Efficient Estimation of Mixed-Frequency Cointegrating Regressions
What is it about?
Temporally aggregating or sampling series suppresses or omits information, causing inefficiency. When one series has been aggregated -- a mixed-frequency data scenario, I derive efficiency bounds conditional on the type of aggregation employed. I show how the unaggregated series may be used to recover some of the lost information and increase efficiency in estimating the cointegrating vector, achieving a gain over aggregating the remaining series to a single frequency.
Why is it important?
The most accepted way of handling series observed at different frequencies is to aggregate all series to the lowest frequency. Along the lines of the proliferation of mixed-frequency approaches in the recent literature, this analysis shows how the high-frequency information can be exploited for a gain in efficiency with a cost of only minimal complication added to estimation.
The following have contributed to this page: Professor J. Isaac Miller
In partnership with: