Housing statistics - some challenges
Posted on February 18, 2017 by Tom Healy
A famous 19th century Irish mathematician and engineer, Lord Kelvin, famously once said: “ I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind; it may be the beginning of knowledge, but you have scarcely in your thoughts advanced to the state of Science, whatever the matter may be ." [PLA, vol. 1, "Electrical Units of Measurement", 1883-05-03 cited here ]
In a previous Monday Blog, “ Housing: no surprises ”, I reported an estimate of new house building in the region of 15,000 in 2016. In fact, the situation is probably far worse than might have appeared from the chart shown in that blog (the title of which has been changed and is reproduced in Chart 1 here). Based, as it was, on annualised ‘House Completions’ (which is the actual term used by the Department of Housing, Planning, Community and Local Government, DOHPCLG, on their statistics page here and accessible in Table 9 here ) the statistics may be a poor guide to the actual number of new houses built and completed in a given year. I was aware that all data were based on ESB connection for ‘new dwellings’. The exact note used on the Department’s website is as follows:
House completions data series are based on the number of new dwellings connected by ESB Networks to the electricity supply and may not accord precisely with local authority boundaries. These represent the number of homes completed and available, and do not reflect any work-in progress.
Now, it would be reasonable to regard the number of ESB connections to ‘new dwellings’ as a good proxy for new housing output over a number of years though any backlog of newly built houses or sudden changes in housing output can lead to lags in the data with catch-up happening in particular years. In other words, the numbers would even out so that connections are a reasonable estimate of new builds. However, there are two significant issues one of which came to light only very recently:
1 Timing factors
The Irish housing market has undergone a dramatic change in recent years with an output (as proxied by ESB connections) of close to 93,000 in 2007 and falling to 8,000 in 2013). The problem in the immediate aftermath of the property crash and implosion of construction activity in 2008-2013 is that a large number of unused dwellings at varying stages of completion and readiness for use were lying in fields and estates across the country. It is plausible to assume that actual new dwellings built in recent years is well below the numbers of ‘house completions’ as measured by electricity connections. Some commentators claim that the true number of builds might be a half of house completions in recent years due to the gradual catching up and release of new dwellings previously built but not completed or occupied.
The expression ‘ the number of new dwellings connected by ESB Networks to the electricity supply’ used by DOHPCLG while not technically incorrect could be misunderstood (or indeed misrepresented). In a statement issued by DOHPCLG last week here it has been clarified and confirmed that ‘new dwellings’ do not mean what one might have thought it meant. In some cases, dwellings which were left unoccupied for at least two years and were reconnected by ESB are counted as ‘new dwellings’. Neither ESB, DOHPCLG or anyone else seems to know just how many such ‘new dwellings’ are contained within ‘completions’.
It is important to distinguish timing factors from double-counting. I suspect that the element of double-counting is limited (possibly less than 10%) while the timing factor is much more significant. The implication is that the total stock of housing is increasing much more slowly than might be the impression given by relying on ‘new completion’ data. DOHPCLG in their stock estimates here confirm a much slower growth year by year between 2011 and 2016.
The matter featured in a recent item by ‘Cantillon’ in the Irish Times (“ Government playing fast and loose with housing statistics ”) where the views of some experts have been cited in support of the claim that the true rate of new house building is running at about 50% of the house completion rate. In the same edition, the DOHPCLG countered these claims (“ Department denies housing statistics unreliable ”). The worrying claim is cited that ‘new electricity connections can be triggered by work to existing buildings, or by formerly vacant units coming back on stream’. In other words, ESB connections do not only refer to ‘new dwellings’ in the sense that most people understand this term (those built in recent years and never occupied) but also dwellings previously occupied but brought back into use after two years. DOHPCLG clarified the matter last week as follows:
ESB connections are used a proxy for house completions as it is the best available indicator that a house has been completed and is now ready for habitation. This data set is available on the Department’s website dating back to 1970 thus providing a longitudinal comparator. ESB Networks has recently confirmed to the Department that the data does not include what it calls “service alterations” where for example an extension is carried out on a dwelling and alterations are therefore required. It does however include re-connections where a dwelling has been vacant for a period of at least two years, and as such all connections represent a residential unit which is now available for occupation.
In the year to end December 2016 an overall total of 14,932 connections have taken place. This represents an increase of 18% on 2015. About 42% of these connections relate to once-off units and these do not tend to show up on the Central Statistics Office Residential Property Price Index volumetric statistics as there is no transaction attracting stamp duty associated with these connections.
It is not possible to tell from the ESB connections dataset at this time which dwellings were never before occupied, and it should be acknowledged that residential units are also permanently disconnected from the grid at a rate of about 0.2-0.5% of the housing stock per annum.
The reference to ‘ESB Networks has recently confirmed to the Department that “ the data does not include what is calls ‘service alterations’ where for example an extension is carried out on a dwelling and alterations are therefore required” suggests that DOHPCLG may not have been fully aware of this fact in recent times. Certainly the footnotes to the relevant Table(s) makes no reference to this important fact. (The data series based on ESB connections runs for decades and has stood the test of time).
The basis for the claim that there were only approximately 8,000 new builds in 2016 are statistics on stamp duty transactions, once-off house commencements and local authority builds.
Why does any of this matter if building output and electricity connections are aligned in the ‘medium-term’? It matters for at least the following reasons:
- The actual number of new builds is even lower than most people have assumed;
- We do not know the number of previously built dwellings being brought into use in any given year; and
- It is very difficult to assess exactly how far short Government is falling compared to the medium annual target of 25,000 contained in Rebuilding Ireland (but it is sure that the current housing output is at least half and possibly one third of what it should be given projected demand)
The shortfall in housing output relative to demand represents a disastrous situation especially in relation to social housing output. A figure of 27,000 new units a year has been used recently by the ESRI as a measure of average annual demand. However, this number may very well be an under-estimate due to higher than expected net inward migration and emerging patterns of household formation.
The problem, in a nutshell, is that we do not know how many new houses are being added to stock in any given year. We have some of total stock from the Census of Population data every 5 years as well as annual estimates published by DOHPCLG here . However, these stock numbers are based on very imperfect estimates of new completions and derived estimates for obsolescence. What we know from the 2011 and 2016 censuses is that total stock increased by 18,981 between April 2011 and April 2016 (the latter figure is provisional). In other words, the net increase in stock was just below 4,000 per annum on average over the entire five year period from 2011 to 2016. Based on DOHPCLG data on ‘house completions’ which is just over 10,000 per annum on average over the same five-year period, the net additional figure is probably somewhere in the region of 40% of the estimated new completions. Assuming an obsolescence rate, based on historical trends, of around 4,000 per annum this leaves a gap of very roughly 1,000-2,000 per annum which may be accounted for by the ‘double-counting’ phenomenon mentioned above. All of these estimates are very rough. All in all, the general public and researchers could be better served by the following:
More comprehensive data on housing flows – new builds, completions and estimated net additions to total national stock as well as breakdowns of these into private, voluntary and local authority components.
To paraphrase Kelvin, how can we know what we are dealing with and how well we are addressing the matter if we do not know the most rudimentary facts of supply never mind demand?