Gravity Model of Migration
Also known as: Migration Gravity Model, Demographic Gravity Model, Zipf P1P2/D Model, Gravity Model of Spatial Interaction (Migration)
The gravity model of migration explains the volume of movement between two places as proportional to the product of their populations (masses) and inversely proportional to the distance separating them, by direct analogy to Newton's law of universal gravitation. Formalized for intercity movement by George Kingsley Zipf in 1946 and embedded in regional science by Walter Isard, it is the workhorse model of human geography for predicting migration, commuting, and other spatial-interaction flows.
Key highlights
- Extremely parsimonious yet empirically robust: a handful of parameters explain a large share of variance in observed migration and commuting flows.
- Parameters are directly interpretable as elasticities — mass elasticities and a distance-decay exponent — that travel across studies and regions.
- Naturally extended with constraints (production-constrained, attraction-constrained, doubly constrained) to reproduce known origin and destination totals.
- Underpins entropy-maximizing and discrete-choice derivations, giving the simple analogy a rigorous behavioural and statistical foundation.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use the gravity model when you have origin–destination flow data — migrants, commuters, trips, trade — together with measures of origin and destination size and the distance or travel cost between them, and you want to explain or predict those flows and quantify distance decay. It is the default first model for inter-regional migration and is well suited to constructing flow forecasts when only marginal totals are known. It is less appropriate when flows are driven mainly by idiosyncratic policy, network, or chain-migration effects not captured by mass and distance, or when the independence-of-irrelevant-alternatives implied by the simple form is violated; competing-destinations and discrete-choice models are then preferable.
Strengths & limitations
- Extremely parsimonious yet empirically robust: a handful of parameters explain a large share of variance in observed migration and commuting flows.
- Parameters are directly interpretable as elasticities — mass elasticities and a distance-decay exponent — that travel across studies and regions.
- Naturally extended with constraints (production-constrained, attraction-constrained, doubly constrained) to reproduce known origin and destination totals.
- Underpins entropy-maximizing and discrete-choice derivations, giving the simple analogy a rigorous behavioural and statistical foundation.
- The Newtonian analogy is descriptive, not mechanistic: it does not model the individual decision process behind a move.
- Simple unconstrained forms ignore intervening opportunities and competing destinations, so they can misallocate flows in dense settlement systems.
- Distance is a crude proxy for the true cost, time, and information barriers that deter migration.
- Log-linear estimation breaks down with zero flows and is biased under heteroskedasticity, requiring Poisson-family estimators.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How is the gravity model of migration different from a general spatial interaction model?
The gravity model is the canonical, unconstrained member of the broader spatial-interaction family. Spatial-interaction modelling generalizes it with balancing factors and constraints — production-constrained, attraction-constrained, and doubly constrained forms — so that predicted flows reproduce known origin or destination totals. The migration gravity model is simply the gravity form applied to people moving between places, with population as the mass term.
Why use Poisson regression instead of taking logarithms?
Migration flow tables contain many zero cells, which have no logarithm and are silently dropped by log-linear OLS, and the log transform also biases coefficients when the error variance depends on the mean. Poisson pseudo-maximum-likelihood estimation models the counts directly, handles zeros, and remains consistent under heteroskedasticity, which is why it has become the recommended estimator.
What does the distance-decay exponent tell me?
The exponent γ measures how sharply interaction falls with distance: a larger γ means flows drop off faster, indicating that distance is a strong deterrent (typical of short-range, cost-sensitive moves), while a small γ indicates that interaction is relatively insensitive to separation (typical of long-range, information-rich migration). Comparing γ across flow types or eras reveals how the 'friction of distance' changes.
Sources
- 1.Zipf, G. K. (1946). The P1 P2 / D hypothesis: On the intercity movement of persons. American Sociological Review, 11(6), 677–686.
- 2.Isard, W. (1960). Methods of Regional Analysis: An Introduction to Regional Science. MIT Press.ISBN 9780262090032
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Gravity Model of Migration. ScholarGate. https://scholargate.app/human-geography/gravity-model-of-migration