The gap between the two nowcasts is the signal, not the noise. Atlanta at 4.7 and New York at 2.3 for the same quarter isn't one model being wrong, it's a methodology tell: GDPNow mechanically extrapolates each incoming series, so it runs hot early and volatile, while the NY smoother is Bayesian and shrinks toward trend. Stacking the Dallas WEI at ~3% year-over-year alongside them effectively triangulates the true read into the low-to-mid 3s.
Which is the useful discipline here, reading three estimates as a distribution rather than picking the friendliest headline. The spread between them tells you how much of the quarter is still being guessed.
The gap between the two nowcasts is the signal, not the noise. Atlanta at 4.7 and New York at 2.3 for the same quarter isn't one model being wrong, it's a methodology tell: GDPNow mechanically extrapolates each incoming series, so it runs hot early and volatile, while the NY smoother is Bayesian and shrinks toward trend. Stacking the Dallas WEI at ~3% year-over-year alongside them effectively triangulates the true read into the low-to-mid 3s.
Which is the useful discipline here, reading three estimates as a distribution rather than picking the friendliest headline. The spread between them tells you how much of the quarter is still being guessed.
Indeed! That's about it!