A Data Science Central Community
PLEASE SEE THE REFERENCES IN THE COMMENTS BELOW TO GET THE WHOLE STORY
THIS IS A 'BLEEDING EDGE' "TECHNOLOGY" DEVELOPED BETWEEN BRUSSELS AND DUBLIN; BUT DREAMED UP BY THE ECB; viz;-
Recent studies undertaken by the ECB show that the so-called mixed frequency models, which are based on the use of monthly, quarterly and annual budgetary data, can provide forecasts of fiscal variables that are in some respects superior to methods relying exclusively on annual or quarterly data. Specifically, the three-frequency structure of these models allows to now cast quarterly figures by using monthly fiscal statistics and in turn to use quarterly and monthly figures to now cast annual data. Therefore, provided data are not taken at face value, a set of intra-annual fiscal data may provide useful information for those in charge of the monitoring of budget execution.
source: Public finances in EMU - 2011
Now-casting Irish GDP
Antonello D’Agostino, Kieran McQuinn and Derry O’Brien
Within individual quarters of the year, the approach, enables the data flow on monthly information during the quarter to be exploited. A pseudo-real time data approach is followed in that the data availability situation, which exists at each quarter is replicated for the model estimates. In evaluating the now-casting model, we perform an out of sample simulation where the estimates of the model are compared with that of a benchmark approach. We find that the mean squared forecast errors for both the now-casts and the back-casts are considerably smaller than those of the benchmark model. Unsurprisingly, the later in the quarter the now-cast or the backcast is generated, the more accurate the estimate is relative to the observed series.
simpler url link below (although the one above should work !!!)
I was originally interested in this approach becasue of my search for pragmatic approaches to Stress Testing and my interest in Layering Macroeconomic heuristics; but here (see the references below) we see that 'NowCasting' is a methodology almost; its a hybrid compromise or tractioning of DSGE models for Reduced Form Positivist purposes, just my kind of thing!