You didn’t let me down 👏. Would be interested to hear your thoughts on how transparent methodology and definitions are in relation to CSO data? I actually enjoy digging through data and sometimes I can find them opaque at best on the site, never mind someone looking to confirm what they already believe (I also think the hierarchy is not particularly intuitive, and it’s easy to miss tables that might be relevant to your search - but that’s another rant). A very general learning from this could be the need for the CSO to more prominently explain “what this data represents”, “how it can be interpreted” and “how it can’t” (of course that gap is, I’m sure, why you started doing this).
Thanks Shane, I actually do think the methodology and definitions from CSO are strong TBH. The doubt has been cast because people didn’t like what they counted, but it stands up to scrutiny. That’s the point of this post.
And also I’m really not keen on forcing the CSO to come up with any more explainers of “how it can be interpreted”. Particularly for a census, their most important job is to compile accurate data, not to steer people in the direction of how to interpret it.
We saw this in the last census when they were pressurised into publishing a caveat with the vacancy figures. That caveat then became a citable reason to ignore the data in favour of poorer measures.
Already there are some noises that they will be changing the way they measure vacancy in the next census, all thanks to the misguided vacancy data debate. This is unlikely to be an improvement, not least because we then cannot usefully compare 2016 vs 2022 vs 2027 if the methodology is different.
Pressurising the CSO to make data politically palatable is a dangerous precedent.
Sorry when I said transparent I may not have been clear. I more so was referring to accessibility. E.g. If I log on to data.cso.ie and navigate to a table such as F2098, I think including a definition of what is being measured here would go a long way. For example, F2098 measures vacancy rate, but there's no easily accessible definition on what "vacancy" means.
I definitely don't agree with making data politically palatable, and that's also not what I meant to imply. However, I do believe that an important role of the CSO is not just to gather accurate data but also to do their best to ensure that data is used and used properly.
Yes, I take your point re data.cso.ie but I think it is fair enough to assume that if you have got to the level of interest to dig deep into the data on there, you would have a solid understanding of definitions and methodology sourced from elsewhere on CSO website or publications. If they were to try and layer on precise definitions for every dataset on data.cso.ie it would impair the simplicity of use. I think it is primarily built for data nerds who know what they’re looking at, and it does that job pretty well IMO.
Your example of F2098 is a good one though. It is just vacancy rate by ED. Eg Mansion House B, Dublin City is 20% (pretty high for prime property!).
If a user is not sure what that is, it can be compared to F2095 which is also vacancy by ED but includes total housing stock, holiday homes etc. The F2095 vacancy rate is also 20% for Mansion House B, being 91 vacant out of 330 total dwellings, but you can see how the stock has been broken down:
• 330 occupied by usual residents
• 3 occupied by visitors only
• 15 unoccupied – residents temporarily absent
• 16 unoccupied – holiday home
• 91 unoccupied – vacant house or flat ← the 20%
So the vacancy rate already excludes holiday homes and the temporarily-absent (people away on census night). It's counting only genuine point in time vacant dwellings — not "everything that wasn't occupied." That's the same census-night measure used throughout my post: a clean count of empties, with the holiday-home and temporarily-away categories stripped out separately. This is what the CSO report as the census vacancy rate.
You didn’t let me down 👏. Would be interested to hear your thoughts on how transparent methodology and definitions are in relation to CSO data? I actually enjoy digging through data and sometimes I can find them opaque at best on the site, never mind someone looking to confirm what they already believe (I also think the hierarchy is not particularly intuitive, and it’s easy to miss tables that might be relevant to your search - but that’s another rant). A very general learning from this could be the need for the CSO to more prominently explain “what this data represents”, “how it can be interpreted” and “how it can’t” (of course that gap is, I’m sure, why you started doing this).
Thanks Shane, I actually do think the methodology and definitions from CSO are strong TBH. The doubt has been cast because people didn’t like what they counted, but it stands up to scrutiny. That’s the point of this post.
And also I’m really not keen on forcing the CSO to come up with any more explainers of “how it can be interpreted”. Particularly for a census, their most important job is to compile accurate data, not to steer people in the direction of how to interpret it.
We saw this in the last census when they were pressurised into publishing a caveat with the vacancy figures. That caveat then became a citable reason to ignore the data in favour of poorer measures.
Already there are some noises that they will be changing the way they measure vacancy in the next census, all thanks to the misguided vacancy data debate. This is unlikely to be an improvement, not least because we then cannot usefully compare 2016 vs 2022 vs 2027 if the methodology is different.
Pressurising the CSO to make data politically palatable is a dangerous precedent.
Sorry when I said transparent I may not have been clear. I more so was referring to accessibility. E.g. If I log on to data.cso.ie and navigate to a table such as F2098, I think including a definition of what is being measured here would go a long way. For example, F2098 measures vacancy rate, but there's no easily accessible definition on what "vacancy" means.
I definitely don't agree with making data politically palatable, and that's also not what I meant to imply. However, I do believe that an important role of the CSO is not just to gather accurate data but also to do their best to ensure that data is used and used properly.
Yes, I take your point re data.cso.ie but I think it is fair enough to assume that if you have got to the level of interest to dig deep into the data on there, you would have a solid understanding of definitions and methodology sourced from elsewhere on CSO website or publications. If they were to try and layer on precise definitions for every dataset on data.cso.ie it would impair the simplicity of use. I think it is primarily built for data nerds who know what they’re looking at, and it does that job pretty well IMO.
Your example of F2098 is a good one though. It is just vacancy rate by ED. Eg Mansion House B, Dublin City is 20% (pretty high for prime property!).
If a user is not sure what that is, it can be compared to F2095 which is also vacancy by ED but includes total housing stock, holiday homes etc. The F2095 vacancy rate is also 20% for Mansion House B, being 91 vacant out of 330 total dwellings, but you can see how the stock has been broken down:
• 330 occupied by usual residents
• 3 occupied by visitors only
• 15 unoccupied – residents temporarily absent
• 16 unoccupied – holiday home
• 91 unoccupied – vacant house or flat ← the 20%
So the vacancy rate already excludes holiday homes and the temporarily-absent (people away on census night). It's counting only genuine point in time vacant dwellings — not "everything that wasn't occupied." That's the same census-night measure used throughout my post: a clean count of empties, with the holiday-home and temporarily-away categories stripped out separately. This is what the CSO report as the census vacancy rate.
The definitions and methodology they use for housing stock, vacancy, etc is clearly explained in the Background Notes here: https://www.cso.ie/en/releasesandpublications/ep/p-cpp2/censusofpopulation2022profile2-housinginireland/backgroundnotes/ and the FAQ here: https://www.cso.ie/en/census/census2022/census2022andvacantdwellingsfaq/ - I really don’t think the CSO are being deliberately opaque, quite the opposite, particularly if you compare them to GeoDirectory.