Nearly a century ago, economist and Maryland congressman Clarence Long called the construction industry “probably the single most strategic factor in making or breaking booms and busts.” Since then, economists have tried to unravel the complex relationship between the housing market and the economy.‌

A flurry of housing activity often precedes economic downturns, from the Roaring 20s to the 2008 global financial crisis, notes Cameron LaPoint, assistant professor of finance at Yale SOM. Property prices and sales volume also often reach a climax a few months before the stock market peaks in economic expansions. But despite these links, understanding the relationship has been hampered by the lack of comprehensive, long-term data on local housing market activity. LaPoint set out to solve that problem, for himself and his fellow economists.‌

In a new paper, LaPoint and co-author Gustavo Cortes of the University of Florida dig into the links between local housing markets and financial conditions by zeroing in on residential building permits. These conditions are important to study, the researchers say, because they reflect the ability of a developer to build at a later date. For the study, LaPoint and Cortes constructed an entirely new dataset of US historical local building permits in all 50 states and 60 major cities from 1919 to 2019. They found that periods of rapid increases in building permits followed by rapid declines predict subsequent booms and busts in both stock- and bond markets.‌

“Building permits give you a good signal about investor confidence in the local economy,” says LaPoint. “They reflect one’s thoughts that it is valuable to have the opportunity to build.” Across 100 years of data and 20 recessionary periods, these movements in local building permit activity consistently predict fluctuations in the stock and bond markets. “It’s a connection between Main Street and Wall Street,” he says.‌

To build their dataset, the researchers spent two years hand-collecting and digitizing the data using deep learning techniques for optical character recognition. They obtained historical building permit data from 1919 to 1957 from company publications Dun’s Statistical Reviewadjusts the settings of the scanning software so that it can accurately read yellowed, vintage periodical pages. To fix inevitable scanning errors, they wrote custom code in Excel. “There’s a lot of trial and error in this process,” LaPoint says. ‌

For permits from 1960 to 1987, the researchers pulled figures from the US Census Bureau’s Building Permit Survey (BPS), which includes paper reports from city and county officials on the number of permits approved. The researchers focused on single-family home permits, because builders and developers tend to execute them with a higher degree of probability, LaPoint said. “They are a more accurate signal of investor activity,” he says.‌

Inevitably, reports had disappeared. So the researchers turned to the HathiTrust Digital Library, a nonprofit cooperative of libraries that preserves millions of digitized items, to find books with the missing records. Two years ago during Thanksgiving, LaPoint spent hours at her childhood desk, clicking on digitized books and waiting for thousands of pages to download. The researchers also tracked down numbers through the Federal Depository Library Program, a government program that collects and organizes federal data. For data from 1988 to 2023, they downloaded state numbers from the US Census Bureau’s website. ‌

Then LaPoint and Cortes compared state fluctuations to stock and bond market volatility, as measured by daily returns. Their hypothesis about the predictive power of conditions held true for historical periods when mortgage loans were scarce, as well as in modern times, when credit is more abundant. ‌

Building permit activity was predictably more volatile in southern U.S. cities and states that have more elastic housing supply — that is, places where developers have more flexibility to build when demand rises and pull back when it falls. Often in these places land is more available, and the rules are not as strict; For example, Atlanta is more elastic than San Francisco. ‌

LaPoint and Cortes found that when building permit activity in Florida became 10% more volatile, stock market volatility increased by 0.2 percentage points and bond market volatility by 0.7 percentage points one year later. This predictive power was not found in heavily regulated states like Connecticut. “Places like Florida have been more lax in their regulatory restrictions on new housing,” says LaPoint.‌

In the data, they could see that in the years before the Great Depression, total permits issued in Miami peaked at $65.7 million in 1925 during a Florida land boom facilitated by railroad expansion, then dropped to $1.2 million on the eve of The 1928 Okeechobee Hurricane. Florida construction permits also peaked five months before the OPEC boom in 1973; another peak led to the Great Recession of 2007–2009 by nearly two years. Las Vegas, another city with an elastic housing supply, also saw declines in 2006.​​

Beginning in 1989, researchers were also able to examine the relationship between local permit activity and movements in the stock prices of individual companies, including Coca-Cola and Home Depot. Based on the companies’ facility locations, they calculated how exposed operations were to fluctuations in permit operations. They found that companies with operations in areas experiencing volatile building permit activity saw more volatility in their own share prices. “It’s like the tremors of the local real estate market spread to businesses that operate in that area,” says LaPoint.​​

The connection means investors can use building permit increases to craft trading strategies that hedge against corporate exposure to real estate markets facing a glut of new housing development, LaPoint said. “It suggests that companies are differently exposed to overbuilding and that building permits can almost act as a risk factor for companies included in the S&P 500.” ‌

Currently, LaPoint and Cortes are working to fill a gap in their data set from 1957 to 1959, when Dun’s stopped publishing, and before the US Census Bureau took over data collection. Once their data set is complete, they plan to post the raw data online. In the meantime, they are building a website that will include interactive graphs of key data. ‌

The research has far-reaching implications, LaPoint says, for investors, housing policy and even the housing affordability crisis. Today, state-to-population ratios are at historic lows in highly regulated states, especially in New England. Data patterns can help consumers predict when prices might peak so they can judge the best time to buy.​​

“When we look at these 100 years of data, it’s worth thinking about, how are we going to go back to where we can meet the demand from the population, if the supply we have now is really weak compared to before. We are trying to translate our measure of building permits into a more direct measure of housing supply over time.” Future research could also predict stock returns, he says, giving investors another forecasting tool.‌

For financial regulators, volatile building permit activity in some regions, and particularly in the south, can act as an early warning of broader financial market instability. “This can help regulators identify when and where to focus attention to limit potential developer enthusiasm and overbuilding risks,” says LaPoint.​​