Mission: Unnecessary
Household formation in England looks set to run at about 150,000 per year up to 2031, well below current official projections.
The prime minister has described it as her ‘personal mission’ to solve the housing crisis by building more. And according to the chancellor, we need 300,000 new houses per year. To many this sounds like Mission Impossible; in fact it’s Mission Unnecessary.
My last blog told the story of why the official projections — the bedrock of the housing need figures — keep overestimating the rate of household formation. In this blog I recalculate the projections to try to correct the source of error. The results suggest that housing need is well below what the government thinks.
There is nothing obviously wrong with the broad approach DCLG takes to projecting household formation. The problems lie in the inputs used. To correct those, two modifications are in order.
Modification 1 — use the latest ONS population projections
The first change I make here is to feed in ONS’s latest population forecasts, which came out last October. DCLG’s current household projections rely on ONS population projections from 2014. But the recent figures from the ONS suggest that there will be one million fewer people in England in 2031 than previously expected. (Incidentally, the primary drivers of the change are half-a-million fewer births and almost 400,000 more deaths than expected.) Naturally, all this takes a chunk out of housing need.
Modification 2 — adopt the ONS proposal on household size trends
DCLG currently bases its projections on household size trends from 1971 to 2011. As I outlined in the last blog, for much of that period households were shrinking. Since 2001 that trend has stalled, but DCLG implicitly assumes it will resume — the source of repeated over-estimates.
The ONS will take over responsibility for producing household forecasts later this year. In its consultation on the methodology, the ONS proposed to base its projections on household size trends only since 2001. While the ONS is yet to announce what it will actually do, this seems like a sensible way to address the past problems.
Re-estimating the household size projections
In re-estimating the most recent DCLG household projections to account for these two changes, I use a simplified version of their methodology, basing it on household size trends by age band.* Average household size by age is reasonably constant over time in this new projection. But older people are more likely to live in one-person households. With increasing numbers of people in the older age bands as the population ages, average household size resumes its decline in my projection, albeit at a slower rate than current official projections. The chart below shows the resulting projection (green line) grafted onto the 2017 ONS estimate for current household size.**
Household formation 30% lower
Combining the revised household size projections with the new population figures has a big impact on the outlook for household formation. By my calculation, adopting the ONS’s suggestion on household size trends cuts the rate of household formation by around 40,000 per year up to 2031. Meanwhile the lower population projection cuts household growth by a further 27,000 per year over the same period. The result is annual household formation running at around 151,000, rather than the current official figure of 218,000 — a 30% drop. This compares to an average household formation rate of about 140,000 per year over the past 20 years.
What does this mean for how many houses we need? It’s reasonable to think of housing ‘need’ as exceeding the rate of household formation. Alan Holmans, for example, suggested that in order to maintain the proportion of vacant and second homes, a extra 19,000 houses would be required each year. Adopting Holmans’ estimate, my projection implies a total housing need of around 170,000 per year up to 2031.
When the ONS decides on its approach later this year, the new projection may look something like this. Or they may make other changes so it looks higher or lower. Either way, tweaking the current approach in the ways they have indicated yields a more plausible, but significantly lower, projection than policy is currently based on.
And there’s nothing wrong with erring on the side of caution and shooting higher with the targets. Hell, let’s throw in another 30,000 houses per year in case household formation turns out higher. But let’s not panic that we ‘need’ any more than 200,000 per year to accommodate new households and keep housing costs under control.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
*I have used a simplified version of the method documented by DCLG. I use DCLG’s ‘headship rate’ data, replacing the post-2011 trend given with a continuation of the trend from 2001–11. In keeping with the DCLG methodology, these headship rates are then multiplied by projections of the household population. This is derived from the ONS’s latest population projections by age group, less DCLG’s estimates of the number of people living in institutions. This process is completed for headship rates and household population by age band. Building it up in this way, rather than simply projecting forwards the average household size trend, allows the projection to capture the fact that, even if headship rates by age are constant, average household size may still fall if the population ages, since older people are more likely to live alone. In its more granular methodology DCLG breaks down trends by age, sex and marital status. For these purposes, however, the age-based break-down is likely (in my view) to be the most important one because the ageing population is probably the single most important determinant of the average household size trend.
**This projection is probably conservative because it assumes that the recent up-tick in household size, seen in the ONS’s surveys, immediately goes into reverse from next year, and a downward trend resumes. We might doubt that, since the recent drift upwards appears to have been due to a bigger migrant population, and the LFS suggests that migrants tend to live in larger households. Consequently this projection may continue to over-estimate household formation.
