mixedfrequency

Mixed frequency refers to data sets or statistical models that incorporate observations gathered at different sampling frequencies. It is common in economics and finance where monthly indicators (such as payrolls or industrial production) are combined with quarterly aggregates (like GDP), or where daily asset prices are analyzed alongside monthly or quarterly indicators. The goal is to leverage information across frequencies without discarding the temporal richness of high-frequency data.

Handling mixed-frequency data presents timing and alignment challenges, since high-frequency observations do not line up neatly with low-frequency targets. Approaches include temporal aggregation, where high-frequency data are collapsed to the lower cadence using averages, sums, or end-of-period values; and temporal disaggregation, where a low-frequency series is reconstructed from higher-frequency indicators using auxiliary information. Direct mixed-frequency models, such as MIDAS (mixed-data sampling) regressions, avoid full aggregation by incorporating high-frequency regressors with flexible lag structures in a low-frequency equation.

Other methods include Denton-type temporal disaggregation, Chow-Lin and related regression-based disaggregation techniques, and state-space or dynamic factor models that fuse information across frequencies through latent processes. These methods differ in assumptions about timing, causality, and measurement error.

Applications of mixed-frequency data are widespread in nowcasting and forecasting macroeconomic variables, policy analysis, and risk assessment in finance. They enable timely use of contemporaneous high-frequency indicators to improve inference about slower-moving targets, while requiring careful model specification to avoid overfitting.

See also: mixed-frequency data, nowcasting, MIDAS, temporal disaggregation, dynamic factor models.