Data


Eoin O’Mahony, UCD and TCD.

I have been working for a while now with the data produced by the InsideAirBnB project. I teach students how to map and analyse these kinds of datasets when they are learning to use geographic information software. The data are really useful to understand how the city changes, how urban unevenness plays out and what can be done to undermine the ‘sharing economy’. That last phrase in particular, the sharing economy, is very pernicious. Sharing usually involves me giving you something and, maybe, you giving me something. In the case of AirBnB, money is given over for a space to sleep and eat.  That doesn’t sound like sharing to me but old fashioned marketised social relations. The same goes for the gig economy: the last time I went to a gig, I wasn’t asked up on stage to pound out a few tunes with The Unthanks.

This morning, I read that Dublin City Council have finally published their report on the impact that AirBnB is having on Dublin City’s housing. One of the more significant reported findings is that there are many individual people renting out multiple short lets. Downey’s report for the Council (which I have yet to read) recommends that two Council committees work together to figure out a way to “tackle the issue”. While we await the Council’s prognostications, let’s examine some of the impacts that the most recent batch of data (February 2017) points to. This is a kind of geography of AirBnB in Dublin, a way in which to help analyse the current housing crisis. This is the housing crisis that Coveney would like to solve part of before June, you know, after winning the leadership race of his party. Priorities, right?

Firstly, within the City Council area, there has been an increase in the number of listings between August last year and the February scrape. In August, there were 4,931 listings for the city area – the vast bulk of all Dublin region listings. By February, this had increased to 6,729, an increase of 36%. There must be few other things in the city that have increased by this amount in this period, except perhaps seagull droppings.  There has not been a 36% increase in the output of social and affordable homes in the city over this period. There is clearly a number of people out there who have apartments in the city who know that if they rent the spare room or the whole apartment they can make some money. Short-term lettings like these allow people the flexibility to rent some weekends and not others but also to pay a mortgage on a second (or fifth or eighth) rental property they just happen to have lying around. It beats having long term tenants it would seem. Perhaps significantly, the proportion of listings that rents the whole property out (as opposed to a room) has remained stable at 47% of all listings.  So where are these listings located?

One of the really good features of a geographic information system (software that allows for spatial analysis) is to be able to see patterns across the city. I conducted a point-in-polygon analysis of the data from the February 2017 listings dataset. As the name implies, this counts the number of listings within each predefined area, in this case electoral divisions (EDs). There are 162 EDs in the DCC area. Location information for these listings are anonymized by Airbnb so any scraping process encounters the following spatial constraints:

    • the location for a listing on the map, or in the data will be up to 150 metres from the actual address.
    • listings in the same building are anonymized by Airbnb individually, and therefore appear “scattered” in the area surrounding the actual address.

I would be interested to see how Downey may have compensated for this in his report for the Council. Any point-in-polygon analysis is therefore compromised by these two constraints. Knowing this, what spatial patterns can we see? The average number of listings per ED is about 34. In the first map below we can see the distribution of listings below, around (±10), and above the average.

Edit: dynamic map is now available here.

number of listings per ED

 

The parts of the city that have above average listings include the docklands, the north inner city around Mountjoy Square and near Stoneybatter. By the far the largest concentrations of listings are seen south of the river, particularly in the south docklands and around Temple Bar. Focusing on those EDs with 100 or more listings, it is clear that the areas south of the river have many more listings than those north of it. This may point to a greater availability in these areas.

EDs with 100 or more

 

Interestingly, the gap between in the southside of the map above contains the areas fancifully known as ‘the Georgian core’. The sabre-shaped ED known as South Dock has well over 300 listings. This takes in an area including the south docklands as well as the area immediately to the south and east of Trinity College. In and around the City Council building on Wood Quay is an area of high concentrations. Thanks to a suggestion by Martin at NCG, I then normalised these listings data by the number of housing units per ED from the 2011 Census. This gave a slightly different geography to the listings data. The average per area is a little under 2% of all housing units. Again, I classified the normalised listings data by below, around and above average but have not displayed the below average areas. We can note a number of differences, as is clear from the final map below.

as a percentage of

 

13% of the units in south inner city are listed as AirBnB-available units. About 9% of the units South Dock are. The Georgian core comes back into play. The heaviest concentrations of listings are therefore found in the south inner city, heading west. I would like to read Downey’s report on this before I do any more work on these data. What’s not clear to me of course is if the Council is going to take any concrete actions to at least curb the power of property to yield profits in the middle of the city’s worst housing crisis.  As Lorcan Sirr has indicated recently, some in control of this city have a strange relationship of denial with data. Action would require the Councillors to push back against the primacy of private property so you know…..not much will happen unless we organise like they’ve done in Barcelona and elsewhere.

Advertisements

The prelim results for Census 2016 make it clear that housing vacancy continues to be a serious issue in Ireland.  Given that new housing units grew by only 18,981 to 2,022,895 and population grew by 169,724 to 4,757,976m between 2011 and 2016, one might have expected vacancy to have fallen substantially.  However, housing unit vacancy fell by only 29,889 to 259,562.  Of these 61,204 are holiday homes (HHs), a slight growth of 1,809 from 2011.  On a base level vacancy of 6%, oversupply is 76,984.

Vacancy and oversupply varies geographically as Map 1 demonstrates.  Excluding holiday homes all but three local authorities – South Dublin (4%), Dún Laoghaire-Rathdown (5.7%) and Fingal (5.3%) – having vacancy rates (exc. HHs) above base vacancy.  In several cases housing vacancy (exc. HHs) is running above 10% and four local authorities have rates above 15%.  The issue is particularly acute in the north west.

Map 1: Housing vacancy (exc. HHs) in Ireland

Map 1: Housing vacancy (exc. HHs) in Ireland

One might expect that the vacancy rate has been declining everywhere, but this is far from the case.  In fact, vacancy has been rising in many EDs.  In Figure 1, each dot is an ED, with each dot above the line representing an increase in vacancy (exc. HHs).  In some cases the increase is quite dramatic.

07_Scatterplot_BxPltSo, the question is what has led some EDs to increase in vacancy?  Some of it is obsolescence – in any housing market 3-5 properties drop out of use in a year.  Some of it might be properties under-construction and on unfinished estates being completed (and thus counted) but are not yet occupied.  And some of it will be related to population change and migration.

Here, we want to look at the latter since a large number of EDs lost population between 2011-2016, especially those in rural areas (with towns in rural counties growing).

Map 2 shows population and vacancy (exc. HHs) categorised into four classes.

  1. (light blue): population decreased and vacancy decreased (687)
  2. (blue): population decreased and vacancy increased (705)
  3. (red): population increased and vacancy decreased (1497)
  4. (light red): population increased and vacancy increased (517)

06_PopChg_and_VacChgThe relationship we would expect would be classes 2 and 3 – where population decreased, vacancy increased, and where population increased, vacancy decreased.  And that happens in 2022 EDs (out of 3406).  However, in 1204 cases (c. a third), something odd happens: as population increases, vacancy increases (517 cases), or as population decreases, vacancy decreases (687 cases).  In the case of the latter this might be explained by obsolescence and household fragmentation.  We would be interested to hear of other possible explanations.

Without further analysis it’s not possible to determine the causes of this inverse relationship.  However, what the data does show us is that how housing vacancy is unfolding is not universal and there are different social and spatial processes at work.

Rob Kitchin and Martin Charlton

Much of the coverage concerning the preliminary Census release from yesterday has focused on vacancy. This has meant distinguishing between those housing units classed as holidays homes in each area and units that are ordinarily vacant. One of the more puzzling statistics to emerge from this distinction is the 190% increase in the number of holiday homes in Dublin city since 2011. In that year, there were 322 vacant holiday houses in the city but those rose dramatically to 937 in April’s census.

What might account for this near trebling in five years? In particular why, in the middle of a housing shortage, is there almost twice as many housing units classed as holiday homes in a dense urban area when compared with five years previously? Speculation with some others on twitter concentrated on the possibility of these being AirBnB properties. I decided to put some focus on this explanation to see if there’s any truth to it.

In recent months there has been a number of online features concerned with how AirBnB (a company which matches bodies with beds across the world) might be affecting rents. If people are renting their city property through AirBnB for much of the year, how might this affect people seeking to live and work in the city full time? For example, in a number of North American cities there are concerns that full-time AirBnB rentals are displacing residents (e.g. see here or here) who are in lower paid jobs and subject to ever-increasing rents. Some city authorities are coming under pressure to restrict the use of full-time rentals through the company. A property owner can make far more renting out short term lets to passers-by than s/he can from locals who are seeking merely to continue living in a city they work and have a life in.

There is a vital politics to this displacement where AirBnB rentals are pricing people of lesser means out of particular areas of a city bustling with tourists. It is an extreme example of gentrification by displacement, almost making the popular term redundant in its bluntness. The uneven geographies of AirBnB rentals hits home for many in this city too.

In Dublin in June, the city council raised the prospect that full-time AirBnB rentals in Temple Bar, a particular zone of intense tourist activity, would be subject to planning permission. The Council argued that a particular property in the neighbourhood was effectively a material change of use from residential to commercial. It insisted of course that this ruling was “site specific” and did not cover the entire Temple Bar area. The prospect of an imposition of a change of use for the area as a whole is remote though: this seemed like a shot across the bow.

Luckily for us, InsideAirBnB allows us all access to data for rentals across a large number of cities to determine if the company is facilitating displacement. I took the January 2016 data from this site and, aside from knowing the first names of each of the renters, the database contains a number of interesting data.

There are 3,772 properties in the AirBnB database in the four local authority areas. Of this number, 3,116 (83%) are in the Dublin City Council area. 1,222 or 39% of this subtotal are for rent, according to the dataset available under Creative Commons, for 300 days or more per year. The heatmap below (Map 1) shows that near-year long rentals are broadly clustered within the Temple Bar, Cow’s Lane and north Docklands areas. Those rented 365 days per year (249 properties) are distributed slightly differently. They are by no means overlooking the splendour of Dublin Bay.

Map 1: a heatmap of the 1,222 properties available for 300 days or more on AirBnB. Data: InsideAirBnB and OSM contributors.

Map 1: a heatmap of the 1,222 properties available for 300 days or more on AirBnB. Data: InsideAirBnB and OSM contributors.

They are scattered across the city with some clusters in Drumcondra, Rathmines and Portobello. Map 2 below shows the distribution of these year-long AirBnB properties across the Dublin City area, Map 3 shows the distribution of entire house/apt available for rent for 300 plus days a year (as opposed to a room in an already occupied dwelling).  It is not beyond the realm of possibility therefore that the >300 days per annum rentals in this database includes a figure of 937 holiday houses recorded in the census. In fact, there are 934 properties rented out for 335 days or more per annum in the database. If you spent the month of December in your own city centre apartment, and rented it out for the remaining 335 days of the year, you might well be among the 937 recorded in the Census.

Map 2: Year-long AirBnB rental properties (n=249) in the Dublin city area. Data: InsideAirBnB and OSM contributors.

Map 2: Year-long AirBnB rental properties (n=249) in the Dublin city area. Data: InsideAirBnB and OSM contributors.

13700067_10209303077727651_7192871880924738136_n

Map 3: 300 plus days per year of entire housing unit /apartment for rent on Airbnb. Data: InsideAirBnB and OSM contributors

But this is a numbers game. We’ll have a better sense of the distribution of the city’s holidays homes when the more extensive data release begins in April 2017.

Eoin O’Mahoney

 

For the past couple of years the housing discourse for Dublin city has been one of housing shortages and a homeless crisis. The preliminary census figures published yesterday reveal that while the vacancy rates (exc holiday homes) for South Dublin (4%), Dún Laoghaire-Rathdown (5.7%) and Fingal (5.3%) are below a base vacancy level of 6% (in a ‘normal’ market we would expect c.6% of stock to be vacant due to selling/rental gaps, deaths, etc), suggesting that they have housing undersupply, Dublin City Council has a vacancy rate of 8.6% (exc. holiday homes).

In total DCC has 21,781 vacant units (20,844 exc holiday homes).  On a base vacancy of 6% (14,544 units) that suggests an oversupply of 6,300.

In other words, there is something pretty odd going on given the homeless rate has been increasing, large numbers are on the housing waiting list, and there’s a widespread belief that the city desperately needs to build housing.

So, what constitutes these 6,300 excess vacant units?

It’s somewhat difficult to know without visiting them and doing an on-the-ground survey, but let’s start with looking at the geography of vacancy in DCC.  Map 1 shows the % vacancy in the city minus holiday homes, and Map 2 shows change in the number of vacant units since 2011.

Map 1: DCC vacancy rates (exc holiday homes)

Map 1: DCC vacancy rates (exc holiday homes)

 

Map 2: DCC vacancy change 2011-16

Map 2: DCC vacancy change 2011-16

In Map 1, all the areas not shaded pale yellow has a vacancy rate (exc. holiday homes) above 6% base vacancy.  Much of the city centre and to the south have rates above 10%, and two EDs have rates above 20% (Mansion House B, Pembroke West B).  In Map 2, the blue areas have seen vacancy rates decline between 2011 and 2016, whereas red areas have seen an increase.  Interestingly, a number of areas have seen quite large increases in vacancy, especially within the canals near to the city centre, Ballsbridge and Rathmines.

Here’s some speculation as to what constitutes the excess vacancy:

  • some unreported airbnb/similar stock;
  • some second homes (used during week, but primary residence recorded as somewhere else);
  • some investment stock left empty;
  • some bedsits not yet converted after change in regulations that made them illegal;
  • some inner city obsolescence.

I’d be interested to hear about other possibilities.

Whatever the reason for the vacancy, it appears that this stock is not presently available to the market and therefore there continues to be a shortage of housing in the capital.

Rob Kitchin

The ESRI published a report this morning concerning housing supply projections up to 2021. Along with the Housing Agency housing supply report published in April 2014 and the CSO regional population projections published in December 2013, it suggests the need to create substantial new supply in the Dublin region and the other principal cities — no surprise to anyone who has been trying to buy in the region or is on the social housing waiting list.

To summarise: housing need projections

The ESRI report details projected housing supply need until 2021. It argues that there will be an increase in household demand of 180,000 units, but because of oversupply in many parts of the country only 90,000 new units will need to be built, some 12,500 per year. 56,000 (60%) of these need to be in Dublin, 8000 per year. 26% more will need to be in the Dublin commuter counties of Meath, Kildare, Louth and Wicklow. Overall, 86% of all new build will need to be in the Greater Dublin region. However, in many counties, the report suggests that new supply will not be needed because of existing oversupply. Indeed, Donegal, Kerry, Mayo, Tipperary, and all the Upper Shannon counties of Leitrim, Sligo, Cavan, Roscommon and Longford are projected to still have oversupply in 2021.

The Housing Agency report analyzed housing need for 272 towns and cities across the country for the period 2014-18. It argued that there was a need for 80,000 new units, or 16,000 per annum. 37,500 units (47%) would need to be built in Dublin, or 7,500 units per annum.

Both reports use a fairly standard housing projection model using housing stock, population projections, household size, vacancy and obsolescence.

The CSO regional population projections gave a mid-term estimate of population numbers in 2031 using two scenarios. The projections predicted that Dublin population would grow by between 96,000 and 286,000, and the Mid-East region by 77,000 to 144,000. In the upper scenario the Greater Dublin region would therefore see its population grow by over 400,000. In contrast, in the lower scenario, the Border region population would increase by just 18,000 and the West by 17,000. Although these figures relate to population, they will clearly need to be housed and these figures suggest the need for substantially more housing stock over the next 17 years.

There is pretty good harmonisation between the ESRI and Housing Agency reports, both suggesting that c.8000 houses need to be built in Dublin per annum to meet demand.  The overall national required rate of between 12,500-16,000 per annum is actually quite modest.  Typically over the past forty five years new build has been 20-30,000 per annum, rising to 40,000+ post 1998.  12,500 is in fact lower that the lowest build rate going back to when DECLG records start in 1970.  In other words, this is by no means an excessive ambition.

So why do we need supply in Dublin given the crash, oversupply, emigration, etc?

In short, the oversupply of the boom for houses in Dublin as a whole was relatively small, and there wasn’t one in South Dublin. There was, however, a reasonably large overhang of apartments. However, since 2008 the three main drivers of housing demand have been growing: natural increase, in-migration to the city, and household fragmentation. These have soaked up the oversupply. On the other side of the equation housing supply has been minimal. In 2013 only 1360 units were built in the four Dublin local authorities (only 8301 nationwide, over half of which were one-offs and generally not for sale on the open market). In short, over the past seven years we’ve moved from having excess supply to excess demand in Dublin and some other urban locations.

So if there is demand why isn’t there supply?

Good question. Housing supply is shaped by a number of factors: demand, available zoned land, planning permission, building costs (materials, labour), regulatory conditions/costs (taxes, levies, fees, etc), finance (for developers and consumers), and ability to make a profit.

In theory a lot of the right criteria for creating supply exist. There is an excess of demand. There are 6400 acres of zoned serviced land available in the four Dublin authorities for 132,000 units. There are a lot of outstanding planning permissions still in effect and LAs want to give permission for developments that meet development plan/zoning criteria. Material and labour costs of significantly lower than the boom time.

And yet, supply does not seem to be coming on stream and there seem to be blockages across the board. With respect to land, it may be the case that owners are not bringing it to development because they bought it in the boom and can’t afford to develop at present house prices. With respect to planning, it may be that developers are seeking permissions that contravene development plans or are trying to alter existing permissions. The property industry also say that the system needs streamlining and simplifying. They also make the case that there are too many taxes and disincentives attached to building such as development levies, VAT, stamp duty, building reg costs, etc that amount to a sizable proportion of any sale price. Finance is a critical issue. Developers need a sizable amount of upfront cash to secure development loans, yet many are bust from the boom or do not have such reserves.

So what are we to do?

The government needs to quickly evaluate each of the potential blockages and work out solutions that are fair and do not undermine good planning and build quality or excessively boost profit at the state’s expense. By quickly I mean weeks, not months and certainly not years. The longer that supply is constrained the more demand there will be on existing stock and house prices will continue to rise. The Construction 2020 strategy is full of task forces, review groups, consultation exercises and very short on actual policy and implementation. We need supply coming on stream as quickly as possible in the Dublin region and some other urban locales (though certainly not in many parts of the country). Construction 2020 thus needs to be fast tracked. After all, 2020 is meant to be an end date, not the date ground is broken.

Rob Kitchin

There is no official data regarding negative equity in Ireland in general, nor its geographical distribution.  By mid-2012, once house prices had fallen to 50% of their 2007 values Davy Stockbrokers estimated that more than 50% of residential mortgages were in negative equity.  Consequently, any house bought from 2000 onwards is likely to be in negative equity.

Negative equity is a significant issue because it creates a spatial trap that restricts mobility. Because the value of the property is less than was paid for it, owners cannot sell and move to another property without realising a loss.  This trap has three consequences.  First, it restricts labour market mobility.  Second, it keeps families in homes that may no longer be suitable to their needs.  Third, it restricts the pool of properties available to the market and limits any recovery to first time buyers, those prepared to realise a loss, those whose property is not in negative equity or have investment capital.  All three have social and economic consequences causing hardship and stress and slowing the recovery of the wider economy.

Negative equity is not evenly distributed because it is determined by the price paid relative to present prices and this is largely shaped by when the house was bought.  So where might this spatial trap be operating most perniciously in Ireland?

This is not an easy question to answer given publicly available data sources.  We have been looking at proxy measures and present one here, though it should be noted that it only captures one kind of property in negative equity – houses that were built post-2001.  It does not include secondhand houses in negative equity, nor buy-to-let properties in negative equity (though the latter can be estimated using a same method).

Our solution is to use two Census 2011 variables at the Small Area level. The first variable is the ‘% of housing units built post 2001’.  The second variable is the ‘% of outstanding mortgages in an area’ (i.e., the property has been purchased not rented privately or from a local authority or voluntary body).  These variables are not perfect, but when combined do give us, we think, a reasonably good proxy.

The figure below is a density smoothed scatterplot of the two variables for all 18,488 Small Areas in the country.  Each Small Area has approximately 80-130 households.  We have divided up the scatterplot into four quadrants, one of which is subdivided based on the clear pattern of points, to create six categories that denote different levels of negative equity (category 1 has very low rates of both post-2001 build and outstanding mortgages), which we have then mapped from the country and for Dublin.

negative equity

????????????

Map_NegativeEq_Dublin

It is important to note that all the Small Areas potentially have some households in negative equity, but that some areas have greater concentrations than others.  In broad terms, categories 5 and 4 are likely to have similar levels of private residential negative equity, but we have left them separate to denote their different characteristics.

When these categories are mapped the pattern that emerges is perhaps what one would expect.  The areas with the highest concentrations of negative equity are in the outer suburbs of the cities and the fringes of commuter towns.  These areas experienced high rates of newly built properties and new household formation all through the boom, but especially in the latter years when the inner suburbs became too expensive for first time buyers and those trading up to a family home.

This pattern is very clear around Dublin, Cork and Limerick, but is slightly different around Galway, where a number of rural Small Areas are highlighted where there was a lot of one-off housing and small nucleated settlement.  This pattern is repeated for many smaller rural towns.

Owner occupiers in these areas are more likely to be spatially trapped, though as noted any individual household in any part of the country could be suffering such a fate.  It is also likely that the same areas will have higher concentrations of mortgage arrears, given that negative equity and mortgage arrears are related.

Whilst further research is needed to refine this analysis it does give a proxy measure of one kind of negative equity in the absence of detailed data from mortgage providers.  We would be interested in any feedback about the approach taken.

Rob Kitchin

Following on from my earlier post on developing scenarios for Ireland’s future, last week saw the publication by the CSO of new regional population projections for Ireland. The latest CSO projections present how the population of the regions may evolve under different scenarios by making assumptions about future trends in migration (both internal and external) and fertility.

In the context of regional development in Ireland, this latest release from the CSO is crucial as all current regional planning policy together with the settlement strategies of all local authorities are currently based on population targets (including those of ‘Gateways’ and ‘Hubs’) originally derived from the previous set of population projections issued by the CSO in 2008.

The CSO projects that, if internal migration patterns return to the traditional pattern last observed in the mid-1990s, the Greater Dublin Area (GDA) will continue to see its population significantly increase by just over 400,000 by 2031 which will account for two-thirds of the total projected population growth in the State over this period. As observed in the post below, this pattern appears  not only to be re-emerging but rapidly intensifying as the Government increasingly focuses its attention on a FDI led job creation strategy which favours agglomeration economies i.e. Dublin and to a lesser extent Cork and Galway.

Chart

In a marked change from the 2008 release, which projected that the population of Dublin would decline by just over 100,000 and that of the Mid-East increase substantially by over 350,000, the population of Dublin is now projected to increase by between 96,000 and 286,000 depending on the scenario applied, while the population of the Mid-East is set to increase more modesty by between 78,000 and 144,000. These trends would have profound implications for spatial planning, housing policy and infrastructure delivery in the capital.

While all regions are projected to experience net population growth, apart from Dublin and the Mid-East all regions will lose population to internal migration and population growth will be primarily driven by natural increase (i.e. birth rate). This will be most noticeable in the Border region with, under one scenario,  projected births of 123,000 and a population increase of just 18,000, and the West which shows projected births of 97,000 and a population increase of  just 15,000. In fact, under some scenarios the Border, West, Mid-West regions are projected to experience population decline regardless of the internal migration pattern applied.

Overall, the CSO projections paint a familiar picture with Dublin and the Mid-East gaining a higher share of the national population (particularly of younger and more highly educated persons) with everywhere else generally losing share. These demographic trends present a key national and regional development challenge and far-reaching questions for planning practice*. Depopulation and changing population structure implies severe impacts on every domain of urban and regional development, including local authority budgets, infrastructure and amenities, housing market and housing mobility, labour market and employment, residential composition, and social inclusion and cohesion – the entire basis of regional planning policy.

More generally, current and projected population trends highlight the urgent need for a new National Spatial Strategy and to transition beyond the current ‘performance of seriousness’ in relation to balanced regional development.

Gavin Daly

 *See Daly, Gavin and Kitchin, Rob (2013) Shrink smarter? Planning for spatial selectivity in population growth in Ireland. Administration, 60 (3). pp. 159-186. ISSN 0001-8325

Next Page »