Are home builders more confident than they should be? There has been some discussion suggesting that the relationship between the National Association of Home Builders/Wells Fargo Housing Market Index (HMI) and single family housing starts is broken, or at least strained lately. Let’s take a look.
The HMI and starts track together well, no question. But they don’t move in lockstep, that’s clear throughout their history. So how much and when should they differ?
The gap between the HMI and starts appears widest at turning points and throughout the most recent boom/bust/recovery cycle. But absent a systematic method of comparison conclusions are in the eye of the beholder.
One way to think about the relationship is to first examine how each series moves over time relative to its own long term average. This normalization is a standard method to compare the dynamics of economic (or non-economic) variables that may vary greatly in scale. Converting the data to deviations from the mean allows a more meaningful comparison. The chart below shows the HMI (red) and starts (blue) normalized by dividing each month by the series average (1985-current).
Since these deviations from the means are just dividing by a constant, you get patterns very similar to the original levels. The green line is HMI deviations minus starts deviations.
Prior to 2005 the deviations track most of the time but the HMI tends to overstate deviations in starts at turning points (91, 94, 95, 98-01 was a long turning point). During the most recent boom, bust and recovery the HMI over-overstates the downturn and similarly over-overstates the recovery. Or, the magnitude of overstating is in proportion to the change in starts (not 1 to 1, but proportional). Given that the recent cycle was more extreme in both its rise and fall, we would expect the divergence between the HMI and starts to be greatest during this period. This proportionality is the point that seems to be missed in the arguments that the HMI is “getting ahead” of starts.
The HMI isn’t broken during the divergences. Starts and builder sentiment track but don’t move in lockstep. It’s just that changes in builder sentiment occur at a higher octane than starts at turning points.
The HMI is currently telling us what it has told us in the past: after falling faster than starts during the downturn it is now rising faster in the recovery. The important point is that the HMI isn’t broken or misleading getting ahead of growth in starts, that’s what it always does.