Massive 1.8 Percentage Point Spread Between GDPNow and Nowcast: What's Going On?
You can almost drive a truck between those estimates.
- In many past quarters, the GDPNow estimate started high then steadily plunged.
- This quarter, the GDPNow forecast started high and the spread widened
GDPNOw Factors by Date
The spread first widened on May 31. We can compare to Nowcast factors.
Nowcast Factors by Date
Construction Comparison June 1 (Yellow Highlights)
- Nowcast: +0.08
- GDPNow: +0.10
The construction report, considered "hard" data, played no part in the discrepancy. The difference is a small rounding error only.
PCE, Trade, Jobs, ISM Comparison May 31 (Yellow Highlights)
- Nowcast: +0.25
- GDPNow: +0.60
I do not have a precise breakdown of GDPNow. But in the past, ISM has goosed GDPNow substantially. ISM is considered "soft data". Nowcast's three ISM categories added only 0.04 to the Nowcast estimate. The three Nowcast employment categories (counting ADP - the last line), added nothing.
Housing Starts and Permits June 22 (Red Highlights)
- Nowcast: -0.10
- GDPNow: +0.10
GDPNow and Nowcast reacted in opposite fashion to the same data. Unlike May 31, there is no confusion as to the origin of the discrepancy.
- The initial GDPNow forecast was 0.9 percentage points higher (4.1 GDPNow vs 3.2 Nowcast).
- GDPNow blasted higher on a mix of both hard and soft data
- Nowcast did not react as positively to ISM as did GDPNow (my estimate, not a certainty)
- Nowcast and GDPnow acted in opposite fashion to the construction report
The "what" is easy to explain the "why" is a guess.
- I maintain that GDPNOw overreacts to ISM. We have seen that behavior before and I have written about it.
- Starting this quarter, Pat Higgins changed the GDPNow model to reduce volatility of reports at the beginning of each month. I believe that may be a factor.
I wrote about point number two on April 30, the initial GDPNow forecast for the quarter: Here We Go Again: GDPNow Projects 4.1% GDP
Pat Higgins Model Change
Today's update incorporates modifications to GDPNow's dynamic factor model as described here. [Mish Note: the dynamic factor model is for economic geeks only].
The dynamic factor is used to forecast yet-to-be released monthly GDP source data. The modifications to the factor model are intended to partially dampen some of the volatility in GDPNow forecasts in the early part of each month.
Model Change Go Haywire?
I do not know.
Was the rise due to volatility suppression? That's essentially the same question. Again, we cannot say for sure.
All we can say for sure at this point is the Nowcast model expected strong reports. The economic reports were not strong enough to send the NOWcast estimates higher.
In contrast, the data exceeded the GDPNow model forecast, sending those estimates higher.
The latest "Blue Chip" (paywall) monthly estimate according to GDPNow is 3.5% as of June 6 or so. Blue Chip is nothing more than a survey of 50+ economists' expectations.
The spread is now so wide that both GDPNow and Nowcast may easily miss their alleged margins of error.
Mike "Mish" Shedlock