Poverty Gap Index
- Introduced by
- James Foster, Joel Greer & Erik Thorbecke (FGT alpha = 1)
- Year
- 1984
- Type
- Money-metric poverty depth measure
- Family
- Income poverty measurement
What does it do?
The poverty gap index is the member of the Foster-Greer-Thorbecke family at alpha = 1 and the standard money-metric measure of the depth of poverty.
Where the headcount ratio merely counts who is poor, the poverty gap averages how far the poor fall below the poverty line, expressed as a fraction of that line and spread over the whole population. It can be read as the per-capita resource shortfall — the share of the poverty line, per person, that perfect targeting would need to transfer to eliminate poverty — making it the natural complement to the headcount when judging the cost and intensity of poverty.
The intuition behind the method
Two countries can have the same number of poor people, yet in one the poor sit just below the line and in the other they are destitute. The headcount cannot tell them apart, but the poverty gap can: it measures, for each poor person, the distance between their income and the line as a fraction of the line, and then averages those distances across everyone. A poverty gap of 0.05 means that, on average over the whole population, incomes fall short of the poverty line by 5 percent of the line — equivalently, that 5 percent of one poverty line per person would, under perfect targeting, be enough to close all the shortfalls. It is the depth of poverty, not just its prevalence.
History
The poverty gap as a normalized, population-averaged shortfall was formalized as the alpha = 1 member of the Foster-Greer-Thorbecke class in their 1984 Econometrica paper. The idea responded to Amartya Sen's 1976 demonstration that the crude headcount ignores the depth of poverty and that a satisfactory measure must be sensitive to the size of shortfalls. By placing the gap inside a decomposable family, FGT made it both axiomatically respectable (monotonic) and operationally convenient (decomposable), and it has since been a standard reported statistic in World Bank poverty assessments and national poverty monitoring alongside the headcount and squared gap.
When to use it
Use the poverty gap index whenever the headcount ratio alone is inadequate because you care about how far the poor fall below the line, not merely how many they are. It is essential for assessing the depth of poverty, for estimating the minimum resources required to eliminate poverty under ideal targeting, and for detecting improvements among the poor that leave the headcount unchanged. Report it alongside the headcount (its prevalence partner) and the squared poverty gap (its severity partner) so that prevalence, depth, and inequality among the poor are all visible. Because it is additively decomposable, it also supports poverty profiling. Be aware that, like all FGT measures, it requires a defensible poverty line and a consistent welfare definition, and that it weights all shortfalls equally, so it is insensitive to inequality among the poor — for that, move to alpha = 2.
Limitations and considerations
- Insensitive to inequality among the poor: a transfer from a very poor to a less-poor person below the line leaves the index unchanged, so it violates the transfer axiom.
- Depends on an exogenous poverty line and a consistent welfare measure; results shift with the line and with equivalence-scale choices.
- As a money-metric measure it ignores non-income deprivations and the quality of price and income data.
- The minimum-transfer interpretation assumes perfect, costless targeting that no real program achieves, so it understates true elimination budgets.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Captures the depth of poverty, distinguishing populations with shallowly poor versus deeply poor members that the headcount treats identically.
- Satisfies the monotonicity axiom: it always falls when any poor person's income rises, rewarding gains that do not cross the line.
- Has a direct policy reading as the per-capita (and, when scaled, aggregate) minimum cost of eliminating poverty under perfect targeting.
- Additively decomposable across subgroups and factorable into headcount times income-gap ratio, aiding interpretation and profiling.
Applications
- Estimating the minimum per-capita and aggregate transfer needed to eliminate poverty under perfect targeting, as a budgetary benchmark.
- Tracking the depth of poverty over time to reveal whether the poor are getting closer to the line even when the headcount is flat.
- Evaluating transfer and social-protection programs by their effect on the depth, not just the prevalence, of poverty.
- Building poverty profiles that apportion the aggregate shortfall across regions and demographic groups for targeting.
- Comparing poverty depth across countries using internationally harmonized poverty lines.
How it works
How it works
For every individual compute the normalized poverty gap g_i = max((z − y_i)/z, 0), where z is the poverty line and y_i the welfare indicator. The gap is zero for anyone at or above the line and equals the fractional shortfall for the poor, rising toward one as income approaches zero. By construction it depends only on the poor (the focus axiom), and normalizing by z makes gaps comparable across lines and currencies.
Formula
g_i = \max\!\left(\frac{z - y_i}{z},\, 0\right), \qquad i = 1, \dots, NThe poverty gap index is the mean of these gaps over the entire population: PG = (1/N) sum of g_i. Averaging over N rather than over the number of poor q is deliberate — it preserves additive decomposability across subgroups and gives the index its per-capita-shortfall interpretation. Because it accumulates the size of shortfalls, the index falls whenever a poor person's income rises (even without crossing the line), so it satisfies monotonicity, which the headcount does not.
Formula
PG = P_1 = \frac{1}{N} \sum_{i=1}^{N} g_i = \frac{1}{N} \sum_{i : y_i < z} \frac{z - y_i}{z}The poverty gap factors neatly as PG = H × I, the product of the headcount ratio H = q/N and the income-gap ratio I, where I is the average proportional shortfall among the poor only. This separates the two ingredients of poverty depth: how widespread poverty is (H) and how intense it is on average among those affected (I). The decomposition clarifies why two populations with equal poverty gaps can differ markedly — one with many shallowly poor people, another with few but deeply poor people.
Formula
PG = H \cdot I, \qquad H = \frac{q}{N}, \quad I = \frac{1}{q}\sum_{i : y_i < z} \frac{z - y_i}{z}Multiplying the index by the poverty line and population recovers the absolute aggregate shortfall, T_min = z × N × PG, equal to the sum over the poor of (z − y_i). This is the minimum total transfer that would eliminate poverty under perfectly targeted, costless redistribution. Real programs cannot target perfectly, so T_min is a lower bound on the budget needed; comparing it with actual program costs reveals targeting inefficiency.
Formula
T_{\min} = z \cdot N \cdot PG = \sum_{i : y_i < z} (z - y_i)Common pitfalls
- Confusing the poverty gap index PG (averaged over the whole population) with the income-gap ratio I (averaged only over the poor); they differ by the factor H and answer different questions.
- Reading the minimum-transfer cost as an achievable program budget, ignoring imperfect targeting, leakage, and administrative cost.
- Treating the poverty gap as if it captured inequality among the poor; it does not — that requires the squared gap (alpha = 2).
- Comparing poverty gaps across surveys with different poverty lines, currencies, or welfare definitions without harmonization.
Frequently asked questions
How is the poverty gap index different from the income-gap ratio?
The income-gap ratio I is the average proportional shortfall computed only among the poor — it answers 'how poor are the poor, on average?' The poverty gap index PG spreads those shortfalls over the entire population, so PG = H × I, where H is the headcount. PG therefore combines how widespread poverty is with how intense it is, and only PG carries the per-capita-shortfall and additive-decomposability interpretations. Reporting I without H can make a population with very few but extremely poor people look the same as one with widespread mild poverty.
Why does the poverty gap not capture inequality among the poor?
Because it sums shortfalls linearly, a one-dollar transfer from a poorer person to a less-poor person (both below the line) reduces one gap and raises the other by the same amount, leaving the total unchanged. The measure is therefore indifferent to the distribution of poverty among the poor and violates the transfer axiom. To make the measure sensitive to such transfers you square the gaps, giving the FGT alpha = 2 squared poverty gap, which penalizes larger shortfalls more heavily.
Can the minimum-transfer cost be taken as the budget needed to end poverty?
Only as an idealized lower bound. The figure z × N × PG is the sum of all income shortfalls and would eliminate poverty only if a planner could transfer exactly each person's deficit at no administrative cost and with no behavioral response. Real programs face imperfect targeting, leakage to the non-poor, administrative overhead, and incentive effects, so the actual budget required is considerably larger. The gap-based figure is best used as a benchmark against which to judge targeting efficiency.
Related methods
Sources
Foster, J., Greer, J., & Thorbecke, E. (1984). A class of decomposable poverty measures. Econometrica, 52(3), 761–766.
Open DOI referenceCopied.