Redfin publishes estimates on the number of homebuyers and sellers active in the U.S. housing market via reports on our news site and a dashboard on our data center. This post details the methodology behind our analysis.
The estimated number of sellers in the market is simply the number of active listings in the MLS. While there is not a similar metric for the number of buyers in the market, we are able to infer the number of active buyers because we know:
- How many sellers there are, based on MLS data
- The probability a seller matches with a buyer each month, based on MLS data
- The probability a buyer matches with a seller each month, based on proprietary Redfin data on buyer search time
Because we know how long the median buyer takes to find a home, we can estimate the probability the typical buyer will find a match each month using our matching model described below.
While this analysis incorporates Redfin data on buyer search time, it estimates the number of buyers and sellers in the market as a whole—not the number using Redfin as their brokerage or real estate platform.
Model
We adopt a standard matching model used in the economics literature on buyer and seller dynamics in housing, including Genesove and Han (2012), Anenberg and Ringo (2024), and Piazzesi et al. (2020). Our unique contribution to this work is leveraging Redfin’s proprietary data, which allows us to estimate the number of buyers at a much higher frequency and with estimates tailored to each region.
We adopt a Cobb-Douglas matching function with constant returns to scale. This is a standard modeling approach in economic models of matching markets and has been found to be a strong empirical fit to the housing market. (Badarinza et al., 2024). In this model, the number of successful matches that occur each month is a function of the number of active buyers and the number of active sellers.
The constant returns to scale assumption implies that a doubling of buyers and sellers would lead to a doubling of total sales. A consequence of this assumption is that the probability of a buyer or seller successfully matching with a counterparty depends only on the ratio of buyers and sellers.
In other words, the rate at which sellers find buyers is equal to the rate at which buyers find sellers times the ratio of buyers and sellers, or:
We treat these rates as hazards, or the probability that a buyer or seller finds a successful match within a given interval of time. Rearranging, the number of buyers is equal to the ratio of the seller hazard and the buyer hazard multiplied by the number of sellers:
Estimation
To estimate the number of buyers based on the final equation above, we need to identify three items:
- The seller hazard rate, or the probability a seller is matched to a buyer in a month.
- The buyer hazard rate, or the probability a buyer is matched to a seller in a month.
- The total number of sellers (active listings in the MLS).
The seller hazard rate is estimated as the share of active sellers whose homes go pending each month, or pending sales divided by active listings, based on MLS data.
The buyer hazard rate is estimated using a region-specific 12-month smoothed measure of the median time from first tour to closing, using Redfin’s proprietary customer tour data. Assuming an exponential survival model, the buyer hazard rate is equal to the natural log of 2 divided by the median time from first tour to close. We use the time from first tour to close as this corresponds with the period of most active search for buyers, in which buyer activity is likely to affect the probability of other buyers and sellers finding matches.
Finally, the total number of buyers is equal to the ratio of the seller hazard and the buyer hazard multiplied by the number of sellers.
Market tightness, or the ratio of buyers and sellers, is a common metric used in academic literature to assess the balance of bargaining power between buyers and sellers. (See, for example, Carrillo et al., 2015). We therefore define a market where sellers far outnumber buyers as a buyer’s market and a market where buyers far outnumber sellers as a seller’s market. A market where the gap is small is considered a balanced market.
Intuition
The matching model is based on a simple intuition about the housing market: that the amount of time it takes each side to find a match depends on how “tight” the market is. The tighter the market (more buyers per seller), the easier it is for sellers to find a match—but the harder it is for buyers to find a match.
Which means that if the number of sellers in the market goes up while the number of buyers stays the same, we should expect to see an increase in sales and:
- A decrease in the amount of time it takes the typical buyer to find a home
- An increase in the amount of time it takes the typical seller to sell their home
Similarly, if the number of buyers on the market goes up, while the number of sellers stays the same, we should expect to see an increase in sales and
- An increase in the amount of time it takes the typical buyer to find a home
- A decrease in the amount of time it takes the typical seller to sell their home
Our model allows us to use information on the amount of time buyers and sellers spend in the market—expressed as a probability of finding a match within a period—along with sales to back out the number of buyers and sellers.
Coffee-Shop Analogy
We can use a simple analogy to demonstrate the logic. Suppose you are standing outside a coffee shop and would like to know how many customers, or “buyers,” are currently inside.
With only two bits of information, you could estimate how many customers are inside without entering and manually counting them. These items are
- How long each customer spends, on average, inside the store
- How many customers successfully leave the store every hour
For example, suppose that each customer spends an average of 15 minutes in the store, and 40 customers leave the store every hour. Then you could infer—based on a well-known formula called Little’s Law—that there are 10 customers inside the store at a given time. (15 minutes per customer x 40 customers / 60 minutes = 10 customers.)
If an additional barista then came off their break and began making drinks, you would then likely observe a decrease in the average time each customer spends in the store. Consequently, you would expect that, if the total number of customers served per hour remained the same, the average number of customers inside the store at a given time is less than before.
Alternatively, if a surge of customers entered the store, you would expect to observe an increase in the average time each customer spends in the store. Consequently, you could infer that the average number of customers inside the store at a given time is greater than before.
Our method for inferring the number of buyers relies on a similar logic. Unlike the coffee shop, the housing market consists of both sellers and buyers looking for a match. But much like the coffee shop, the amount of time it takes each side to find a match depends on how “busy” the shop is. The more buyers there are for each seller, the longer it takes for a given buyer to find a match. Our model of the housing market also accounts for how sellers compete with each other—with more sellers available, the buyers have an easier time but the sellers are more likely to slow each other down.
Correlation with Home Prices
History has shown that a change in the balance of buyers and sellers is often a signal of what’s to come with home prices; home prices tend to cool when sellers increasingly outnumber buyers, and heat up when buyers increasingly outnumber sellers:
References
Anenberg, Elliot and Daniel Ringo, “Volatility in Home Sales and Prices: Supply or Demand?” Journal of Urban Economics, Vol. 139, January 2024.
Badarinza, Cristian and Balasubramaniam, Vimal and Ramadorai, Tarun, “In Search of the Matching Function in the Housing Market,” SSRN, June 2024.
Carrillo, Paul E., Eric R. de Wit, and William Larson, “Can Tightness in the Housing Market Help Predict Subsequent Home Price Appreciation? Evidence from the United States and the Netherlands,” Real Estate Economics, Vol. 43, No. 3, January 2015, pp. 609-651.
Genesove, David and Lu Han, “Search and Matching in the Housing Market,” Journal of Urban Economics, Vol. 72, No. 1, July 2012, pp. 21-45.
Piazzesi, Monika, Martin Schneider, and Johannes Stroebel, “Segmented Housing Search,” American Economic Review, 2020, Vol. 110, No. 3, pp. 720–59.
