may influence national happiness levels, with more demographically stable and developed populations expected to show higher average happiness scores.

Result: Partially rejected. Total population (r = −0.04) and population density (r = 0.08) show virtually no linear relationship with happiness — large countries are not happier, and dense countries are not sadder (Singapore and the Netherlands rank highly despite extreme density). However, median age (r = 0.68) is a meaningful predictor: countries with older populations score substantially higher on average. This is because median age is a strong proxy for demographic maturity — nations with older populations have typically sustained decades of investment in healthcare, nutrition, and institutional stability.

---

Hypothesis 3 — Profile of High-Happiness Countries ✅ Confirmed

Hypothesis: Countries with high happiness scores are expected to share common structural characteristics, including strong economic performance, low perceived corruption, high personal freedom, and strong institutional trust.

Result: Confirmed. The top 10% of happiest countries share a clear and consistent profile compared to the bottom 10%:

Variable	Top 10%	Bottom 10%	Ratio
GDP per capita	1.37	0.37	3.7×
Life expectancy	0.89	0.29	3.0×
Government trust	0.28	0.12	2.4×
Freedom	0.58	0.28	2.0×
Family support	1.38	0.65	2.1×
Generosity	0.33	0.23	1.4×

The "dream country" formula is: wealthy + healthy + free + trustworthy institutions. Generosity shows the weakest gap, suggesting it is more a by-product of happiness than a driver of it.

---

🗺️ Regions Covered

10 world regions are represented:

Region	Mean Happiness (2015–2019)
Australia and New Zealand	7.30
North America	7.18
Western Europe	6.76
Latin America and Caribbean	6.02
Eastern Asia	5.65
Central and Eastern Europe	5.43
Middle East and Northern Africa	5.35
Southeastern Asia	5.34
Southern Asia	4.58
Sub-Saharan Africa	4.19

---

📈 Technologies Used

Tool	Purpose
Python 3	Primary analysis language
Pandas	Data cleaning, merging, and manipulation
NumPy	Numerical operations and correlation calculations
Matplotlib & Seaborn	Data visualization and statistical graphics
pycountry	ISO country name standardization
Jupyter Notebook	Interactive development environment
MySQL Workbench	Relational schema design and SQL-based EDA

---

🔮 Recommended Next Steps

• Run multivariate regression models to quantify each variable's independent contribution to happiness, controlling for GDP
• Add a geospatial layer to map happiness scores interactively across the world
• Extend the dataset beyond 2019 to capture the impact of the COVID-19 pandemic on global wellbeing
• Build a policy simulator: model which specific interventions would move a country up the happiness ranking

---

✒️ Authors

Irene Fafián and Bibian González