The Government wants 1.2 million new homes built in the five years to June 2029. It is not going to happen on current trends. The first five quarters delivered around 219,000 homes against a run rate that needs roughly 280,000 a year. The National Housing Supply and Affordability Council now expects the target to be met around September 2030, more than a year late, and that estimate predates the latest commodity-price shock.
So the country is behind. The instinct is to want a stronger economy to dig out of the hole. A simple regression model of building approvals suggests the opposite. To get approvals growing faster, you might need the economy to get worse first.
What the model says
I built a small quarterly model of house approvals. It is deliberately plain. Quarter-on-quarter growth in approvals is regressed on two things: its own lagged value, and the twelve-month change in the unemployment rate.
The picture that comes out is roughly half momentum and half labour market. The lagged term means approvals carry their own inertia, so a strong quarter tends to be followed by another strong one. The unemployment term does the rest of the work. The fit is modest, with an adjusted R-squared of 0.35, but for a two-variable model on noisy approvals data that is a reasonable share of the variation explained.
The chart below shows the actual series against the modelled one. The model tracks the major cycles well, including the deep COVID trough in 2020 and the sharp rebound that followed.
The counterintuitive part
Here is the bit that makes people stop. A rising unemployment rate predicts stronger approvals growth, not weaker. Read literally, that says the labour market deteriorating is good news for home building. That cannot be a real causal story, and it is not.
The unemployment rate is standing in for something else. Approvals lead the economic cycle rather than follow it, and the thing that drives them is interest rates. The RBA cuts rates precisely when unemployment is climbing, because a weakening labour market is exactly what pulls the cash rate down. So "unemployment rose over the past year" really means "the RBA has been easing." Easier money lifts approvals a few quarters later.
In other words, the unemployment change is a clean, low-noise proxy for the monetary cycle. It captures the easing impulse better than the cash rate itself does. I tested the cash rate directly, both the three-month and twelve-month change, and it performed worse. The unemployment rate wins because it smooths out the noise and lines up with the turning points.
The cruel irony
Put the two pieces together and the policy implication is uncomfortable. The model says approvals growth accelerates when the RBA is cutting, and the RBA cuts when the labour market is weakening. The fastest path to a building boom, on this logic, runs through an economic downturn.
A government that wants to hit a housing target is therefore in an awkward spot. The same government also wants low unemployment and a strong economy. The model suggests those two goals pull against each other in the short run. To get the surge in approvals that would close the gap to 1.2 million, you would want the conditions that produce rate cuts, and those conditions are a softening economy.
This is not advice. Nobody should engineer a recession to build houses, and the model is a crude two-variable description rather than a forecast of policy. The deeper point is that approvals are a monetary-cycle variable. They respond to the price of money far more than to any housing program. A target framed in number of homes is partly hostage to what the RBA does, and the RBA is reacting to the broader economy, not to the housing target.
What did not survive
For anyone interested in the model build, plenty of alternatives were tried and rejected on AIC and BIC grounds. The list includes the approvals level and its log, year-on-year approvals growth, one-month and three-month approvals growth, a monthly version with stepped quarterly GDP, GDP quarter-on-quarter growth, GDP per capita on the same basis, year-on-year versions of both, GDP per capita lagged one and two quarters, the three-month and twelve-month change in the cash rate, twelve-month population growth on a smoothed fifteen-plus base, and a lagged dependent on a year-on-year specification.
GDP in all its forms added little once the unemployment change was in the model, which fits the story. Approvals lead activity, so contemporaneous GDP has nothing to tell them. Population growth, the variable most people reach for when explaining housing demand, did not earn its place either. Over the horizons that matter for approvals, the monetary cycle dominates the demographic one.
I left out finance commitments on purpose. They would lift the fit, but they are not an external driver. Finance is a step inside the build pipeline, typically just before approvals. Being a step you take to build, finance is not meaningfully independent of the thing it would explain. In short, if I were forecasting approvals, yes, include it. If I am trying to understand why approvals occur, it is not an independent explanation.
The winner is the boring specification: momentum plus the monetary cycle, captured through the unemployment rate. It is not a forecasting tool and it will not tell you where approvals go next quarter. What it does say is that the housing target sits on top of the interest rate cycle, and that is a more fragile foundation than a five-year target number implies.
No comments:
Post a Comment