Exogenous Variable
An exogenous variable is a variable whose value is determined outside the model or system under study. In an economic model, exogenous variables are inputs whose values are taken as given, while endogenous variables are determined within the model. The distinction between exogenous and endogenous variables is fundamental to model building, identification, and the interpretation of causal relationships.
Definition and contrast
In any model, the analyst chooses which variables to explain and which to treat as given. Endogenous variables are those the model seeks to determine, such as the equilibrium price and quantity in a supply-and-demand model. Exogenous variables are those whose values are supplied from outside, such as consumer income, technology, or government policy in a simple market model. A change in an exogenous variable shifts the equilibrium, while a change in an endogenous variable is the result of the model.
The distinction is model-specific, not absolute. A variable that is exogenous in one model may be endogenous in another. For example, government spending may be treated as exogenous in a short-run macroeconomic model but as endogenous in a political-economy model that explains fiscal decisions. The choice depends on the question being asked and the time horizon.
Role in economic models
Exogenous variables serve several purposes. They provide the shocks that drive model outcomes, such as changes in preferences, technology, resource availability, or policy. They define the environment within which agents optimise. They also serve as instruments in comparative statics, allowing the analyst to trace how a change in an external factor affects endogenous outcomes, holding other exogenous factors constant.
In econometrics, identifying causal effects requires exogenous variation in a variable of interest. If a variable is endogenous, for example because it is jointly determined with the outcome or affected by omitted factors, ordinary regression estimates may be biased, and techniques such as instrumental variables, natural experiments, or randomised controlled trials are used to isolate exogenous variation.
Comparative statics and dynamics
Comparative statics examines how the equilibrium of a model changes when an exogenous variable changes, holding other exogenous variables constant. For example, an increase in exogenous income raises the demand for a normal good, raising its equilibrium price and quantity. The method is widely used in microeconomics and macroeconomics to generate predictions that can be tested against data.
In dynamic models, exogenous variables may follow specified processes, such as a random walk or an autoregressive process, and the model traces the response of endogenous variables over time. In growth models, the rate of technological progress is often treated as exogenous, a limitation that endogenous growth theory seeks to overcome by explaining it within the model.
Interpretation and criticism
The choice of exogenous variables reflects theoretical assumptions and is subject to criticism. Lucas critique argues that treating policy parameters as exogenous and fixed is problematic because changes in policy alter the behaviour of agents, changing the very relationships the model assumes constant. This critique motivated models with microfoundations in which behavioural parameters, rather than reduced-form relationships, are treated as structural.
Deciding what is truly exogenous also raises the problem of causal inference. Variables commonly treated as exogenous, such as weather or oil prices, may be correlated with other determinants of outcomes, complicating identification. Robust model building therefore requires careful justification of exogeneity assumptions and, where possible, empirical tests.
Exogeneity in structural and reduced-form models
The treatment of exogeneity differs between structural and reduced-form modelling. Structural models specify behavioural relationships grounded in theory and seek to estimate deep parameters, such as preferences and technology. In these models, variables such as policy instruments and external shocks are treated as exogenous, while outcomes are endogenous. Reduced-form models summarise relationships among variables without specifying the full structure, and care must be taken not to interpret correlations as causal.
The concept of strict exogeneity in econometrics means that a variable is uncorrelated with the error term in all periods, a strong condition rarely satisfied in observational data. Weak exogeneity, a less demanding condition, is sufficient for certain estimation and inference procedures. Distinguishing among these concepts is important for valid inference and for policy analysis.
Sources of exogenous variation
Identifying exogenous variation is a central challenge in empirical economics. Researchers exploit natural experiments, such as changes in policy that are plausibly unrelated to the outcome, randomised controlled trials in development and field settings, and instrumental variables that affect the endogenous variable only through a specified channel. Discontinuities in eligibility rules, lotteries, and weather shocks are common sources of exogenous variation.
The credibility revolution in applied economics has emphasised research designs that isolate exogenous variation and thereby support causal claims. This has shifted the emphasis from elaborate structural models toward transparent identification strategies, while recognising that external validity, the extent to which results generalise, remains a separate concern.
Exogenous shocks in macroeconomics
In macroeconomics, exogenous shocks such as changes in oil prices, technological discoveries, or natural disasters are used to trace the dynamic response of the economy. Vector autoregression models, identified using restrictions derived from theory or from timing assumptions, estimate the effects of shocks on output, inflation, and interest rates. The distinction between shocks that originate outside the economic system and those generated by the system itself is fundamental to business-cycle analysis.
Policy variables occupy an ambiguous position. They are chosen by authorities and can be treated as exogenous to a private-sector model, but they also respond to economic conditions and are therefore endogenous in a broader political-economy sense. Careful modelling distinguishes the policy rule, which describes how instruments respond to conditions, from the shocks that move instruments away from the rule.
Exogeneity and policy evaluation
Policy evaluation relies on identifying exogenous variation in policy. When a policy change is announced and implemented independently of the outcome it is intended to affect, the change can serve as a source of exogenous variation, allowing estimation of the policy's causal effect. Randomised evaluations, in which treatment is assigned by lottery, provide the cleanest identification, but they are not always feasible for macroeconomic or large-scale policies.
Quasi-experimental methods, including difference-in-differences, regression discontinuity, and instrumental variables, exploit features of the data or institutional setting to approximate random assignment. These methods have transformed empirical economics by enabling credible causal inference from observational data, though they require careful attention to assumptions and to potential threats such as confounding and anticipation effects.
Exogeneity in structural macroeconomic models
In modern macroeconomic models, the distinction between exogenous and endogenous variables is central to the analysis of shocks and policy. Dynamic stochastic general equilibrium models specify exogenous shocks, such as technology or preference shocks, and derive the endogenous responses of output, inflation, and interest rates. Policy rules describe how instruments respond to conditions, and the effects of shocks are traced through impulse response functions.
The treatment of expectations is crucial. In models with rational expectations, agents use all available information, so the distinction between anticipated and unanticipated shocks matters. Anticipated policy changes can be neutral if agents adjust in advance, a result known as policy ineffectiveness, whereas unanticipated changes can have real effects. These insights shape the design and communication of macroeconomic policy.
Criticisms and refinements
The exogeneity assumption is frequently challenged. Variables treated as exogenous may be influenced by the system through feedback, omitted variables may confound relationships, and measurement error can bias estimates. Researchers address these issues with robustness checks, sensitivity analyses, and alternative identification strategies, and increasingly report the range of estimates consistent with the data rather than a single point estimate.