Email Exchange With GDPNow Creator Pat Higgins on the Model
GDPNow vs Nowcast Evolution
Compared to the Blue Chip forecast these models look ridiculously optimistic.
The GDPNow website offered this tidbit.
There are no subjective adjustments made to GDPNow—the estimate is based solely on the mathematical results of the model.
In particular, it does not capture the impact of COVID-19 beyond its impact on GDP source data and relevant economic reports that have already been released. It does not anticipate the impact of COVID-19 on forthcoming economic reports beyond the standard internal dynamics of the model.
Stranger things have happened but I pinged Pat Higgins specifically about unemployment claims as I thought that may be the source of the error.
Here is the response from Higgins who is always generous with his time.
The model does use initial unemployment insurance claims among the 126 series used to estimate the monthly factor used to forecast yet-to-be released data.
However, the spike in March claims is identified by the code as an outlier and replaced with a more typical value.
Our policy is to not change the code in the middle of a quarter, so we can’t change how the code handles outliers until we begin forecasting the second quarter.
Whether or not March claims is treated by the code as an outlier does not have an outsize impact on the first quarter forecast, in part because it is only 1 of 126 series and because much of the first-quarter data has been released.
I can’t really say how a spike in unemployment would effect the forecast because of the FOMC blackout which ends a week from Friday. I can follow up after that however.