Notícias FX

Eventos que mexem com o mercado, condições de cada sessão e cenário macro. Escrito para quem opera com sistema, não com palpite.

Acompanhe pares marcando-os com ☆ na página de sinais.
Tag: building permits Limpar filtro

Ontem · 1 notas

The Field With No Error Bar Missed by Eighty Thousand. The Field With a 400,000-Unit Error Bar Beat by Fifty-Nine — and the Beat Is Not Significant.

US October housing starts printed 1,309,000 at an annual rate on 18 November, beating all four published consensus values — 1.23, 1.245, 1.25 and 1.27 million — by between 39,000 and 79,000 units.

Then the Census Bureau undercut the beat on its own line: the 9.9 per cent monthly rise is stated as plus or minus 16.8 per cent, with a footnote saying the 90 per cent confidence interval includes zero, so the change is not statistically significant on the issuer’s own test. Measured in levels that interval is 400,180 units wide against a four-publisher consensus spread of 40,000 — a ratio of 10.004, which is the cleanest direct confirmation this desk has had that vendor dispersion is not an uncertainty estimate. A release blackout sized by consensus disagreement is sized by a number ten times too small. Two further findings. One vendor’s consensus column read 1.27 million when this desk recorded it before the release and 1.25 million after it, while that page’s own separate model forecast stayed at 1.20 — so the surprise you compute depends on when you read the consensus, and the same print scores as a beat of 39,000 or of 59,000. And the September prior was revised away inside the same document: every consensus was built against 1,221,000 and the release replaced it with 1,191,000, which makes the 9.9 per cent rise 7.207 per cent against the base the forecasters actually used. Permits are the inverse case. Three publishers agreed to within 5,000 at 1.375, 1.377 and 1.38 million; the print was 1,295,000, missing all three by about 82,000, or 1.281 times the starts beat — and that series carries no margin of error at all, because the issuer says it is not subject to sampling error. The absence of an error bar is not a tighter error bar.

Opere a reação, não a manchete
A notícia mostra quando as condições mudaram. Combine com o calendário econômico e confira os sinais atuais antes de mexer em qualquer coisa.