Ask HN: How to tell if a home is unoccupied? | Hacker News Reader

Ask HN: How to tell if a home is unoccupied?

9 pointsspoonie15 comments

Premise: a city wants to know how many homes are unoccupied. How can they reliably measure this?

Sending city staff door to door to see if anyone is home is probably expensive, disruptive, invasive, and inaccurate if residents keep irregular hours. Sending snail-mail surveys is probably also expensive and would get a lot of false negatives. Can utilities services help? Does electricity use imply occupation? Does this give false negatives for households with low electricity use and false positives for empty houses with a few token appliances plugged in? What about natural gas and/or water usage? What about more indirect measurements like traffic/transit data or tax status (for rental housing at least)? Could telephone-pole mounted cameras measure foot and motor traffic?

(Disclaimer: this question isn't about whether unoccupied homes are a problem for a city, it's just a thought experiment about how this question could be answered with technology. I admit this question is inspired by speculation that empty homes in places like Vancouver, BC and SF are affecting rents and availability.)

Define unoccupied:

- By original owner?

- By original owner or renter?

- Cleaning lady coming once in 2 weeks for 5 years is ok?

- Duration of time no living being is present within the premises?

- Neighbor taking mail once a day?

- Average number per district is ok or precise addresses are required?

etc .... Clear definition is needed

I was originally thinking 100% lived-in, by anyone wether they are the owner or not, and precise data down to each housing unit. That definitely gets into shades of grey when you consider vacations and second homes, or even granny flats! What if someone owns a house but goes on vacation for 2 months every year? If you just want data you can probably trust residents to self-report this but if you were trying to implement a vacancy tax that might not work.

Our city gets reliable statistics via utility data. I believe this is common.

This sounds like a reliable method (lots of data points!), and using different utilities should decrease false positives.

A common method in Melbourne for measuring apartment occupancy is measuring water usage.

I worked for a utility and this was an important problem for us as a customer would close their account and move out of a rental and another tenant may move in and start using gas without having an account with one of the utilities (we don't lock meters on moves, this would be too expensive.) The utility company only finds out something is using gas when a meter is read every few months (we had a threshold.) Here is when we contact the tenant and try to figure out whats going on, unfortunately we don't have phone numbers and sending someone around to check was too costly at the time.

You could feature extraction per target home. I wonder if you could use neighbors' information to have better posterior beliefs about the home in question.. Like, suppose Alice lives in the house you're trying to decide, and Bob, Cathy, ... Yvonne, and Zara are all benign 'adversaries' that are willing to help you learn about Alice's occupancy. Is there some kind of utility/wiring/plumbing/logical singleton shared across the whole neighborhood, which could be used for deductive purposes.

Cheapest method: Assume the government really want this number, they can have all mailman help keep track if there are leftover mail(a letter is created/given if resident has no mail) after x days.

Technology wise, I cannot think of an easy and cheap way. Camera with vision for each residential door will be a robust approach. Accuracy can be precise by more investment in software from detecting opening-of-door-event to facial recognition of the expected household members.

First get the low hanging fruit. Wifi activity, utility use.

Then look for commonalities in the homes that you can't tell about.

Tricky edge cases

If someone lives in a home for 6 out of 12 months per year, and the rest of the time it is empty does it qualified as occupied or unoccupied or do we count it as 0.5?

What about air bnbed for 75% of the time.

Do you care about 1 person living in a mansion (most of which is unoccupied) vs. 10 people living in a 1 bed apartment?

Indeed! It would depend on whether the city wants data for further analysis or will act directly on the data (e.g. new taxes). If the city just wanted a punitive tax on non-occupied housing, they'd probably be better with LVT or a revenue-neutral shift away from sales tax or payroll tax towards higher property taxes. This would indirectly affect non-occupying owners without needing to target them directly.

Not that I condone it but you could get the public to report unoccupied homes using their phones. Especially neighbors who could established a history of vacancy.

Body heat detection equipped drones could help.

How would you tell between a home that is legitimately occupied and an abandoned property that has been taken over by squatters?

so theyd report 53.56% occupied rather than 53.57.. is that sucha big deal when its very very close and performs much better than alternative methods on an extremely technically challenging problem?

Reply on news.ycombinator.com