Ten years ago, Ireland's ghost estates epitomised the bursting of the property bubble - the obvious symptom of over-building. Today, the homelessness crisis is the obvious symptom of the opposite failure in the housing market - under-supply.
Why don't we build the right number of homes in the right places? Poor data is a big issue. Without reliable data, it is impossible to know how much housing we need.
It can be tricky to keep a tally of how many homes are built each year. Knowing how many old houses are lost from the stock of housing - so-called obsolescence - is even harder.
Like many people, I believe that the resolution of Ireland's current housing problems ultimately requires more supply. However, the challenges of getting supply right are compounded by the reality that some properties get knocked down every year and some of them must be replaced.
Unfortunately, industry commentators disagree about the extent of these housing demolitions. Recent estimates range from 6,000 to 16,000 units per year.
However, whatever the correct number, the assumption that all of these units need to be replaced is illogical. If we believe it, we will create housing completion targets greater than the actual requirement - and maybe even new ghost estates.
While it is not always clear how commentators are deriving their housing obsolescence estimates, economist John FitzGerald proposes a practical and transparent method. FitzGerald begins by adding up the number of housing units built between two census years. The extent to which this number exceeds growth in the housing stock over the same period must represent the number of units that went obsolete.
Applying this approach to the latest data, we find that 29,500 new homes were built between April 2011 and April 2016. However, census records show that the housing stock only rose by 8,800. This suggests 20,700 units (4,140 per annum) must have become obsolete - an annual depreciation rate of 0.21pc.
But this is only the tip of the iceberg. Because, in addition to the fact that fewer dwellings are being demolished, we can challenge the logic that they all need to be replaced.
Housing obsolescence varies dramatically by region. As illustrated by the green areas of the map, Dublin, Cork, Galway and their surrounding commuter belts have very low depreciation rates - very few homes in these areas are abandoned or knocked down.
In contrast, as shown by the red areas, obsolescence is much higher in the rural West of Ireland. Again, one might infer that this reflects slower (and in some places falling) population growth. This results in weaker housing demand, higher vacancy rates and lower capital values, which may not justify the expense of maintaining superfluous properties.
Counties with stronger population growth tend to have fewer demolitions.
The counties of Connacht tend towards sluggish population growth and high depreciation of the housing stock.
Munster is similar to Connacht, but less extreme.
Obvious exceptions to this pattern are Cork and Galway, where urban centres act as magnets for population growth, leading to preservation of the much-needed housing stock.
Ultimately, there are two conclusions. Firstly, demolition is much less of an issue in terms of supply than many commentators are claiming.
At most, it can be no more than 4,100 units per annum. But it is also clear from this data that many properties that are being knocked down do not need to be replaced.
They are in places where demand for housing is low.
By my calculations, fewer than 3,000 new dwellings need to be built each year to cover obsolescence. This is really good news, because it means that - relative to the most extreme estimates - we are already 13,000 units closer to meeting our annual home- building target.
Secondly, this analysis highlights the futility of building too much in the wrong place. In this context, it is very positive that residential development is now happening where it is most needed: 39pc of dwelling completions and one-third of planning permissions in the past year have been in Dublin.
However, we must remember that it is not just about the number of units, but also the type of units. That, perhaps, is an article for another day.
Dr John McCartney is director of research at Savills and is on the editorial board of the 'Journal of Property Research'