Census 2011 revealed that there were 289,451 vacant properties (14.5% of total stock) in April 2011. Of these 59,395 (3%) were classed as holiday homes. That means that 230,056 properties are vacant. In any housing market there are always some houses vacant, termed base vacancy. In Ireland the Dept of Environment (DECLG) expect 6% of properties to be vacant at any one time. In other words, given the 1.99m units they’d expect 119,691 units to be vacant. That still leaves us with an oversupply of 110,365 units (plus 17,900 under-construction units as recorded by the DECLG unfinished housing development survey for 2011).
So what constitutes vacancy and oversupply?
The unfinished housing development survey for May 2011 revealed that there were 18,638 complete but vacant units on the 2,876 unfinished estates in the country. So, unfinished estates are a part of the vacancy issue but they are only 8.1% of vacant stock nationally (excluding holiday homes) and 15.6% of oversupply. As our new mapping module that overlays unfinished estates over vacancy at Small Area level reveals, the vast bulk of 2,791 Small Areas with vacancy rates over 25% do not contain an unfinished estate. This is as true for the cities and towns as it is for rural areas (and there are quite high vacancy rates in some urban locations as the map for Dublin City below reveals where there are 7,995 vacant houses and 16,321 vacant apartments (excluding holiday homes)).
So are the vacant units derelict or underconstruction properties? No. The CSO are quite clear about this and they put a lot of effort into training enumerators how to record housing. As they state:
In identifying vacant dwellings, enumerators were instructed to look for signs that the dwelling was not occupied e.g. no furniture, no cars outside, junk mail accumulating, overgrown garden etc., and to find out from neighbours whether it was vacant or not. It was not sufficient to classify a dwelling as vacant after one or two visits. Similar precautions were also taken before classifying holiday homes. Dwellings under construction and derelict properties are not included in the count of vacant dwellings. In order to be classified as under construction, the dwelling had to be unfit for habitation because the roof, doors, windows or walls had not yet been built or installed.
In other words, units had to be fit for habitation to be recorded in the Census. And vacant units were visited several times to make sure they were unoccupied.
As to why the 211,418 houses that are not on unfinished estates are vacant it is difficult to say with certainty. It seems likely that they are combination of:
1) empty unsold one-off housing units
2) bereavement properties (a widower has died and the unit is deliberately being kept off the market as the family wait for prices to rise, or it either hasn’t been rented or sold)
3) empty investment properties (either (i) unable to find tenants or (ii) unable to sell due to negative equity, (iii) left deliberately vacant (at least initially) – many properties were left vacant in the bubble because they were bought to flip and rapid asset appreciation outstripped rental income or they were waiting for major refurbishment or demolition to make way for larger developments despite being habitable
4) the owner is in long-term nursing care or retirement home (and the house is retained to move back to or is being kept off the market – temporary vacant for health care would have been recorded in the Census as temporarily absent)
5) vacant properties due to short-term or long-term migration (where the owner has not been able to let or sell or does not want to let/sell due to negative equity)
6) units presently for sale or available for rent (a look at daft.ie and myhome.ie suggest there are over 100,000 such units in the state although the majority are presently occupied).
The base vacancy is in effect no. 6; oversupply is principally nos. 1-5 plus unfinished estates (plus the 17.9k houses in estates underconstruction, plus one-offs underconstruction). Given falling property prices, we have had five years of some of these units being built up, which would have normally been released to the market but are either being held back or cannot be rented/sold. The 110,365 oversupply would have been mostly absorbed if we had not overbuilt new housing developments in the period 2005-08 (there are 85,538 occupied units on unfinished estates).
To what extent are these vacant properties available to the housing market?
All the property in the state is owned by somebody (brand new unsold units, whether on unfinished estates or one-offs are owned by builders, developers or banks) and they are all potentially part of the housing market. Even if a property is not for sale now, they could be if the market improves. When the market does start to recover – which it will do in time – held back properties will start to be released to the market (along with occupied homes who are in the same waiting game). In other words, just because they are not available now, does not mean that will never be available and we shouldn’t start to build new stock until market conditions are sufficient to first reduce oversupply and second to entice excess stock onto the market.
Even if they were available to the market, is there a demand for them?
The 230,036 vacant units are excess to present household numbers. Even occupying 110,365 of them (the oversupply) will be a challenge. Population growth is presently low (11.4K in 2010; 13.6K in 2011 according to the CSO) and this is predominantly made up of new births who won’t be buying anything for a long time. The falling average household size (from 2.81 to 2.73 between 06-11) will create some demand, as the smaller the average size whilst population grows the more units will be needed to accommodate households. In March 2011 the DECLG revealed that there were 98,318 households on the social housing waiting list. However, 65,643 of these were in suitable housing, but they could not afford the rent and were receiving rent supplement. Moving them from their present units to oversupply units will make no difference to the vacancy rate as they’re old accommodation will become vacant on leaving. The need for additional social housing stock is up to c.33,000 (though some would be best addressed through alterations to existing accommodation). To top it all off, getting access to credit is difficult at present and impossible for some kinds of property such as apartments outside of the principle cities.
Where does this leave us?
With a very weak housing market that will remain weak for some time, especially in areas with high vacancy as supply exceeds demand. We would expect demand to rise in and around the cities first as the economy recovers (though it must be noted that population growth in the centre of cities is weak, especially Cork and Limerick where population declined between 06-11; it is more robust in their suburbs). Demand in rural areas is likely to be much weaker and it may take many years for supply and demand to harmonize.
The census provides us with a valuable insight into the total numbers of vacant dwellings, however, it does not directly tell us why units are vacant. We can speculate and use other census data (e.g. relating to migration) to draw some conclusions, but what would be really useful if some research aimed at determining specific reasons for a few case study areas.
Regardless of the reasons, vacancy and oversupply remain a serious issue. It is one of the principle factors underpinning falling property prices. The only long term solution is to harmonize supply and demand, and this has to be the main aim across the country, including in the cities. If we don’t do this, then the market will take much, much longer to recover.
Rob Kitchin
April 3, 2012 at 5:14 pm
The figure of 110,365 “excess” housing units is probably a significant underestimate, for two reasons. Firstly, the DECLG estimate that 6% of housing units can be expected to be vacant at any one time seems rather high, and one wonders where it came from. Before the 2011 census, the only previous comprehensive vacancy figure of which I am aware came from the 2006 census which (excluding holiday homes and homes temporarily vacant) found that 12.6% of dwellings were unoccupied on census night. However, this was at the height of the construction boom when the housing overhang was beginning to reach its peak.
From a review of international literature, Williams et al. (2010) reckon that a maximum vacancy rate of 5% (excluding holiday homes) is appropriate for Ireland in the context of a “normal” property market. This accords well with the situation in Northern Ireland where the unoccupied proportion of the total housing stock has hovered around 5% over the last 15 years.
Secondly, this proportion should be applied to that portion of the total housing stock which should be considered normal i.e. excess housing units should be excluded before applying the normal vacancy rate. Given the total housing stock of 1.936M (excluding holiday homes) and total vacancy level of 230,000 units reported in the 2011 census, and a “normal” vacancy rate of 5%, I have calculated arithmetically that the number of “excess” units is approximately 140,000, 30,000 more than suggested here. If we subtract this from the total housing stick of 1.936M we get a “normal” housing stock of 1.796 units. Applying the “normal” vacancy rate of 5% to this normal stock gives a “normal” vacancy level of 90,000 units which, when added to the estimated 140,000 excess units gives the total vacancy level of 230,000 reported in the census.
Reference: Williams B, Hughes B, and Redmond R (2010) Managing an unstable housing market. Urban Insitute Ireland, UCD, Working Paper No. 10/2.
Proinnsias Breathnach
April 4, 2012 at 8:16 am
The 6% rate is a generous rate. As you note, the general international norm is to expect a base vacancy rate of 3-5%, especially for urban areas and where there has been high new build rates and strong population growth (there really is no excuse for high vacancy when you have very strong household growth). It is slightly different for rural areas which might have had very low rates of growth or even negative growth in some locales where you might expect more ‘abandoned’ or obsolete properties. I have used the 6% rate because it illustrates that even against the DECLG’s generous rate there is a huge oversupply issue. I will look into the other issue.
April 3, 2012 at 7:16 pm
As someone who lives in a “small area” in rural County Longford with a vacancy rate of 21.3% (16/75), the vast majority of vacant properties there would come under category 2 (former family home – usually pre-1919 – left vacant after the older generation has passed on and either used for rental or visitor use or otherwise not on the market but not, strictly speaking, derelict). I suspect something similar might be true for a lot of the rural areas that have seen long-term population decline.
Are vacancies categorised by date of construction in the same way as occupied houses are?
April 4, 2012 at 8:18 am
The age of construction is a question filled out by the house occupier not the enumerator. Given that the house was vacant, little other information would have been collected in relation to it. I will look into it, but I suspect there is no age information for vacant stock.
April 4, 2012 at 4:43 am
Hi Rob and Justin,
I am interested in how exactly a dwelling or household is defined within the census (I have looked at the background notes but cannot find specific reference to this). For example, and assuming I am interpreting the colouring correctly, the area roughly bound by Nassau Street, Grafton Street/Harcourt Street, Hatch Street and Merrion Square in Dublin indicates a 25% vacancy rate on the Vacant Ireland map. As much of this might once have been in residential use (georgian houses/converted flats, flats above shops etc.), but has since become used for other functions, I am wondering what the rate is based on. Is it based on 25% of what has recently been known to be in residential use or does it end up being a more general pattern of vacancy based upon judgements of what appears to be a dwelling?
Philip
(Edit: I have just noticed that if I click on the area it gives exact vacancy. The above mentioned area seems to have a vacancy level of 36.5%.)
April 4, 2012 at 8:25 am
Philip, a dwelling in the Census is a unit in which people reside – not a commerical property unless it is also used as a residential dwelling. The enumerators are all well trained and will not have included dwellings that have been converted to other uses. As noted above, they have very clear guidance on determining vacancy and holiday homes. Part of that would be to first identify that the units normal use is as a dwelling. In the area you describe there clearly is still some residential units, despite conversions, and a chunk of it is vacant (perhaps falling into category 3(iii) above).
“In identifying vacant dwellings, enumerators were instructed to look for signs that the dwelling was not occupied e.g. no furniture, no cars outside, junk mail accumulating, overgrown garden etc., and to find out from neighbours whether it was vacant or not. It was not sufficient to classify a dwelling as vacant after one or two visits. Similar precautions were also taken before classifying holiday homes. Dwellings under construction and derelict properties are not included in the count of vacant dwellings. In order to be classified as under construction, the dwelling had to be unfit for habitation because the roof, doors, windows or walls had not yet been built or installed.”
April 4, 2012 at 8:37 am
Thanks for that. I would have assumed it was a more difficult process for the enumerators within a city centre area where a buildings use might be less clear-cut than somewhere like a suburban residential estate. That was why I asked about it.
Thanks again,
Philip
April 4, 2012 at 10:51 am
The policy if an enumerator is unsure is to ask other people in the building or the neighbours. And they go back multiple times. I’d be confident that the approach is robust.
April 5, 2012 at 9:14 am
Ever since this post was published I have been itching to speculate wildly about the nature and causes of vacancy. Only one thing inhibits me – the facts. There is something amiss with the data published on AIRO and on which this post is based. I live in the ED of Kileely in County Clare. According to the AIRO Census Mapping Module for the Mid West, this ED has in 2011;
247 Permanent Private Housing Units
34 Unoccupied Dwellings
28 Vacant Dwellings
281 Housing Stock (Units?)
My first problem is that I am unable to distinguish between the concept of vacant dwellings and unoccupied dwellings. Surely both definitions must be coincident but if so, why do the numbers differ?
My second problem is that the “Housing Stock” is evidently the sum of 247 and 34 and this implies that the figure 247 denotes “Occupied” housing units although this is not clearly stated.
Third and perhaps more disturbing, is the fact that in 2006, this ED had only 248 dwellings in total including vacant houses and holiday homes. The data presented above implies that either the stock of dwellings increased by 33 or in the compilation of the table, vacancies have been erroneously added to the total of all stock. Since I know that thirty houses were not built in this ED in the interval between 2006 and 2011, I am forced to conclude that the dataset is in error.
The final question must be, is this a systemic error affecting the whole dataset or does it merely affect my humble ED?
April 5, 2012 at 9:46 am
In relation to point 1 there is a clear distinction between vacant and unoccupied. Vacant is that it is not occupied as a primary residence. Unoccupied is it is usually occupied but was temporarily absent on Census night (see pt 2).
Point 2, there are several categories of recorded occupancy and vacancy as set out in Table 45 of the Census 2011 results. See http://www.airo.ie/spatial-indicators/view/2508 Housing stock is amalgam of all of them.
Point 3, we will need to check this, but I will be very surprised if I check Geodirectory and find that there are no new properties in the Kileely ED. There was new build across the whole country.
April 5, 2012 at 10:19 am
Thank you for the response.
I note that the category “Unoccupied” in fact denotes “Usual Residents Temporarily Absent”. Perhaps it might be a good idea to document this somewhere.
However the response does not define what is meant by “Permanent Private Housing Units”. I assume that it implies “Occupied Units” and that it must perforce include also “Usual Residents Temporarily Absent” but your response does not clarify this.
I indeed presume that “Housing Stock” is the total of all housing units (occupied and unoccupied) but again a clear definition would be helpful.
By the way, I too had a look at Geodirectory. According to it there were less houses in Kileely ED in 2009, than were counted by the Census-takers in 2006.
April 5, 2012 at 11:38 am
As a follow on to the points above I’ve had a look at the data on AIRO, CSO and GeoDirectory. The figures on AIRO for Perm. Private Households (PPH) (247), Unoccupied (34), Vacancy (28) and Housing Stock (281) all match up for the Killeely ED so there is no problem with the statistics there.
The PPH figure for Killeely in 2006 is 224 (196 Houses, 17 Flats/Apartments and 11 Not Stated) according to the CSO, this is the same figure as on AIRO. This would give an increase of 23 new PPHs between 2006 and 2011.
To check this I’ve had a look at the GeoDirectory for Killeely and checked the entry date. Entries to the database with a create date of 2006, 2007, 2008, 2009, 2010 and 2011 total to 37. Some of these may be vacant and a small proportion of the number with the 2006 date (18) may have been included in the 2006 Census count. Either way I’d say it easily accounts for the increase in PPH figures from 2006 to 2011.
Justin
April 6, 2012 at 5:20 pm
Hi Justin,
Your absolute certainty and faith in the data by reference to steady increases in Geodirectory gave me pause. And so I checked again.
In 2009 Geodirectory showed 234 address points in Kileely ED. Since there had been 226 occupied dwellings in the ED in the 2006 census, this seemed, in the immortal American phrase, “close enough for government work”. But then I analysed the Geodirectory address point categories. Nine of these points were defined as “Commercial” with no residential content and a further two as “Unknown” which in fact proved to be derelict houses. Thus 234 became 223 – situation improving and seeming to validate a comparison between the census and Geodirectory since the census deals only with residences.
But then the census of 2006 had said that there was a total of 248 dwellings in Kileely in 2006 of which 22 were defined by the census to be vacant. So innocently I began a search for the 22.
By an odd coincidence Kileely ED has clean Google photography dated April 2006 and some rather poorer quality photography for 2010. The recently launched myplan.ie has a more up to date copy of Geodirectory but displays it on a background of rather older OSI photography. Nevertheless, it was quickly apparent that by comparing Google’s photography to Geodirectory there had been a further nine houses in course of construction in April 2006 which could not have been occupied on census night.
But perhaps the census counted those among the vacant? This is unlikely since it was apparent from Google’s 2006 photography that some consisted of mere site works or foundations. Thus the 2009 Geodirectory total has now fallen to 214. But worse is to come. Geodirectory contains another six address points which are stated to be residences but which I can only call “test points” since they correspond to no ground features which might be even accidentally mistaken for dwellings. Perhaps they are artefacts deliberately introduced errors to detect unauthorised copying? Certainly they are not houses.
In any event, Geodirectory 2009 is now down to 208 points.
But there is some hope. There are another seven dwellings clearly visible on google’s 2006 photography for which there were no corresponding address points in 2009. So Geodirectory climbs back to 215 points. Using myplan.ie as a reference, and comparing with Google photography from 2010, we can find a further four dwellings that must have been completed after 2006. So we arrive at a figure of 219 dwellings .
Now here is where I have trouble suspending disbelief. According to the census of 2011 published on AIRO, the total number of houses, occupied and vacant, has increased to 281. That is thirty-three more than the grand total of occupied and vacant in 2006 but it is perfectly clear from Google’s photography that that there are not thirty-three further new houses in Kileely ED. It is also clear that some of the points which you report as having been added to Geodirectory between 2006 and 2011 are simply correcting omissions although a few are genuine new additions giving the impression of steady growth. But none of these amendments will bring the grand total to 248 (2006) nor yet to 281 (2011).
It is rather difficult to conceal thirty-three houses behind a furze bush so perhaps you can now understand why I am somewhat skeptical of the data published.
One reasonable conclusion I believe that I can validly draw from this examination is that comparisons between Geodirectory and the census are not safe and the one should not be used to ground-truth the other. Indeed the CSO itself concluded this in its field tests for the 2011 census.
I began this examination because the data supplied by AIRO to the Mid West Region will be used for policy development but unless the limitations of these data are well understood, erroneous policy proposals may result.
The only place that you might hide thirty-three houses if not behind a furze bush, is in plain sight. An hotel, now closed but surrounded by what may have been self catering apartments. If these apartments were let for longer rather than shorter periods then the available housing stock could overnight expand and contract by the required tens of units. If the hotel is converted to a nursing home (as envisaged by a recent planning permission) then the units will become bona-fide long term residential accommodation.
Policies designed to limit “one-off” houses for example would be misdirected when the problem of unexpected growth is the result of conversion of a commercial premises.
The census may be a fiction in which we all politely profess varying degrees of belief, but we should beware the unexpected effects of “mashing” it together with other data like “pandy”.
Dick
April 6, 2012 at 5:37 pm
We’ll take a look at this after Easter, but just to respond to this statement:
“One reasonable conclusion I believe that I can validly draw from this examination is that comparisons between Geodirectory and the census are not safe and the one should not be used to ground-truth the other.”
My understanding is that the CSO and Geodirectory worked very closely during Census 2011 and every single property in the Census is in Geodirectory and carries a Geodirectory georeference. In other words, Geodirectory was used to identified properties that enumerators would visit and enumerators on the ground updated any gaps in Geodirectory – they are ground-truthed against each other. I’m not saying that each is absolutely 100% perfect, but in my view they are pretty close (much closer than you’re suggesting) and we can plan policy very robustly on the basis of them.
April 6, 2012 at 5:59 pm
Some Quotes from the 2009 pilot test;
… With a view to assessing the impact of the GeoDirectory on the Census field operation, the Pilot tested the integration of the GeoDirectory into the methodology of Census geography.
… The GeoDirectory provides a flag indicating whether a building is under construction, vacant, a holiday home or derelict. This flag is based on the date it was last validated by An Post staff member. The results of the Pilot suggest that these flags cannot be taken as a reliable guide to the current status of these buildings. This pattern is more pronounced in rural areas.
… Approximately 4% of dwellings enumerated during the Pilot were not listed on the GeoDirectory. However, a greater number of dwellings not on the GeoDirectory were found by enumerators using blank ERBs (4%) compared with those using pre‐addressed ERBs (3%).
April 6, 2012 at 9:06 pm
You cannot take the issues reached in the pilot to apply to the full census; that’s why it’s a pilot! That’s why the CSO do pilot studies on all their data generation – to look into all this stuff and then take appropriate actions – like working with Geodirectory between pilot and Census to make sure that any problems are addressed. Remember, enumerators went to every building to assess it (several times if it was vacant) – they did not do it simply from geodirectory flags or looking at google satellite images. My understanding is If they found a dwelling not in Geodirectory it was verified and added. As I’ve already said, ALL survey’s have a margin of error, but in this case I believe it to be a very small margin that enables robust policy formulation. And I’ll take a look at your specific example after the Easter break.
April 7, 2012 at 12:35 am
Happy Easter!
So what else does a sample mean?
Dick
February 19, 2013 at 9:48 am
[…] with, there is a strong correlation with the presence of unfinished estates. However, as we have discussed elsewhere, unfinished estates are just one element of vacancy given that there are only 16,881 vacant […]