Two things are being reported about the rental market and they do not sound like they belong to the same year.
Asking rents have gone almost nowhere for three years and fell slightly in some quarters. Completions are running down toward a low point later this decade. And the share of renters spending more than a third of their income on housing is close to the highest it has been recorded.
The ordinary response to a pair like that is to pick the one that matches what you already believe and treat the other as somebody's spin.
Johannes Kepler spent six years refusing to do that, on a discrepancy of eight arcminutes.
The leftover is not the nuisance
He had a circular model of Mars that fitted Tycho Brahe's observations almost perfectly. Almost. The error that would not go away was about eight arcminutes — a quantity so small that every astronomer before him would have absorbed it into the general untidiness of observation.
He did not, and his reason is the whole method. Tycho's instruments were good to about two arcminutes. So the leftover was four times larger than the known limit of the measurement, which meant it could not be the measurement. It had to be the model. Six years later the circle was gone and the orbit was an ellipse.
The discipline is two numbers rather than one. Not just how big is the discrepancy, but how big is the error the instrument can actually produce. A residual is only a finding when it is larger than the instrument.
Two instruments, two populations
Run that on the housing pair and the contradiction stops being one.
The rent number that is flat is, in most reporting, asking rents on professionally managed apartments: what a new tenant is quoted on a vacant unit in a building large enough to be tracked. That is a real, carefully constructed quantity with a narrow aperture. It does not observe what a sitting tenant is charged at renewal. It does not observe the small landlord with four units. It does not observe the bottom of the stock at all, which is precisely the part that has been shrinking as older cheap units are renovated, converted or demolished.
The affordability number is a share of households, weighted by people rather than by units, and dominated by the lower half of the income distribution — the half least likely to be living in the buildings the first number tracks.
So the gap between them is not noise and it is not spin. It is the two instruments pointing at different populations, and the size of the gap is a measurement of how far apart those populations have drifted. That is a finding about the market rather than a problem with the reporting, and it is invisible to anyone who resolves the contradiction by choosing a side.
Watch out for
This method has a bad failure and Kepler ran into all of it.
Pointed at an instrument that is not good enough, it manufactures theories out of ordinary error. He spent years chasing residuals through hypotheses about atmospheric refraction and never arrived at the law that Snell found shortly afterward. If you do not know the precision of the thing you are reading — and most published figures do not state it — a leftover is not evidence of anything.
The worse one is closer to home. In the chapters where he clinches the ellipse, he presented numbers computed from his own theory as though they were independent confirmation of it, and nobody noticed for three and a half centuries. The person most able to fool you with a residual is the one who built the model, and that includes you. When a gap between two figures happens to support the conclusion you already held, that is the moment to establish who constructed each series and what they were constructed for.
He was also, by his own account, a poor observer and a hasty calculator working entirely from another man's nights. The method depends completely on somebody else having measured well.
Answer this next
Take the housing number you last quoted and answer one question about it: which population does it cover, and which does it exclude?
If the answer is that it covers the units that are easy to count, you have not been reading about the market. You have been reading about the part of it that is instrumented.
