Repo link

Some Notebooks

Data (1,2,3)

TLDR: readily available home sales data allows us to prod at a variety of real estate “folk wisdom” within very specific markets.

Self indulgent introduction

Those lucky and unlucky enough to be attempting to purchase homes right now have likely felt like they were “getting shot in the face with money.” There's no shortage of broad analysis and guidance about housing (e.g. this), but this advice is generally unspecific to an industry that's notoriously location specific.

This “gap” in guidance is historically closed by real estate agents and the experientially-grounded opinions they are able to provide. Some things you might have heard:

It's a mistake to buy a less desirable house — on a main road, near train tracks, etc — in a hot market. You'll have a harder time selling it in a cold market, and its value will be more susceptible in such downturns.
You should seek the least expensive house in an upward trending neighborhood, as it has the most room for growth and improvement.
Alternatively, you don't want to buy either the most expensive or the least expensive house in a neighborhood. Neither will appreciate over time as much as a closer-to-median house.
You can forecast appreciation in a neighborhood by observing the frequency of teardowns/new constructions.

These are ultimately hypotheses, albeit backed by experience and the aggregated knowledge of the realtor community.

We’ll share a straightforward way of softly “validating” some of these hyper-local hypotheses using publicly available data. The answers have wide error bars, but are still actionable.

Specific hypotheses

To highlight the hyper-locality of this approach, let’s consider a set of hypotheses focused on a few insanely pricey neighborhoods in the SF south bay area.

H1	Houses close to CalTrain tracks in Palo Alto underperform.
H2	Newly built houses in The Willows neighborhood of Menlo Park appreciate slower than older houses.
H3	Houses within a block of highway 101 in The Willows appreciate slower than those further away.
H4	Houses on a main street (Middlefield) appreciate more slowly than those on more “desirable” side streets in Palo Alto.

All of these hypotheses involve some sort of “affected” location or house attribute — oldness, closeness to a highway / main street / train tracks / etc. In general, one would expect these properties to be more volatile — a great deal in an economic downturn, and inflated in value during an economic upturn.

Things have obviously trended upwards since the 2008 bubble, so on that basis we’d generally expect affected locations to appreciate faster. But how has that played out over the last 5-15 years of real estate transactions in these specific neighborhoods?

Self indulgent introduction

Specific hypotheses

Methodology