Income Distribution, Economic Inequality & Poverty Measurement

The mathematical study of personal income and wealth distribution forms a cornerstone of welfare economics and social policy. In an economy, aggregate production and national income may expand substantially, yet if gains accrue disproportionately to an elite minority, broad societal well-being and social cohesion may deteriorate. Statistical economics provides the formal analytical toolkit to quantify the size distribution of income, evaluate the degree of economic inequality, and measure the depth and severity of poverty.

In empirical microeconomics, the Size Distribution of Income arranges individual earners or households into ordered income brackets regardless of the institutional source of income (wages, profits, rents, or transfer payments). The resulting empirical income distributions invariably exhibit pronounced positive skewness (right-skewed): the vast majority of citizens cluster at low-to-moderate income levels, while an extended tail stretches into extremely high income brackets. Consequently, the arithmetic mean income $\bar{x}$ systematically exceeds the median income $x_{\text{med}}$, rendering standard symmetric Gaussian distributions inapplicable.

In 1897, the Italian mathematical economist and sociologist Vilfredo Pareto published his seminal empirical investigation into the upper tail of income distributions across European nations. Pareto discovered a striking empirical regularity: the number of individuals possessing an income greater than or equal to a specified threshold $x$ follows a power-law decay.

Mathematical Formulation

Pareto's Law is formulated as:

where:

• $x$ represents the level of income.
• $N(x)$ denotes the number of individuals earning an income of at least $x$.
• $x_0$ denotes the lower truncation threshold (the minimum subsistence income above which the power law holds).
• $A$ is a positive scale parameter proportional to total population size ($A = N x_0^\alpha$).
• $\alpha$ is the Pareto Index (or Pareto coefficient), a dimensionless shape parameter governing the speed of income decay.

Taking natural logarithms of both sides converts Pareto's equation into a linear specification:

When plotted on a double-logarithmic scale with $\ln(x)$ on the horizontal axis and $\ln[N(x)]$ on the vertical axis, Pareto's Law traces a downward-sloping straight line with constant slope $-\alpha$.

Probability Density Function and Moments

Normalized as a continuous probability distribution for a continuous random income variable $X \in [x_0, \infty)$, the Cumulative Distribution Function (CDF) is:

The complementary CDF (survival function) is:

Differentiating $F(x)$ with respect to $x$ yields the Probability Density Function (PDF):

The Expected Value (Mean Income):

The theoretical arithmetic mean income $\bar{x} = E[X]$ is given by:

Evaluating this improper integral:

Thus, the theoretical mean income is:

If $\alpha \le 1$, the integral diverges to infinity, indicating that the income distribution possesses an infinite mean—a condition reflecting hyper-concentrated aristocratic or plutocratic wealth. In empirical modern economies, $\alpha$ typically ranges between 1.5 and 2.5.

The Median Income:

The median income $x_{\text{med}}$ satisfies $F(x_{\text{med}}) = 0.50$:

Because $2^{1/\alpha} < \frac{\alpha}{\alpha - 1}$ for all $\alpha > 1$, the median is strictly less than the mean: $x_{\text{med}} < E[X]$, confirming the deep positive skewness of the Pareto distribution.

Economic Interpretation of $\alpha$:

The parameter $\alpha$ is an inverse measure of income inequality:

• A large $\alpha$ (e.g., $\alpha \ge 2.5$) signifies rapid decay in the upper tail, meaning very few individuals possess extreme fortunes, which corresponds to greater equality.
• A small $\alpha$ (e.g., $\alpha \to 1$) signifies a fat upper tail where vast wealth is concentrated among a tiny oligarchy, indicating severe inequality.

Introduced by American statistician Max Otto Lorenz in 1905, the Lorenz Curve is the universal graphical and mathematical device used to visualize and analyze economic inequality.

Continuous Mathematical Definition

Let $f(x)$ be the probability density function of income with finite positive mean $\mu = \int_0^\infty x f(x) dx$. For any given income cutoff $x^*$:

1. The Cumulative Proportion of Population earning income up to $x^*$ is:

1. The Cumulative Proportion of Total National Income earned by those individuals is:

Eliminating the intermediate threshold parameter $x^*$ defines the Lorenz curve as the continuous functional relationship:

Fundamental Mathematical Properties of the Lorenz Curve:

1. Boundary Conditions:

Zero percent of the population earns zero percent of income, and 100 percent of the population earns 100 percent of income.

2. First Derivative (Monotonicity):

By the chain rule of calculus:

The slope of the Lorenz curve at any point $p$ is equal to the ratio of the marginal individual's income $x^*$ to the national mean income $\mu$.

3. Second Derivative (Convexity):

Because $\mu > 0$ and $f(x^*) > 0$, the second derivative is strictly positive everywhere. Hence, the Lorenz curve is strictly convex (bowed downward toward the horizontal axis).

4. The Line of Perfect Equality:

If every individual in society earned the exact same income ($x = \mu$), then $x^* / \mu = 1$ everywhere, yielding $L(p) = p$. The 45-degree diagonal connecting $(0,0)$ to $(1,1)$ represents the Line of Absolute Equality.

5. The Line of Absolute Inequality:

If a single individual captured the entire national income while the remaining population earned zero, $L(p) = 0$ for all $p \in [0, 1)$ and $L(1) = 1$. The Lorenz curve coincides with the lower horizontal axis and the vertical boundary.

Formulated by Italian statistician Corrado Gini in 1912, the Gini Coefficient ($G$) is the most widely cited scalar measure of income inequality.

1. Geometric Definition via the Lorenz Curve

In the unit square $[0, 1] \times [0, 1]$, the total area under the 45-degree line of perfect equality is:

Let $B$ represent the area lying directly underneath the Lorenz curve:

Let $A$ represent the area of the concentration lens enclosed between the 45-degree line of equality and the Lorenz curve:

The Gini coefficient is defined as the ratio of area $A$ to the total area under the line of equality ($1/2$):

• Boundedness: $0 \le G \le 1$.
• $G = 0$ indicates perfect equality ($L(p) = p$).
• $G = 1$ indicates total inequality ($L(p) = 0$ for $p < 1$).
• In national economies, Gini coefficients range from approximately 0.24 (egalitarian Nordic nations) to over 0.60 (highly unequal developing or transition economies).

2. Gini from the Pareto Distribution

If income follows a continuous Pareto distribution with parameter $\alpha > 1$, the continuous Lorenz curve takes the explicit closed functional form:

Integrating $L(p)$ over $[0, 1]$:

Substituting into the Gini formula:

This closed-form formula illustrates the inverse relationship: as $\alpha \to 1$, $G \to 1$ (complete inequality); as $\alpha \to \infty$, $G \to 0$ (complete equality).

3. Gini from Grouped Empirical Data (Trapezoidal Rule)

When analyzing census or household income surveys grouped into $K$ discrete fractiles (e.g., quintiles, deciles), the empirical Lorenz curve is a piecewise linear polygon connecting coordinates $(p_i, q_i)$ for $i = 0, 1, \dots, K$, with $(p_0, q_0) = (0, 0)$ and $(p_K, q_K) = (1, 1)$.

The area under the piecewise linear Lorenz curve is the sum of $K$ trapezoids:

The Gini coefficient is computed via the Brown Formula:

For equal decile groupings where each fractile represents an equal population share $p_i - p_{i-1} = 1/K$:

While inequality measures dispersion across the entire income spectrum, Poverty Measurement focuses on the left tail—identifying individuals whose economic resources fall below a socially or biologically defined subsistence standard: the Poverty Line ($z$).

1. The Headcount Ratio ($H$)

The Headcount Ratio is the simplest and most widespread metric. Let $N$ be the total population, and let $Q$ denote the number of poor individuals whose income or consumption expenditure $y_i$ falls strictly below the poverty line ($y_i < z$):

Limitation: The Headcount Ratio captures only the incidence of poverty. It is completely insensitive to the depth of poverty: if a catastrophic shock halves the income of already destitute individuals, $H$ remains unchanged. Furthermore, it violates the Pigou-Dalton transfer principle: a regressive cash transfer from a destitute person to a marginally poor person that pushes the latter above $z$ paradoxically causes $H$ to decrease.

2. The Income Poverty Gap Ratio ($I$) and Poverty Gap Index ($\text{PGI}$)

To incorporate depth, the Income Gap Ratio ($I$) measures the average proportional shortfall of the poor from the poverty line:

where $\bar{y}_p$ is the mean income of the poor.

The Poverty Gap Index ($\text{PGI}$), also denoted as $P_1$, aggregates this shortfall across the entire national population:

$\text{PGI}$ measures the total financial resources per capita (expressed as a fraction of $z$) required to lift all poor individuals to the poverty line through perfectly targeted transfers.

3. The Foster-Greer-Thorbecke (FGT) Metric Family

In 1984, Erik Foster, Joel Greer, and Erik Thorbecke formulated the FGT Family of Poverty Indices ($P_\alpha$):

The parameter $\alpha$ acts as an inequality aversion parameter:

• $\alpha = 0$: $P_0 = Q/N = H$ (Poverty Incidence / Headcount Ratio).
• $\alpha = 1$: $P_1 = \text{PGI}$ (Poverty Depth / Poverty Gap Index).
• $\alpha = 2$: $P_2 = \text{SPGI}$ (Squared Poverty Gap Index or Poverty Severity Index). By squaring the proportional shortfalls, $P_2$ places progressively higher mathematical weight on individuals living furthest below the subsistence threshold. Unlike $P_0$ and $P_1$, $P_2$ satisfies both the Monotonicity Axiom and the Pigou-Dalton Transfer Axiom.

While the Gini coefficient provides an intuitive geometric measure of inequality, it suffers from a major mathematical limitation: it is not strictly sub-group decomposable. If a population is divided into distinct geographic regions or ethnic subgroups, national Gini inequality cannot be cleanly partitioned into the sum of within-group inequality and between-group inequality without generating an intractable overlapping residual term.

To overcome this, mathematical economists deploy the Generalized Entropy (GE) Family of Inequality Measures, rooted in Claude Shannon's information theory.

Mathematical Formulation

For a discrete population of $n$ individuals with incomes $y = [y_1, y_2, \dots, y_n]$ and arithmetic mean $\mu = \frac{1}{n} \sum y_i$, the Generalized Entropy Index $GE(\alpha)$ is defined as:

The parameter $\alpha \in (-\infty, \infty)$ represents sensitivity to income disparities at differing parts of the distribution:

• For large positive $\alpha$ (e.g., $\alpha = 2$), the index is sensitive to income concentration at the very top of the distribution.
• For $\alpha = 1$, the index is the celebrated Theil's $T$ Index.
• For $\alpha = 0$, the index is Theil's $L$ Index (also known as the Mean Log Deviation, MLD), which is particularly sensitive to distress in the lower tail.
• For $\alpha = 2$, $GE(2) = \frac{1}{2} \left( \frac{\sigma}{\mu} \right)^2 = \frac{1}{2} CV^2$, which is half the squared coefficient of variation.

Perfect Sub-Group Decomposability

Suppose a nation's population is partitioned into $K$ mutually exclusive subgroups ($k = 1, \dots, K$), where subgroup $k$ has population $n_k$, mean income $\mu_k$, population share $v_k = n_k / n$, and aggregate income share $s_k = (n_k \mu_k) / (n \mu)$. Every Generalized Entropy index decomposes into:

where:

For the Mean Log Deviation ($\alpha = 0$), the weights simplify to population shares:

This property enables policymakers to evaluate what percentage of national inequality stems from disparities within provinces versus average income gaps between provinces.

In 1970, British economist Anthony B. Atkinson formulated an explicit welfare-theoretic approach to inequality measurement. Atkinson argued that every statistical inequality metric embodies an implicit ethical value judgment regarding social welfare.

1. Equally Distributed Equivalent Income ($y_{\text{ede}}$)

Let society possess an additive, utilitarian Social Welfare Function over individual incomes:

where $U(y)$ exhibits constant relative inequality aversion $\epsilon \ge 0$:

Atkinson defined the Equally Distributed Equivalent Income ($y_{\text{ede}}$) as that hypothetical level of uniform per-capita income which, if enjoyed by every single individual in society, would generate the exact same level of total social welfare as the currently observed unequal income distribution:

Solving for $y_{\text{ede}}$:

Because $U(y)$ is strictly concave for $\epsilon > 0$, by Jensen's inequality $y_{\text{ede}} < \mu$, where $\mu$ is the actual arithmetic mean income.

2. The Atkinson Inequality Index ($A_\epsilon$)

The Atkinson Index is defined as the proportional welfare loss attributable to unequal income distribution:

• Interpretation: An Atkinson index of $A_\epsilon = 0.20$ means that if income were distributed completely equally, society could achieve the same level of aggregate social welfare with only $80\%$ of its current total income. The remaining $20\%$ of national income is effectively "wasted" on inequality.
• The Inequality Aversion Parameter $\epsilon$:
• $\epsilon = 0$: Society is indifferent to inequality ($y_{\text{ede}} = \mu \implies A_0 = 0$).
• $\epsilon \to \infty$: Society adopts a Rawlsian maximin criterion, focusing exclusively on the welfare of the poorest citizen.

Traditional poverty measurement relied exclusively on unidimensional monetary indicators: whether a household's income or consumption expenditure falls below a monetary threshold $z$. However, as Nobel laureate Amartya Sen emphasized in his Capabilities Approach, poverty is inherently multidimensional: an individual may enjoy income above the subsistence line but suffer from severe deprivations in clean drinking water, basic healthcare, literacy, sanitation, and electrical energy.

To operationalize multidimensional poverty, Sabina Alkire and James Foster (2011) developed the Alkire-Foster Method, which underpins the United Nations Development Programme's global Multidimensional Poverty Index (MPI).

1. The Dual-Cutoff Identification System

Consider an achievement matrix $X$ of size $N \times D$, where $x_{ij}$ denotes the achievement of individual $i$ in dimension $j$ ($j = 1, \dots, D$). Each dimension has an explicit deprivation cutoff $z_j$ and an importance weight $w_j$ satisfying $\sum_{j=1}^{D} w_j = 1$.

• First Cutoff (Within-Dimension Deprivation Cutoff $z_j$):

Individual $i$ is deprived in dimension $j$ if $x_{ij} < z_j$. We construct the deprivation matrix $g^0$ where:

• Deprivation Score ($c_i$):

The weighted deprivation score for individual $i$ is:

• Second Cutoff (Cross-Dimensional Poverty Cutoff $k$):

An individual is identified as Multidimensionally Poor if their total weighted deprivation score equals or exceeds the cross-dimensional threshold $k$ (e.g., $k = 0.33$ or $33\%$ of dimensions):

We construct the Censored Deprivation Score $c_i(k)$:

Deprivations of non-poor individuals are censored to zero.

2. The Alkire-Foster Aggregation Metrics

1. Multidimensional Headcount Ratio ($H$):

The proportion of the population identified as multidimensionally poor:

where $q_k$ is the number of individuals with $c_i \ge k$.

2. Average Intensity of Poverty ($A$):

The average proportion of dimensions in which poor individuals are deprived:

3. Adjusted Headcount Ratio ($M_0$ / MPI):

The foundational Alkire-Foster poverty index is the product of incidence ($H$) and intensity ($A$):

3. Dimensional Decomposition of $M_0$

A vital mathematical property of $M_0$ is that it is strictly decomposable by dimension. The censored headcount ratio for dimension $j$, denoted $h_j$, is the proportion of the total population that is both multidimensionally poor and deprived in dimension $j$:

The overall index $M_0$ is an exact weighted sum of the dimensional censored headcounts:

The percentage contribution of dimension $j$ to national multidimensional poverty is:

This allows planners to identify whether national poverty is primarily fueled by lack of educational schooling, child malnutrition, or lack of clean water infrastructure, directly informing targeted sectoral investments.

In modern welfare economics and social justice theory, economists distinguish between inequality of outcomes (disparities in observed income or wealth) and Inequality of Opportunity. Pioneered by John Roemer (1998) and operationalized by François Bourguignon, Francisco Ferreira, and Vito Peragine, this framework partitions the determinants of an individual's economic success into two distinct categories:

1. Circumstance Variables ($C$): Factors entirely outside an individual's personal control, determined at birth or during childhood. These encompass parental educational attainment, parental socioeconomic status, geographic birth region, race, gender, and inherited wealth.

2. Effort Variables ($E$): Factors that are within the realm of individual agency, personal choices, and voluntary decisions—such as hours worked, educational investment, career occupational choice, and willingness to assume entrepreneurial risk.

Mathematical Formulation

Let an individual's income $y$ be determined by circumstances $C$, efforts $E$, and unobserved luck $u$:

To measure inequality of opportunity:

• Society is partitioned into $M$ mutually exclusive Types, where all individuals within type $m$ share the exact same circumstance profile $C_m$.
• Under the principle of Equality of Opportunity, disparities in income between types (attributable to circumstances) are ethically objectionable and represent social injustice, whereas disparities within a type (attributable to varying effort) are ethically legitimate.

Measuring the Opportunity Share

Using the Mean Log Deviation ($GE(0) = MLD$), which is perfectly sub-group decomposable:

The between-type inequality $MLD_{\text{between}}$ is computed by replacing each individual's observed income with the average income of their circumstance type $\bar{y}_m$:

The Inequality of Opportunity Ratio ($IOp$) is defined as:

In empirical studies across developing economies, between 30% and 60% of total national income inequality is directly attributable to circumstance variables determined at birth, demonstrating that high headline inequality largely reflects structural inequality of opportunity rather than differential effort.

Complementing static inequality metrics, Intergenerational Income Mobility measures the degree to which an individual's economic status is determined by the economic position of their parents.

1. The Intergenerational Income Elasticity (IGE)

The standard empirical model of intergenerational mobility regresses the log of child's adult permanent income ($y_{\text{child}}$) on the log of parental income ($y_{\text{parent}}$):

The parameter $\beta$ is the Intergenerational Income Elasticity (IGE):

• $\beta = 0$: Complete intergenerational mobility (a child's expected adult income is entirely independent of parental wealth; complete equality of opportunity).
• $\beta = 1$: Complete intergenerational immobility (a child's income reproduces their parents' position with zero social fluidity).
• Intergenerational Mobility is measured as $1 - \beta$. In egalitarian societies (Denmark, Norway), $\beta \approx 0.15$ to $0.20$. In highly immobile societies (United States, United Kingdom, Brazil), $\beta \approx 0.45$ to $0.60$.

2. The Great Gatsby Curve

Named by Alan Krueger in 2012, the Great Gatsby Curve maps the empirical cross-national relationship between static income inequality (measured by the Gini coefficient) on the horizontal axis and intergenerational income persistence (measured by IGE, $\beta$) on the vertical axis. The curve reveals a strong, positive empirical correlation: nations with higher static income inequality systematically exhibit lower intergenerational social mobility. When income gaps between social strata widen, wealthy families invest vastly greater resources in private schooling, elite tutoring, social networks, and bequest capital, cementing advantage across generations and locking disadvantaged children into persistent poverty traps.

Standard static poverty metrics measure whether a household's income falls below the subsistence threshold $z$ at a single survey instant. However, households constantly transition into and out of poverty due to agricultural droughts, macro-financial crises, illness, unemployment, or seasonal price spikes. To incorporate future uncertainty into social policy, economists analyze Vulnerability to Poverty.

Mathematical Formulation of Vulnerability as Expected Poverty

Pioneered by Stefan Dercon and Shubham Chaudhuri (2002), vulnerability $V_{it}$ is defined as the ex-ante probability that a household will experience poverty in the future period $t+1$:

where $f_{it}(y_{i, t+1})$ is the conditional subjective probability density function of future household consumption, dependent on observable household asset endowments $X_{it}$ and covariate community shock variables $S_t$.

Econometrically, log household consumption is modeled as:

where disturbances exhibit heteroskedasticity capturing risk exposure:

Assuming log-normality, the estimated vulnerability probability is:

where $\Phi(\cdot)$ is the standard normal cumulative distribution function.

Policy Taxonomy: Chronic vs Transient Poverty

A household is classified as vulnerable if its vulnerability probability exceeds a critical risk threshold (typically $V^* = 0.50$):

1. The Chronically Poor: Households whose expected long-run consumption is strictly below the poverty line ($E[y_i] < z$). They lack productive capital, arable land, and marketable education. Their poverty requires permanent structural interventions (asset transfers, education, land reform).

2. The Transiency Vulnerable: Households whose expected mean consumption is above the poverty line ($E[y_i] > z$), but whose exposure to uninsured environmental, macroeconomic, or health volatility frequently plunges them into poverty. Their vulnerability requires counter-cyclical safety nets (crop insurance, subsidized health coverage, emergency credit lines).

In 1976, Amartya Sen demonstrated that the classical Headcount Ratio ($H$) and Poverty Gap Index ($I$) violated two fundamental ethical criteria:

1. The Monotonicity Axiom: A reduction in the income of any poor household, holding all other incomes constant, must strictly increase the poverty index. (Violated by $H$).

2. The Transfer Axiom: A pure regressive transfer of income from a poor household to any richer household (even if both remain poor) must strictly increase the poverty index. (Violated by both $H$ and $I$).

The Sen Normalized Poverty Index ($S$)

Sen formulated an index explicitly integrating incidence ($H$), depth ($I$), and inequality among the poor ($G_p$):

where:

• $H = Q / N$ is the Headcount Ratio.
• $I = (z - \bar{y}_p) / z$ is the Income Gap Ratio.
• $G_p$ is the Gini coefficient calculated exclusively among the poor population.

Axiomatic Interpretation:

If all poor individuals have identical incomes ($G_p = 0$), Sen's index reduces to $S = H \times I = P_1$ (the standard Poverty Gap Index). As inequality among the destitute widens ($G_p > 0$), the index increases, satisfying both the Monotonicity Axiom and the Transfer Axiom by penalizing severe internal deprivation among the poorest of the poor.

While Corrado Gini's concentration coefficient weights the vertical distance between the 45-degree line and the Lorenz curve uniformly across the population, welfare economists often wish to place greater ethical sensitivity on disparities at specific locations along the income spectrum. Farhad Mehran (1976) and Walter Piesch (1975) formulated alternative linear rank-dependent inequality metrics.

1. The Mehran Inequality Index ($M$)

The Mehran Index assigns declining linear weights to individuals based on their population rank, placing greater mathematical emphasis on income inequality among the poorest citizens:

• In discrete population samples:

• The Mehran index is particularly sensitive to changes in income at the bottom of the distribution: a transfer from an impoverished citizen to a middle-class citizen produces a larger numerical increase in $M$ than in the Gini coefficient.

2. The Piesch Inequality Index ($P$)

Conversely, the Piesch Index assigns increasing linear weights to population rank, focusing sensitivity on concentration at the top of the income distribution:

• In discrete population samples:

• The Piesch index is sensitive to oligarchic wealth concentration and windfall executive earnings among the top percentiles.

By examining the triad of Gini, Mehran, and Piesch indices simultaneously, national statistical offices can ascertain whether rising national inequality is being driven primarily by falling wages at the bottom ($M$ rising fastest) or accelerating wealth accumulation at the top ($P$ rising fastest).