Hacker News new | past | comments | ask | show | jobs | submit login
Can AI Predict the Next Area to Gentrify? (citylab.com)
77 points by laurex on Dec 23, 2018 | hide | past | favorite | 34 comments



Given that a bloke with a list of fried chicken shops and coffee houses can do it for London, I reckon the answer is 'yes'.

https://medium.com/@Sam_Floy/how-to-know-if-where-you-live-i...


That only works if your city is ethnically diverse.


That's funny - I literally couldn't find a coffee shop in Peckham yesterday while my kids were at the cinema.

Plenty of lunatics, drug-takers, and drunks, though.


Seriously? If you were at the Peckhamplex then you were 100 yards from three or four hipster joints, at least.


Probably made a mistake going to the high street.


Really? There are at least two coffee places within ten metres of the train station on Rye Lane, right across from the Cinema. You're talking absolute rubbish.


Peckham, home of the terminator[1].

1: https://www.youtube.com/watch?v=LMACpa_oXjk


It is well known that the stock market is extremely difficult to predict, because any model that has an advantage quickly loses that advantage once it is put into practice.

Real estate, however, has not traditionally had the intense sort of algorithmic focus that the stock market has had. Are there still opportunities to profit based on obvious patterns, and if so, why hasn't this generally been done already? Lack of available of real estate data perhaps? I would think a hedge fund would have no problem contacting a few hundred MLS for the necessary listing data though.


Patterns in real estate aren’t as predictive as you think, especially given the long time to close a deal.

Down at the neighborhood level, a single policy decision or decision to sit on a big plot of land instead of developing it can completely halt gentrification and reverse it.


Also, it's likely a lot of the trends easily incorporated into a computer model (price movements, demographic changes, transport changes, retail outlet changes etc) significantly lag the policy decisions that make them possible. So if you want to profit from real estate price growth in a city, I'd imagine human understanding of why an area might gentrify in future outperforms machine understanding of indicators the area is starting to gentrify. Still, the long time to close a deal also means real estate markets are far from efficient, so there's likely to still be profit to be made with a machine telling you where to look, especially if you do the rest of your research properly.


Real estate is very expensive to transact. Certainly one could not execute short term trading strategies like is common with securities markets.


Don't you usually start a derivative market if the transaction costs of the original market are too high?


Derivative of what, though? There's no index to track that people will agree on or can't be gamed.


In the US there's the Case-Schiller index which is the underlying for several classes of derivatives.


This is true for retail. However if a fund structures it the right way i.e. one property one company, the transaction costs could be similar like a share purchase deal.


Real Estate speculation can have very long timelines.

The data on upcoming developments and rezonings is all public and trivial to find but is also available years in advance. It can take years for public works projects to be completed even if they're not delayed. This means one needs to be able to hold for a long, long time.

For example several years ago the City of Vancouver in partnership with the public transit authority formalized a decision to make a rapid transit line across the city. It's the sort of thing that will increase land values. However due to politics the line was delayed and construction is estimated around 2020 I think. The plan has been floating around since the 1990s and it was already cancelled once in 2001. Similarly there's another line that will likely be eventually built and it's mentioned in the region's 2040 transportation plan but the timeline is so far off.

The expected existence of these improved public transit options are likely to some degree already baked into the selling prices of properties in this area.


There are probably signs people are looking to move somewhere, like an increase in retail/dining activity, so it might be more like predicting who's going to have a baby [1] than predicting the stock market.

1: https://www.forbes.com/sites/kashmirhill/2012/02/16/how-targ...


The money to be made on a property is peanuts to traders. They’d have to buy blocks, but then they are real estate developers and that’s a different game. Most of the edge will be in marketing and negotiation for a property development rather than buy and hold for 10 years hoping the poor people get priced out because there is decent pressed coffee sold from a bicycle friendly hip cafe with exposed brick walls.


Back around 2012 or so I interviewed with a company doing just that to price mortgage backed securities. So it is being done.


i can predict it with near 100% accuracy anywhere in the world. just show me where all the artists and musicians in your city hangout and pay less than 900/mo on rent.


Essentially correct. You can see way ahead of time where this will happen.

That said - there's a new kind of 'suburban gentrification' which is another thing altogether: middle class people not looking to be so much hip, rather than just live in a decent affordable area near downtown that's not crackburg.


In the 90s in Seattle there were bumper stickers that read, "Artists are the shock troops of gentrification"


> just show me where all the artists and musicians in your city hangout and pay less than 900/mo on rent.

Well, that is the trick bit.


That's where Google and FB have an advantage. They track everyone's location and interests. They could probably do it.


> That's where Google and FB have an advantage. They track everyone's location and interests

I was thinking about this two summers ago when there was that migration “crisis” in Europe, i.e. that Facebook knew better than anyone else on the planet how many immigrants there were (give or take) and especially from where exactly they were coming (that was a very contentious point back then) based on their location data and on the location data of their friends and families left at home. Come to think of it, I bet they could also “guess” the general level of education for many of those immigrants, that was also a very, very hot topic back then.


good plan... but I think we hoping you would show us :)


and (of course) lgbt.


Why _of course_? In all the western cities I know the “gayer” districts usually had their share of gentrification around a decade in the past.

Of course artists and creative people in cities tend to be more liberal (because cities in general tend to be more liberal). But in my experience it is rare that the lgbt epicenter and the art epicenter are colocated. Of course art people might hang a rainbow flag out, but that doesn’t mean there are more gays/lesbians/queers living there.


From my observations, although much has been written about this subject as well. For example, Boston's South End (struggling and transient), Provincetown, Ma (cape cod) a forgotten blue collar village on the decline were all "discovered" by the gay culture and now, paradoxically, their gay character is threatened by big money moving in.


Personally I’m less interested in looking for neighbourhoods where gentrification is going on than I am to discover what makes places become “not sketchy”.

What small changes influence the crime levels and feel of a place when walking around that could improve it for everyone living there?

When those things improve house prices naturally rise.


Isn't that effectively the same thing? The long tail of gentrification is rising prices and (somewhat) lowered crime.

I've long suspected you can predict this with ML given the right features but also realized you don't really need any autonomous process, it's generally an understood pattern.

Perhaps if you wanted to buy thousands of properties across the country.

* I've long noticed that people in gentrified areas will accept higher crime rates not just because the inevitability of being right next to a not-yet-gentrified area, but because it makes their area feel more "real" or "gritty."


'Walkability' is a concept that comes up on HN quite often. Things like awnings and doorways can make somewhere feel more comfortable to walk in and I guess that would make it feel less sketchy.


The authors of the cited study posted their ipython notebook on github:

https://github.com/jreades/urb-studies-predicting-gentrifica...


Based on Betteridge's law, I would say “no”. At least not reliably.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: