For the rest of us, as we wade through the tail end of the pandemic, witness gravity-defying labor markets, experience inflation rates not seen in four decades, and see unprecedented interest rate hikes, some clarity about the housing market would be helpful.
Perhaps, even about our local market?
Opinions and math
It turns out that creating timely and actionable insights about local housing markets is not hard to do. Just ask real estate agents for their opinion. Well, really, lots of them and do a little math with all the responses.
It makes intuitive sense. Real estate agents are out there daily, seeing, listening, observing, and transacting. Their opinion is ground truth.
All we need to do is collect their opinion and do that math. HousingIQ did that.
Ears to the ground
We analyzed data from the HousingIQ Survey of Kentucky REALTORS®. To gauge what survey responses can tell us about future house price trends, we combined five survey questions concerning house prices into a Sentiment Index and compared it with the Freddie Mac House Price Index (FMHPI) for Kentucky.
And the result? Real estate agents can foretell changes in house price appreciation two months in advance. Just five questions posed to a group of practitioners are sufficient to predict the future trajectory of house prices.
The monthly surveys were conducted between February 2020 and July 2022. A thirty-month period that started before the lockdowns; included a remarkable recovery and unprecedented interest rate hikes. And through this all, the survey could foretell house prices.
The Sentiment Index has been trending down since March 2022. The forecast for August is a month-over-month decrease in Kentucky house price appreciation of -0.74% +/- .22%. The forecast for September is a decrease of -0.79% +/- .26%. As the August numbers trickle in, we can confirm that the Sentiment Index accurately forecasts price changes.
The wisdom of the crowd can create housing market intelligence. With merely five questions, the collective wisdom of real estate agents can be harnessed to predict the future trajectory of house prices.
Since house price indices are released with a delay of a couple months, survey-based insights provide more timely intelligence. By targeting surveys to specific regions and market tiers, we can create actionable insights unavailable via published house price indices.
The monthly survey includes five forced-choice questions concerning respondent opinion about the direction of house prices, price cutting by sellers, sales below the asking price, days on market, and foot traffic. Responses are one of Increase, Not change, or Decrease.
We converted responses into a sentiment value using factor analysis. Each month’s Sentiment Index is the average of the individual sentiment values.
We compared the Sentiment Index with the Freddie Mac House Price Index (FMHPI) for Kentucky.
The Sentiment Index leads price changes.
The cross-correlation of sentiment and month-over-month changes in house price appreciation shows that the Sentiment Index leads by one, two, and three months. Remarkably, sentiment does not appear to respond to house price changes, as evidenced by the Sentiment Index not being a lagging indicator.
The Sentiment Index provides new information about house price changes not contained in the house price index.
House prices are persistent. Price changes in one period are strongly influenced by changes in past periods. A Granger causality test shows that incorporating the prior two month’s Sentiment Index values improves forecasts of house price changes.
The Sentiment Index can independently predict future changes in house prices.
A regression model based on only the 1 and 2-month lags of the Sentiment Index and not including lags of house prices has an adjusted R-squared of 80% and a p-value of 9.0e-9.
By the numbers
The survey was conducted between February 2020 and July 2022. The median number of responses each month was 265.
Cronbach’s Alpha = 0.77
Kaiser-Meyer-Olkin (KMO) Index (Overall) = 0.8
Bartlett’s Test For Sphericity Chi-square test statistic=10,273
Measure of Fit
Tucker Lewis Index = 0.95
RMSEA index = 0.08
Order = 2, F test statistic = 16.4, p-value = 3.7e-05