An economics and business concept defining a measure, method, or organizational practice used for analysis and decision-making. It specifies how information is generated or used to guide allocation of resources and evaluation of outcomes. It does not ensure correctness without clear assumptions, reliable inputs, and appropriate review of results. It materially affects planning, performance, and risk by shaping decisions and incentives within organizations and markets. The concept is generally stable, though methods and tools evolve over time.

Principle

When residuals are serially correlated (Cov(εi, εj) ≠ 0 for i ≠ j in an ordered sequence), OLS standard errors are invalid and OLS may be inefficient; in dynamic models, autocorrelation often signals omitted lags or persistent shocks.

Demonstration

Quarterly GDP growth regressions can exhibit positive autocorrelation in residuals because economic shocks and adjustments persist across quarters, so a large positive residual this quarter increases the chance of a positive residual next quarter.

Misapplication

Estimating time series regressions with lagged dependent variables by OLS without accounting for autocorrelation, then using conventional t-tests and confidence intervals as if residuals were independent.

Consequence

Detecting autocorrelation leads to remedies such as including lagged variables, employing autoregressive integrated models, using HAC (Newey–West) standard errors, or applying GLS; these restore valid inference and improve model fit by capturing dependence.

Reversal

Independence of residuals: errors are uncorrelated across the ordering, which validates the use of standard OLS inference procedures based on i.i.d. or independent errors assumptions.

Boundary

Specifically relevant to ordered data (time series, panel with time dimension, spatial data); ordinary cross-sectional analyses without ordering typically do not invoke autocorrelation though spatial dependence is a related concept.

Semantic Tension

Tension between addressing autocorrelation through model augmentation (adding dynamics) and treating it purely as a nuisance to correct standard errors; the former changes structural interpretation while the latter preserves coefficient estimates but adjusts inference.

Synthesis

Autocorrelation indicates dependence across ordered observations; practitioners must detect its presence, determine whether it reflects substantive dynamics or misspecification, and choose modeling or inference corrections that reflect the underlying data-generating process.