In what is destined to become a series, the next set of data we’ve decided to forecast is the infamous median San Francisco rent price. We used historical median rent data from Zillow, which you can find here if you’d like to play with the same data. If you’re not in the bay area, here’s a recap: rent is very expensive here. The median rent price in November according to Zillow was $3400.
So How Do We Do It?
Our forecast utilizes ARIMA and Seasonal Trend Decomposition via LOWESS (STL). We have covered this method here in greater detail, but essentially, we are deconstructing time series data into an overall trend component and an oscillating seasonal component. We forecast future points in the time series using auto-regression and/or moving averages.
What We Found
Long story short – historical rent data mathematically suggests that rent has found a sweet spot between $3000 and $3500, and it should remain in that range for the next twelve months. In fact, we’re forecasting that San Francisco rent will hit a median low of around $2800 this summer, a price we haven’t seen in nearly two years.
Sources of Variation
There are some significant factors that account for the variation in the median rental price in this data set, and we want to be clear that this is in no way incredibly “predictive.”
- Mix of Available Unit Types
This dataset includes all rental listings regardless of number of bedrooms, square footage, etc. If the number of studio apartments increases compared to the number of three bedroom apartments, the median price could move downward drastically.
- Variation in Listing Volume by Neighborhood
Zillow provides data at the neighborhood level. When listing volume in more expensive neighborhoods like Pac Heights or Nob Hill makes up a higher percentage of total units listed, the median rent value is going to increase. Conversely, a high proportion of listings in areas like the Tenderloin are going to bring the median down.
- Seasonal Trends
The historical data suggests that median rent typically drops from March to April each year and then climbs steadily through the fall months. While our data forecasting methods account for this, it’s difficult to be completely accurate with a relatively small number of data points (47 months).
Although there has been historical increase in rent prices in San Francisco, this forecast suggests that further increases are starting to slow, and we should see lower prices in the summer of 2014. The market appears closer to a price point where supply and demand are more balanced.
This is a strictly mathematical forecast – there are infinite social and economic factors when attempting to predict something like rent price, and while we hope to be able to build some of those in to a future model, creating a truly accurate prediction is nearly impossible.
Thanks to Scott Yacko for yet again doing most of the work on this experiment.
Subscribe to our blog.
We'll occasionally send you an email when we have featured posts and great stories about analytics and data forecasting.
If you liked this post, you'll enjoy using Levers.
Rent isn't the only thing you can forecast with Levers. We provide simulation software for marketing analytics, ecommerce, lead generation, social media and more. Learn more about Levers…