What Can Alternative Data Tell Us About the US Housing Market?


Key Takeaways

  • MacroX’s real estate model predicted the positive surprises in the housing market in February
  • We continue to anticipate near-term price declines in home prices but an increase in residential construction activity
  • Student-loan forgiveness is an important topic driving the real estate market in the medium-term

Activity in the US housing market rebounded last month

The US Housing market experienced a rebound last month as a number of housing indicators experienced strong month-on-month growth. Housing starts, building permits, and existing home sales rose more in February than real-estate analysts expected. Our real-estate model which uses social data, however, correctly predicted this positive surprise and it did so in December! This is due to social data predating actions in the construction life-cycle enabling accurate predictions of the real-estate market months in advance. We are also bullish on pending home sales for February for which data will be released next week.

Figure 1: MacroX’s alternative data and AI-powered real estate models have been months ahead and accurate.

Our outlook on house prices

We continue to think that US house prices will decline slightly this year. Our real-time estimate of housing demand shows that levels have moderated from their post-pandemic highs as historically high mortgage rates bite. Still, as last month’s US aggregate (dotted line) demand ticked up a bit, the question is whether this moderation will continue.

Figure 2: MacroX’s housing demand nowcast and overall aggregate demand nowcast show a recent divergence.

US housing inventory shortages have also softened providing further scope for a moderation in house price growth. Indeed the latest National Association of Realtors data agrees with us and shows a 0.2% yoy decrease in median house prices.

Figure 3: Housing inventories across the US have ticked up recently.

Why do social data make us cautiously optimistic on residential activity?

In spite of the pressure on prices, our social-data (twitter, search, news etc.) based models are picking up on increased residential activity such as a dramatic rise in home inspection related social data, but not much increase in new construction or construction loans. Taken together these patterns suggest that the uptick in permits is either people preparing their homes for listing or affluent people substituting away from buying a new home to improving their current home.

Figure 4: Home inspection activity has been rising dramatically while construction and borrowing activity is flat.

Medium-term real estate themes

By crunching through news, social media and the blogosphere we are able to identify themes driving the real-estate market. Over the past 18 months, student loan forgiveness seems to have risen in importance and is an important topic to follow when forming a medium-term view for the sector.

Figure 5: Student loan forgiveness is an important medium-term topic since it affects, the debt capacity and credit scores of potential home buyers.


We construct a “user journey” for the construction life-cycle – starting from motivation such as a new-born or marriage, to getting the permit, and finally the inspection. A deep understanding of this user journey gives us a more granular picture of real-estate activity and motivations. Fun fact – a rise in social activity for termite inspections does not necessarily indicate a termite outbreak, but actually an indicator of the imminent inspection activity!

Stay Tuned

We will continue to update our models and also do a deep-dive on the longer-term trends shaping real-estate such as the digital economy, remote work, and the “rising star” cities.

Data sources – US census bureau, FRED, national association of realtors (NAR), Realtor.com, MacroX social models ingest twitter, search, facebook, postings, news, and blog data.

Sign up for our newsletter