US housing starts fell 2.6% in August but permits are still above last year. Does housing data belong in your USD analysis?

The Census Bureau said on 17 September that housing starts fell 2.6% in August to a seasonally adjusted annual rate of 1,275,000, while building permits slipped 2.7% to 1,394,000, which is still 3.5% above a year earlier. Single-family starts rose 7.6% to 918,000, and completions fell 11.9% (full report).

Two things stand out. The Bureau's own confidence ranges show that the monthly changes in starts are within the margin of error, so they can't be called significant, and it says it may take three months to see a trend in permits and six months for starts. Around the same time, builder confidence fell to 32 with 38% of builders cutting prices (report), and existing-home sales fell 2.0% (report).

Housing is one of the parts of the economy most sensitive to interest rates, and the Federal Reserve raised its target range to 3.75%–4.00% on 16 September (report). But it rarely moves the dollar on the day.

Questions for the room:

• Do you use housing data at all in your dollar analysis, or do you skip it?
• How do you treat a number that comes with a wide margin of error: ignore it, average several months, or watch the trend in permits?
• Which US releases do you rank above housing, and why?
• Have you ever changed a view because of housing data?

Please share your own approach, not predictions sold as certainties. Signal selling and promises of direction are removed.