# Thread: Linear regression and apartments.

1. ## Linear regression and apartments.

Here's the situation. I'm thinking of buying an apartment. That will, of course, require me to take a credit.

I've also been through masters program lately. Finished all of my exams succesfuly, much to my surprise, I hasten to add. One of those exams was statistics. And you may call me boring, but I found it interesing, for a first time in my life.

Linear regression was a one part of it. And the part I most liked, not really sure why. But... would it be useful in my situation? I'm looking at many offers given to me and I would like to put some perspective to them. Can linear regression help me? I'm used to thinking with stats in medical terms(my masters, you know), but this is pure economics, so it's kinda foreign to me.

How would one set regression model for this? Any help would be appreciated.

2. ## Re: Linear regression and apartments.

Originally Posted by Mordokai
Here's the situation. I'm thinking of buying an apartment. That will, of course, require me to take a credit.

I've also been through masters program lately. Finished all of my exams succesfuly, much to my surprise, I hasten to add. One of those exams was statistics. And you may call me boring, but I found it interesing, for a first time in my life.

Linear regression was a one part of it. And the part I most liked, not really sure why. But... would it be useful in my situation? I'm looking at many offers given to me and I would like to put some perspective to them. Can linear regression help me? I'm used to thinking with stats in medical terms(my masters, you know), but this is pure economics, so it's kinda foreign to me.

How would one set regression model for this? Any help would be appreciated.
Linear regression helps you track correlations between some random (or seemingly random) variables and is often used to falsify some theoretical models that predict specific correlation. With appartaments you have two key quantities that can be expressed by numbers: size of the appartament and its price. The thing is, we already know that those are linearily dependent as most offers show price per m^2 already (at least where I come from), so linear regression might not be all that useful at least at first glance. If you have time to aggregate data about many offers, you might use linear regression to see if there is, for example, some correlation between the size and price per m^2 or you can factor in distance from city center and track correlation of prices with that.

I know that feeling of obtaining a metaphorical hammer and itching to find some nails, but it is not always as simple.

As I went through appartament hunting, I have some general and unsolicited advice:

Preparations:
1. Know your budget! You will take a credit, so you have to know, what is your limit from two sides: what banks would be willing to give you (and on what terms including the down payment for the appartament) and how much are you able to safely pay every month even in bad circumstances. Sometimes it is difficult to get specific numbers here, but you should have some solid estimates ready.
2. Spend some time pondering on where you want to live. Condensed residential area in the city center or sparsely populated suburbs? What do you consider important? Mass transit availability, car traffic, shops and other services nearby (including healthcare, schools etc. depending on your future plans), parks, lakes or whatever is important for you. Making a short list of key features of a given area to check with every appartament offer might be a good idea. Thinking about everyday things you might need to do often helps and so does checking the typical commute routes as living in a nice neighbourhood means not so much, if you lose a lot of time daily stuck in traffic whenever you go to work due to for example a single road leading out of the whole area.
3. How big an appartament you want? Just a small one-person studio or do you plan to have roomates/family with you at some point?
4. (optional) Confront point 3 with point 1 and cry yourself to sleep.

Now you have the information needed to efficiently sort appartament offers:
1. You can buy a new appartament or a used one. The former are often sold in a state that needs some additional work (painting, flooring and obviously furnishing), but the standard varies between developers and affects the price. Nevertheless, you need to be aware that you will need more money than just for the appartament as getting even just the basics done costs a lot. Used appartaments more often come ready or almost ready to use, but this varies between offers and sometimes the state of the appartament warrants an immediate renovation and striping walls and floors takes additional time on top of what you would need to do with a new appartament.
2. As you browse the offers, try to get a feel of the average prices and look for some lucky bargains. Here is where statistics can help you as you can easily build a distribution of prices per m^2 to help you compare the offers. There will surely be strong correlation with the location, so adding that detail to your data table would not be a bad idea.
3. Lucky bargains come in two forms: truthful (someone is in a rush, or you caught a good offer from an appartament flipper as not all of them do luxury homes for the rich) or with a nasty surprise. You will only be able to tell the difference, if you make a throughout check of the appartament and its surroundings.
4. Price is not everything, so study the layout of the appartament and try to picture in your mind the furnishing and in general how can you use the space. Some appartaments are so badly designed that you can barely fit anything in despite seemingly large size that is wasted on some twisted corridor or weirdly shaped rooms. If you plan a throughout renovation or buying a new appartament that is not yet built, look which walls are not load bearing as you can in principle take them down and make some new ones elsewhere. Developers often give an option of redesigning such details in not yet built appartaments. So a bad design can sometimes be fixed. If you have a hard time picturing what goes where, big furniture shops often have same exemplary displays on how to furnish a given ammount of space. They sometimes also have online services for building a 3D model of an appartament, which can be useful for this as well.
5. If the appartament already exists and looks interesting on the photos, check it yourself dilligently. Very often you can find flaking paint and more serious damage once you see things up close. You will also get an idea on how loud or smelly the surrounding area is, how well the widows filter that and so on. One of the more important things is thermal isolation, which to be honest is not easy to check without proper equipment, but if you happen to check out some offer in winter, you might be able to notice for example poor quality windows as you will be able to feel cold air close to them even when they are shut. Another trick that sometimes works: if you close all the windows and close a door to some room said door might start making noticable sounds because of the air flow through that small crack at the bottom - this is an indicator that at least some windows are shoddy or there is some other problem with ventilation.
If you spot mold on a wall somewhere, look for a different appartament. More often than not it is due to a leak or an inherent thermal isolation problem - not something you can fix on your own.
6. If you buy a used appartament it is also a good idea to get some information about the way the whole building is organised (as there has to be some form of homeowner association) and how well it is run: is the building as a whole in good shape, are the common areas kept clean and so on. If you can, get some information on what is an average monthly cost of having an appartament from potential neighbours as those things also vary heavily: old buildings tend to need more heating in the winter as well as serious renovations more often.
7. Checking if there is a private parking space or a garage hall might also be a good idea if you have a car or plan to have one.
8. Once you found some appartament that you might actually buy there is one very important thing to do: read the full land and mortgage registers! I cannot stress this point enough.

After that you usually sign some contingent contract (obviously read that carefully) with which you can go to the bank and ask for credit, which can be a journey on its own.

TL;DR Good luck. You will need it.

3. ## Re: Linear regression and apartments.

Taking another perspective on linear regression, it's a tool for prediction. So if you have something that you want to predict, but which you have historical data about, you can use it to make predictions in the case of things where the answer is currently uncertain. For that kind of usage, you need to have a thing you want to predict, and you need to have past data.

For example if you were buying an apartment both to live in and as an investment with an eye towards selling it in 10 years or something like that, you might predict the maturation of value of the apartment you're buying based on historical trends in apartment values for different sizes, base costs, proximity to various amenities, etc.

Or you might really care about the expected repair costs of the place (though you'd need to get a dataset of other apartments and how often/how much were paid for repairs, and I don't know how to find such a dataset).

4. ## Re: Linear regression and apartments.

Originally Posted by NichG
Or you might really care about the expected repair costs of the place (though you'd need to get a dataset of other apartments and how often/how much were paid for repairs, and I don't know how to find such a dataset).
YMMV but here it is mandatory for an apartment building, effectively the tenant's asociation, to have a renovation plan for like 10-20 years. Basically, it'll list if roof, major pipeing, windows, elevator etc has been done, when and when they plan to or expect it to be needed.

In general the only things I can think of that statistics can look at is how price and size relates to the few tangible factors involved. Remember: location, location, location. Sometimes that is abit artifical, a house on the corner having a different address because it's "better", maybe you belong to a better school area or something, or your postcode just so happens to include 99% of the bad part of town. Should be possible to spot outliers in price, both expensive and cheapo ones?

5. ## Re: Linear regression and apartments.

If you call the local police department, they might be able to tell you the crimerate or something about it. My mom worked at a police academy, so he had a contact, but I found it helpful when house-shopping to get feedback, because some places looked nice but we got a big NO about living there. (And, driving through anyway, yeah, place looked shadey. Though one time the neighbors offered me a beer around 2 pm while we were looking at it!)
But, back to serious: the local police might have some info. Not sure how publicly they'd share it, but no harm in calling and asking if you have just a few properties you care about.

Originally Posted by snowblizz
In general the only things I can think of that statistics can look at is how price and size relates to the few tangible factors involved. Remember: location, location, location. Sometimes that is abit artifical, a house on the corner having a different address because it's "better", maybe you belong to a better school area or something, or your postcode just so happens to include 99% of the bad part of town. Should be possible to spot outliers in price, both expensive and cheapo ones?
That does sound like a useful statistical analysis. Not linear regression itself, but simply map price vs. square mileage for each zipcode/postcode (e.g., one chart per postcode, or school district zone, something like that) and see what looks like an outlier. Then investigate to see if it's a good buy in an otherwise expensive area or a bad buy in an otherwise good area.

On bedroom/closet/bathroom count: someone above mentioned noting the layouts, and I agree. I had an apartment where it said "2 room", but it was really one big room with a... sorta high cubicle-esque wall dividing the 'bedroom' from the 'living room'. Also looked at a house where one of the bedroom was a walk-through room, being the only way to access the master bedroom. Which could work well, but not if you want one bedroom for each of your kids. From what I can tell in the US, a "bedroom" has to have walls so high and a closet, so sometimes things are technically 'bedrooms' even if not practically such. I know you might be in a different country, but there could be similar technicalities of language that could make a place look great but be disappointing.

Also, if you can somehow talk to the super (e.g., repairman) of the apartment complex, if it has one, that could be a boon. I reckon most places won't let you easily do that, but I was looking to rent an apartment and the super gave my wife and I the tour. He noted how the walls were really thin and you could usually hear the neighbors. We probably would have passed on that place anyway (the kitchen was super skinny), but that was definitely a factor. And not something we could have realized during the tour.
I still am rather surprised the super was so honest, but I reckon his pay is not tied to how many units get rented.

I'll also add I've never looked into buying an apartment. Rented some, but only looked at houses for buying. Some maybe some of this isn't relevant, but I'd think most would copy over.

6. ## Re: Linear regression and apartments.

Some things also depend on local laws, so some things also depend on where you live.
And if you like an apartment, try to take a professional (a contractor, an architect you know) with you the second time you go visit to see things someone which non-professionals don't see.

7. ## Re: Linear regression and apartments.

Originally Posted by farothel
Some things also depend on local laws, so some things also depend on where you live.
And if you like an apartment, try to take a professional (a contractor, an architect you know) with you the second time you go visit to see things someone which non-professionals don't see.
This is a very good advice. If you know a professional, who could look a prospecting appartament over, it could be a huge help - especially in noticing some hidden problems. On that note, an infrared camera will tell you a lot about an appartament, but almost noone has one.

8. ## Re: Linear regression and apartments.

There has been a lot of good advice so far, even if has been unsolicited Keep 'em coming, you guys. Be it apartment wise(practicality and all that) or statistics(I welcome further education, after all).

I visited my first potential buy today and I liked it. I will visit more... this is me buying, after all, not merely renting. But yeah, more debate is welcome.

If mods feel like moving the thread is needed, I'm ok with that as well.

9. ## Re: Linear regression and apartments.

Originally Posted by Mordokai
Here's the situation. I'm thinking of buying an apartment. That will, of course, require me to take a credit.

I've also been through masters program lately. Finished all of my exams succesfuly, much to my surprise, I hasten to add. One of those exams was statistics. And you may call me boring, but I found it interesing, for a first time in my life.

Linear regression was a one part of it. And the part I most liked, not really sure why. But... would it be useful in my situation? I'm looking at many offers given to me and I would like to put some perspective to them. Can linear regression help me? I'm used to thinking with stats in medical terms(my masters, you know), but this is pure economics, so it's kinda foreign to me.

How would one set regression model for this? Any help would be appreciated.
You need a degree in neighborhood evolution or whatever covers it to do good here.

Generally I'll say unless you think you're better than the average flipper, buy where you think your kids can graduate from school in a couple decades. Rent every place else. upkeep expenses are real if you aren't pretty good with some tools.

IF you are moving to an area you are unfamiliar with you'll want to look at murder maps by zip code in 1980, y2k and 2020. I am familiar with St Louis and can tell you where NOT to buy if you plan on selling in X or XX years. That roughly correlates with the murder map BUT you'll not notice the histories of the 'flighties' we'll call them and where they have moved now. No reason to think they'll do any better keeping up their new neighborhood than their old one.

10. ## Re: Linear regression and apartments.

Speaking as a statistical consultant, you're approaching the question backwards. You don't decide to use a specific statistical tool and then look for variables to use it on. You look at a problem, figure out what question you need to answer to solve it, see what meaningful variables you could measure to estimate that answer, and then use the tool that will solve the problem with the available data.

You don't decide to use regression and then look for some variables to regress on, for the same reason that you don't decide to use a picture hanger without first having a picture to hang.

If you have a specific numeric variable that you want to predict (y), and you have data of that variable and one or more related variables(x1, x2, ... xk) that have some degree of linear correlation (R2> 0), and if the xi variables can be measured before you have to predict the y variable, then linear correlation is a useful tool.

But deciding to use linear regression without first having variables to regress on is like deciding to calculate a mean when you don't have a random variable, or deciding to use dish detergent before determining if you have any dirty dishes.

11. ## Re: Linear regression and apartments.

Originally Posted by Jay R
Speaking as a statistical consultant, you're approaching the question backwards. You don't decide to use a specific statistical tool and then look for variables to use it on. You look at a problem, figure out what question you need to answer to solve it, see what meaningful variables you could measure to estimate that answer, and then use the tool that will solve the problem with the available data.

You don't decide to use regression and then look for some variables to regress on, for the same reason that you don't decide to use a picture hanger without first having a picture to hang.

If you have a specific numeric variable that you want to predict (y), and you have data of that variable and one or more related variables(x1, x2, ... xk) that have some degree of linear correlation (R2> 0), and if the xi variables can be measured before you have to predict the y variable, then linear correlation is a useful tool.

But deciding to use linear regression without first having variables to regress on is like deciding to calculate a mean when you don't have a random variable, or deciding to use dish detergent before determining if you have any dirty dishes.
Yup, there are other cool statistic tools besides linear regression that could potentially fit the bill better (cool though it is as a starting point). All depending on the size and nature of the data available I guess.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•