📈 economics

Global Ease of Living Index: a machine learning framework for longitudinal analysis of major economies

This paper introduces a reproducible machine learning framework that addresses data gaps and utilizes dimensionality reduction to construct a comprehensive, longitudinal Global Ease of Living Index for major economies since 1970, providing policymakers with a transparent tool to identify and improve key socio-economic and infrastructural areas.

Original authors: Arun Kumar Selvaraj, Tanay Panat, Rohitash Chandra

Published 2026-06-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Arun Kumar Selvaraj, Tanay Panat, Rohitash Chandra

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine trying to grade a student's report card, but instead of just looking at their math and reading scores, you want to know how happy, healthy, safe, and well-off they are overall. Now, imagine doing this for every country in the world, but with a huge problem: some of the report cards have missing pages, torn corners, or entire subjects that were never graded because the data wasn't collected.

That is exactly what the authors of this paper set out to do. They built a new "report card" for the world called the Global Ease of Living Index (Global-EoLI). Here is how they did it, explained simply:

1. The Problem: Missing Pieces of the Puzzle

The world is full of different ways to measure how good life is (like the "Happiness Index" or "GDP"), but the authors argue these are flawed.

The "Happiness" Trap: Some indices rely on asking people, "Are you happy?" But the paper says this is like asking a shy person if they are loud; cultural biases and how people want to be seen can skew the results.
The Missing Data: Many countries, especially developing ones, don't have records for everything. Maybe they forgot to track the cost of bread for 20 years, or they didn't count how many doctors are in a specific region. If you just throw away the countries with missing data, you lose a huge chunk of the picture.

2. The Solution: The "AI Repair Shop"

To fix the missing pages in the report cards, the authors used Machine Learning as a repair shop.

The Analogy: Imagine you have a jigsaw puzzle with 40% of the pieces missing. You can't just guess randomly. Instead, you look at the pieces you do have. If you see a pattern of blue sky and green grass, you can reasonably guess that the missing piece in the middle is probably also blue or green.
The Tool: They used two smart computer models (called Random Forest and MICE) to act as expert guessers. These models looked at the relationships between different facts (like how GDP usually relates to life expectancy) to fill in the blank spots with the most likely numbers. They tested these models to make sure the "guesses" fit the original picture perfectly.

3. Building the Index: The Four Pillars

Once they had a complete set of data, they didn't just add everything up. They organized the data into four main "pillars" or categories, like the four legs of a sturdy table:

Economic: How much money is there? (Growth, jobs, cost of living).
Institutional: How well does the government run? (Is the law fair? Is there corruption? Can people speak freely?).
Quality of Life: How healthy and safe are people? (Life expectancy, healthcare access, crime rates).
Sustainability: Is the planet being treated well? (Pollution, renewable energy, carbon emissions).

4. The "Magic Filter": Finding the Real Story

The authors had hundreds of different numbers (indicators). To make sense of them, they used a statistical technique called Factor Analysis.

The Analogy: Imagine you have a bag of 100 different colored marbles. Some are red, some are blue, some are green. Factor Analysis is like a magic filter that sorts them into four distinct buckets based on their color patterns. It tells you, "Okay, these 25 numbers all move together and represent 'Economic Health,' while these 15 numbers represent 'Government Quality.'"
This allowed them to create four clear sub-scores for each country, which they then combined into one final score.

5. What They Found: The Big Picture

They ran this index from 1970 to 2022 and looked at major economies like the US, China, India, Australia, and Germany.

The Developed Nations: Countries like Australia, Germany, and Japan consistently scored high. They have strong "tables" with all four legs (economy, government, life quality, and environment) standing tall.
The Rising Giants: Countries like China and India showed massive growth in their Economic leg. They are building the table very fast! However, their Sustainability and Quality of Life legs are still catching up, meaning they are growing rich but still face challenges with pollution and healthcare access.
The "Happiness" vs. "Ease of Living" Clash: The paper compared their new index to the famous "World Happiness Index." They found a big disconnect. For example, India ranks much higher on the new "Ease of Living" index (51st) than on the "Happiness" index (139th).
- Why? The authors argue the Happiness Index is flawed because it relies on small surveys that don't represent billions of people and focuses too much on money (GDP) without considering how expensive life is. Their new index says, "India might not be the 'happiest' in a survey, but the actual conditions of living there are improving and are better than the Happiness Index suggests."

6. The Takeaway

The authors didn't just make a list of numbers; they built a transparent, reproducible tool.

Open Source: They put all their code and data on the internet (GitHub) so anyone can check their work or use it to update the index.
For Policymakers: This tool helps leaders see exactly where their country is weak. If a country is doing great on money but terrible on pollution, the index lights up a red flag on the "Sustainability" leg, telling them, "Hey, fix this specific part of the table."

In short, this paper is about building a fairer, more complete report card for the world, using smart computers to fill in the blanks, so we can see the true state of how people live, not just how much money they make.

Technical Summary: Global Ease of Living Index (Global-EoLI)

Problem Statement
Existing indices for assessing quality of life and ease of living, such as the World Happiness Index or the Global Livability Index, suffer from significant limitations. These include a heavy reliance on subjective self-reported data (which introduces cultural bias and social desirability effects), a lack of transparency in weighting methodologies, and an over-reliance on aggregated data that may not reflect local realities. Furthermore, many indices fail to adequately capture the multidimensional nature of well-being, often isolating economic factors from socio-economic and infrastructural conditions. A critical technical challenge is the prevalence of missing data in historical economic and social indicators, particularly for developing nations and specific time periods (e.g., Cost of Living Index missing from 1970–2012, Healthcare Index missing up to 92.44% in certain periods). Simple imputation methods like mean substitution or listwise deletion can distort underlying data patterns and lead to information loss.

Methodology
The authors propose a machine learning framework to construct a longitudinal Global Ease of Living Index (Global-EoLI) covering major economies from 1970 to 2022. The methodology proceeds through seven distinct steps:

Data Collection: Data was aggregated from sources including the IMF, WHO, and World Bank, covering 70 countries. The dataset includes four primary sub-indices: Economic, Institutional, Quality of Life, and Sustainability.
Data Imputation: To address missing values (ranging from ~2% to over 90% for specific variables), the study compares two advanced imputation techniques:
- Random Forest Regressor (RFR): An ensemble method capable of capturing complex non-linear relationships without strong distributional assumptions.
- Multiple Imputation by Chained Equations (MICE): A statistical technique that generates multiple plausible datasets to account for imputation uncertainty.
- Selection: The study evaluates both methods using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) on synthetic missing data. RFR demonstrated superior performance, particularly in handling non-linear and complex environmental data, and was prioritized for the final index construction. Kernel Density Estimation (KDE) plots were used to verify that imputed distributions aligned with observed data.
Preprocessing and Standardization: Data was grouped by country and year. Ordinal and categorical variables were removed prior to standardization. Continuous variables underwent z-score standardization ( $x - \mu / \sigma$ ) after imputation to ensure the scaling parameters were derived from the complete dataset, preventing bias.
Dimensionality Reduction and Sub-Index Construction:
- Principal Component Analysis (PCA): Used as an exploratory tool to determine the appropriate number of latent factors based on eigenvalue analysis (Kaiser criterion) and to assess the variance explained.
- Factor Analysis (FA): Employed with varimax rotation as the primary method to extract latent dimensions and compute factor scores. This step identified four distinct factors corresponding to the four sub-indices (Institutional, Quality of Life, Economic, and Sustainability). The Kaiser-Meyer-Olkin (KMO) test was used to verify sampling adequacy; notably, the Sustainability sub-index showed weaker adequacy (KMO < 0.5 with RFR), leading the authors to treat its results with caution and prefer MICE-based specifications for this specific dimension.
Weighting and Aggregation: The four sub-indices were aggregated into the final Global-EoLI using fixed weights inspired by the Indian Ministry of Housing and Urban Affairs' Ease of Living Index but adjusted to reduce reliance on purely economic factors. The assigned weights were:
- Quality of Life: 0.35
- Economic: 0.25
- Institutional: 0.25
- Sustainability: 0.15
Longitudinal Analysis: The resulting index was used to rank countries annually from 1970 to 2021, with a focus on G20 members and a mix of developed (e.g., US, Germany, Australia) and developing (e.g., India, China, Malaysia) nations.

Key Results

Imputation Performance: RFR outperformed MICE in terms of RMSE and MAE across most sub-indices, particularly in the Sustainability and Quality of Life domains where data relationships are non-linear.
Factor Structure: Factor Analysis successfully identified four coherent latent factors. The Institutional factor was dominated by governance indicators (Rule of Law, Control of Corruption), while the Quality of Life factor was driven by human development metrics (HDI, Life Expectancy).
Country Rankings (2021): Developed nations (Australia, Japan, Canada, Germany) consistently ranked high in the Global-EoLI. Australia ranked 4th, followed by Japan (8th) and Canada (17th). Emerging economies like India (51st) and China (41st) showed lower overall rankings despite strong economic growth, highlighting gaps in institutional stability, quality of life, and sustainability.
Comparison with Happiness Index: The study found significant disparities between the Global-EoLI and the World Happiness Index. For instance, India ranked 51st on the Global-EoLI but 139th on the Happiness Index. The authors attribute this to the Happiness Index's reliance on small sample sizes (2,000–3,000 respondents) for massive populations and its heavy weighting of GDP per capita without adjusting for cost of living.
Robustness: Sensitivity analysis using alternative weighting schemes (e.g., equal weights, higher sustainability weights) showed that rankings for top-performing developed nations remained relatively stable, whereas emerging economies showed more volatility, reflecting their uneven performance across the four dimensions.

Significance and Claims
The paper claims to offer a transparent, reproducible, and comprehensive framework for assessing living standards that addresses the limitations of existing indices. Key contributions include:

Methodological Rigor: The integration of machine learning (RFR) for handling high levels of missing data in longitudinal datasets, ensuring data completeness without resorting to simplistic deletion.
Multidimensional Perspective: By combining Economic, Institutional, Quality of Life, and Sustainability dimensions, the index provides a more holistic view of well-being than GDP-centric or single-factor indices.
Policy Utility: The framework serves as a practical tool for policymakers to identify specific areas requiring intervention (e.g., healthcare systems, public safety, environmental sustainability) across different economic contexts.
Transparency: The authors emphasize that the code and data are open-source, allowing for independent verification and application to other contexts.

The authors modestly acknowledge limitations, noting that variables with extremely high missingness (e.g., >80%) rely heavily on model-based reconstruction and should be interpreted with caution. They also note that the Sustainability sub-index exhibited weaker factor structure reliability compared to other dimensions. The study concludes that while the Global-EoLI improves upon existing metrics, future work should explore Bayesian inference for uncertainty quantification and incorporate more granular environmental metrics.