For this particular example (housing) I agree that median isn't that informative. It doesn't cover the range of prices available and a theoretical city with a high median, but a glut of housing available cheaply, would look expensive according to this stat.
However, this is not my impression of housing in NYC at all. In fact - quite the opposite. The very low vacancy rate means that the floor is very high. That is why you see stories about essentially closets going for more than luxury apartments in other cities.
The point of the median is that it's an actual existing price, and that it filters out extreme outliers.
If you add a datapoint where somebody is renting a castle for $100K/month, that skews the average way up. Similarly, the existence of a few people being paid $1M/month can make it look like the area is a lot wealthier than it is, because the average is pulled up.
The median is useful because it gives you an actual datapoint in the middle. It's an actual price that exists.
Of course, the devil is in the details, and there are situations in which the result is not entirely satisfactory. But on the whole is an useful measure especially when there's more than 5 datapoints.
Of course any kind of statistics is a simplification of the actual data, so sometimes you have to dig into the details. But often times we want something like checking 10 potential cities to see which place is desirable to stay in -- it doesn't matter if things are not entirely right so long they're right enough to quickly produce a reasonable answer.
2 comments
[ 3.8 ms ] story [ 17.1 ms ] threadHowever, this is not my impression of housing in NYC at all. In fact - quite the opposite. The very low vacancy rate means that the floor is very high. That is why you see stories about essentially closets going for more than luxury apartments in other cities.
If you add a datapoint where somebody is renting a castle for $100K/month, that skews the average way up. Similarly, the existence of a few people being paid $1M/month can make it look like the area is a lot wealthier than it is, because the average is pulled up.
The median is useful because it gives you an actual datapoint in the middle. It's an actual price that exists.
Of course, the devil is in the details, and there are situations in which the result is not entirely satisfactory. But on the whole is an useful measure especially when there's more than 5 datapoints.
Of course any kind of statistics is a simplification of the actual data, so sometimes you have to dig into the details. But often times we want something like checking 10 potential cities to see which place is desirable to stay in -- it doesn't matter if things are not entirely right so long they're right enough to quickly produce a reasonable answer.