📈 economics

A Novel Subgroup Decomposition of Income Inequality: Resolving the Information Mixture Problem

This paper proposes a novel subgroup decomposition methodology based on the Unequally Distributed (UD) and Relative UD (RUD) income framework to resolve the "information mixture problem" in conventional approaches by isolating a foundational egalitarian baseline, thereby enabling a mathematically consistent partitioning of income inequality into pure within-group and between-group components.

Original authors: Joongyang Park, Muhammad Hamza, Youngsoon Kim

Published 2026-08-03

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Joongyang Park, Muhammad Hamza, Youngsoon Kim

Original paper licensed under CC BY 4.0 (https://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

The Great Economic Puzzle: Why We Need a Better Ruler

Imagine you are trying to figure out why a group of friends has such different amounts of candy. Some have a mountain, some have a few pieces, and some have none. Economists call this "income inequality," and they spend a lot of time trying to measure it. But here's the tricky part: to understand why the candy is uneven, you have to break the problem down into smaller pieces. You want to know: Is the unfairness because the groups themselves are different (like one group getting a huge haul of candy while another gets a tiny one)? Or is it because, even within a single group, some kids are hoarding more than others?

For decades, the standard way to measure this has been like using a ruler that starts at the "average" height of the group. If you measure everyone's height from the average, you get a number that tells you how spread out they are. But there's a hidden flaw in this method. It mixes up two different things: the fact that everyone has some candy (the baseline) and the fact that the extra candy is distributed unfairly. It's like trying to measure how messy a room is by measuring the distance of every object from the center of the room, even though the floor itself is covered in a thick, uneven layer of carpet that everyone is standing on. If you don't account for that carpet, your measurement of the mess is wrong. This paper dives into that specific problem, proposing a new way to measure the "mess" that ignores the carpet and focuses only on the scattered toys.

The Carpet and the Toys: A New Way to Measure the Mess

The authors of this paper, Joongyang Park, Muhammad Hamza, and Youngsoon Kim, argue that the old way of measuring income inequality is suffering from what they call the "information mixture problem." To fix this, they invented a new mathematical tool based on a concept called "Unequally Distributed (UD) income."

Here is how their new method works, using a simple analogy. Imagine a society as a group of people standing on a floor. The old method looks at how far everyone is from the "average" height. The new method says, "Wait a minute! First, let's cut off the floor."

The Egalitarian Baseline (The Floor): The authors say every person has a "minimum" amount of income that is shared by everyone, like a floor everyone stands on. In their math, this is the lowest income in the group.
The Unequal Surplus (The Toys): Once you remove that shared floor, what's left is the "surplus." This is the extra income that some people have and others don't. This is the "unequally distributed" part.
The New Ruler (RUD Income): They then take this surplus and measure it against a "perfectly equal" state where everyone has zero surplus (because they all just have the floor).

The paper proposes a specific formula, called the $L^2_{2,q}$ index, to measure the distance between the actual distribution of these "toys" and a perfect state of zero toys. Think of it as measuring the total energy of the scattered toys. If everyone has the same amount of extra candy, the toys are neatly stacked (low inequality). If one kid has a mountain and the rest have none, the toys are scattered everywhere (high inequality).

What They Found: The Magic of Pure Math

The authors didn't just guess; they built a mathematical proof to show why their new ruler is better. They demonstrated that their method solves two big headaches that plague the old methods:

No More "Overlap" Confusion: Old methods, like the famous Gini coefficient, often create a confusing "overlap" term when you try to split the data into groups (like splitting the candy into "boys" and "girls"). It's like trying to add two numbers and getting a third, mysterious number that doesn't belong to either. The authors' new method is "strictly additive." This means if you calculate the inequality for Group A and Group B separately, and then add them up, you get the exact total inequality of the whole society. No mystery numbers, no leftovers.
The "Floor" Doesn't Break the Math: A major problem with other methods is that if the "floor" (the minimum income) changes, the whole measurement gets messed up. The authors proved that their new index, the $L^2_{2,q}$ , stays consistent even when the floor moves. It respects a rule called the "Pigou-Dalton Transfer Principle," which basically says: "If you take a dollar from a rich person and give it to a poor person, inequality should go down." Their math shows this happens perfectly, even if the poorest person's floor rises.

A Test Drive with Five People

To prove their point, the authors ran a simulation with a tiny society of just five people with incomes of 1, 2, 3, 4, and 5.

The Old Way: Would measure the spread around the average (3).
The New Way: First, it removes the "floor" of 1. The new incomes become 0, 1, 2, 3, and 4. Then, it scales them so they are relative to the average.
The Result: They split this group into two: a "Lower Tier" (1 and 2) and an "Upper Tier" (3, 4, and 5). Using their new formula, they could perfectly calculate how much of the total "mess" was caused by the difference between the two tiers versus the mess inside each tier. The math added up perfectly: the sum of the parts equaled the whole, with no errors.

Why This Matters (And What's Next)

The paper concludes that by using this new "Relative Unequally Distributed" (RUD) framework, we can finally see the true structure of economic inequality without the distortion of the baseline. It's like cleaning the lens of a camera so you can see the actual picture, not the glare.

However, the authors are careful to note that while the math is solid, using this in the real world has a challenge. To use this method on real data, you need to know exactly where the "floor" (the minimum income) is. In a real economy, finding that exact minimum is tricky and depends on how you model the data. The authors suggest that future research needs to figure out the best way to estimate this floor, perhaps by looking at how other types of data (like the size of raindrops or the lifespan of lightbulbs) are modeled.

In short, this paper doesn't just offer a new number; it offers a new way of thinking. It suggests that to truly understand why some people have more than others, we first have to agree on what "having nothing" looks like, and then measure everything else from there. It's a fresh perspective that promises to make our economic maps much clearer, provided we can figure out how to draw that starting line.

Technical Summary: A Novel Subgroup Decomposition of Income Inequality

Problem Statement
The paper identifies fundamental theoretical flaws in conventional income inequality decomposition frameworks, specifically the "information mixture problem" and the "variation within an income distribution problem." Traditional methods, including those based on the Generalized Entropy class and the Gini coefficient, typically anchor decomposition on the arithmetic mean of total income. The authors argue that this approach conflates two distinct components of income: an egalitarian baseline (the minimum income shared by all individuals, representing realized equality) and an unequal surplus (the income remaining above this baseline). By measuring dispersion around the mean, conventional indices inadvertently mix the evaluation of the egalitarian baseline with the unequal surplus. Furthermore, because these indices measure internal statistical dispersion rather than the pure geometric departure from a state of perfect equality (the zero-vector), any additive decomposition of these indices inherently inherits and magnifies this mismeasurement. This leads to structural distortions, such as the complex "overlap" or "transvariation" terms found in Gini decompositions, which violate strict additivity and subgroup consistency.

Methodology
To resolve these issues, the paper proposes a novel subgroup decomposition methodology based on the Unequally Distributed (UD) and Relative UD (RUD) income framework. The core of the approach involves the following steps:

Extraction of the Egalitarian Baseline: The authors define the egalitarian baseline as the minimum income ( $y_1$ ) in a distribution. This baseline is subtracted from every individual's income to generate the UD income vector ( $x_i = y_i - y_1$ ). In a state of perfect equality, the UD income vector is the absolute zero-vector.
Normalization for Scale Invariance: To ensure the measure is scale-invariant, the UD income is transformed into RUD income ( $z_i = x_i / \mu_y$ ), where $\mu_y$ is the mean of the original total income. This preserves the zero-vector baseline while guaranteeing that the index remains constant under uniform scaling of incomes.
Selection of the $L^2_{2,q}$ Index: The authors utilize the $L^2_{2,q}$ index, defined as the mean of the squared RUD incomes ( $L^2_{2,q} = \frac{1}{n}\sum z_i^2$ ). This index is mathematically equivalent to $\mu_z^2 + V_z$ , where $\mu_z^2$ represents the squared geometric distance between the societal mean and the zero-vector (the departure from equality), and $V_z$ represents the variance of the RUD income (the internal variation of the surplus).
Decomposition Strategy: The paper proposes two decomposition approaches:
- Variance Decomposition: The total variance ( $V_z$ ) is partitioned into between-group ( $V_{z, Between}$ ) and within-group ( $V_{z, Within}$ ) components using standard ANOVA logic applied to the RUD incomes. This isolates structural disparities without residual cross-terms.
- Population-Weighted Decomposition: Following Hamza et al. (2026), the total index is expressed as a population-weighted sum of subgroup-specific $L^2_{2,q}$ indices, allowing for a direct comparison of inequality levels across subgroups.

Key Contributions

Resolution of the Information Mixture Problem: By extracting the egalitarian baseline prior to decomposition, the proposed framework ensures that only the pure, unequally distributed surplus is partitioned. This prevents the contamination of inequality measures with the magnitude of the shared baseline.
Mathematical Consistency and Additivity: The $L^2_{2,q}$ index satisfies Strict Additivity and Subgroup Consistency. Unlike the Gini coefficient, which produces complex overlap terms, this method allows for a perfect additive partition of total inequality into within-group and between-group components.
Axiomatic Robustness: The paper demonstrates that the proposed decomposition satisfies essential axioms, including Scale Invariance, Population Independence, and the Pigou-Dalton Transfer Principle. Notably, the authors show that the $L^2_{2,q}$ index correctly captures the systemic impact of raising the foundational baseline (e.g., when a progressive transfer increases the minimum income), a feature where CDF-based indices (like those relying on the Gini coefficient) may fail.
Geometric Interpretation: The methodology reframes inequality as the squared geometric distance of the RUD income vector from the absolute zero-vector, providing a clear Euclidean interpretation of structural disparities.

Results
The paper validates the methodology through a stylized numerical example involving a society of five individuals partitioned into two subgroups. The results demonstrate:

Exact Decomposition: The total $L^2_{2,q}$ index is calculated as 6/9 (equivalent to 2/3). This value is perfectly decomposed into the sum of population-weighted subgroup indices and into the sum of between-group and within-group variance components.
Structural Clarity: The decomposition successfully isolates the share of total inequality attributable to the variation within the RUD distribution and further partitions that variation into differences between subgroups versus internal subgroup dispersion. Specifically, the share of total inequality attributed to the variation within $z$ is 1/3, and the share of total inequality attributed to the difference between subgroups is 1/4.
Verification of Properties: The numerical application confirms that the index remains invariant under population replication and that the decomposition holds without residual terms.

Significance
The paper claims that this methodology provides a mathematically rigorous and "uncontaminated lens" for viewing societal stratification. By proposing a framework that avoids the structural distortions inherent in variance-based measures around the arithmetic mean, the authors offer a precise tool for identifying the root structural drivers of economic inequality. They argue that this approach is essential for designing targeted redistributive policies, as it distinguishes between inequality caused by the size of the egalitarian baseline and inequality caused by the distribution of the surplus. While the paper acknowledges the future challenge of estimating the minimum threshold ( $y_1$ ) in empirical microdata, it posits that this decomposition framework is the necessary theoretical foundation for resolving the "information mixture problem" that has plagued conventional inequality analysis.