In London, rent contracts are typically one month or 6 months rolling. When we want to look at the short term demand for housing in Central London, the rental market is more appropriate than house prices for sale.
Lets look at how rents have changed this year. My data comes from Zoopla listings until the end of October 2020.
Dataset
In 2020 there were 110k listings for rental properties within London travel zone 1 - 3. My historic data is somewhat limited and I only have 31k listings per year between 2015 - 2019.
Average Rents
By Zone
The chart below shows the average rent of advertised properties in London Zones 1, 2 and 3. The orange line is the 2020 average per month, whereas the green line is the 2015-2019 average.
Since WFH (work from home) started rents have fallen in Zone 1 and Zone 2 but not in zone 3. Before WFH, rents in Zone 1 and 2 were around 25% higher in Zone 1 and Zone 2, but since then they have fallen 10 - 25% below the 2015-2019 average. That is around a 30% contraction in advertised rent prices.
Zone 3 prices have not changed much at all.
All of the above is expected given that many people no longer need to live in Central London for work purposes.
Note that the mix of properties changes every month (and every year), so any trends in prices could just be reflecting that instead of actual change in prices.
Zone 1 by postcode
Since Zone 1 prices have fallen the most, I broke down the changes by post outcode. Ie. EC1X XXX becomes EC1. If there were not enough listings in a postcode I merge the postcode into the postcode district (Eg. EC). To simplify the charts I plot the percent change in rents in 2020 each month from the 2015-2019 average that month.
The postcodes where rents have fallen the most are shown first, going right and then down.
It appears West London and South West London have suffered the most, with average rents falling by almost 50%! East London seems to have fared a little better.
Rental Price Index
For each of the 16k properties (in Zone 1) I have rental price history over the years. See below the rental price histories for 8 random properties.
To compare the rents across properties I normalise the price by dividing the price at listing by the average price over all listings.
Based on this sample of 8 properties, the rental prices tend to stay roughly flat since 2010.
We could do these for each property in our dataset but it quickly gets messy. See below with just 200 properties.
So instead, we will extract the price index from a linear model which uses the following regression to predict rent.
$$ log(rent) = \beta_1 PropertyID + \beta_2Outcode + \beta_3Year + \beta_4Month + \beta_5Outcode:Year$$
Note: outcode is the first portion of the postcode. eg for EC1 XXX it will be EC1.
In essence, we can model the rent on a particular date as the property’s average rent plus the general postcode uplift, plus the effect of year/month of sale. We have an interaction between outcode and sale year so that we can extract average rent changes over times in each outcode.
The price change for each outcode since 2010 can be seen below.
