Federal Reserve Economic Data

There is a lot of talk about how wild the housing market has been, including houses sold on the day of listing without any inspection and for much more than the asking price. Can we see evidence of this in the data?

This housing fervor should be reflected in the number of days that houses are on the market, which is what our FRED graph shows. We sampled three housing markets generally considered to be hot—San Francisco, Denver, and Austin—but the story is similar elsewhere. While the median days on the market are lower than usual, they’re not dramatically lower and certainly don’t show that most houses are sold unusually quickly. In fact, the typical seasonal pattern seems to persist.

Before trying to understand what’s really going on, we need to understand what the data are actually measuring. There are two ways to compute median days on the market:

1. Take all transactions in a month, look at how long the houses were on the market, take the median.
2. Take a snapshot on a particular day of the month, look at how long the houses currently for sale have been on the market, take the median.*

The first method takes into account all the quick sales, which is the methodology for the data shown in the graph. Here’s how the source, Realtor.com, defines the data:

The median number of days property listings spend on the market within the specified geography during the specified month. Time spent on the market is defined as the time between the initial listing of a property and either its closing date or the date it is taken off the market.

If this method includes the quick sales, why are the reported medians still high? First, the stories of quick sales may be for particular submarkets and not reflective of the overall market. Indeed, their could be a mismatch between what is offered and what is demanded. Second, there may be a cognitive bias at work that is similar to the frequency illusion: If many people are interested in a few houses, even if few or no people are interested in other houses, the perception is still that all houses are hot. A similar bias occurs on highways: They’re congested during rush hour, when many people are on them, but empty otherwise. So people think they’re always congested, whereas they’re actually almost always empty. Third, there may simply be a minimum number of days needed to close a sale and, thus, the data cannot go below that number.

* This second data methodology misses most of the quick sales, but underreports slow sales, which haven’t yet been concluded. So it may actually report a lower number of days on the market.

How this graph was created: Search FRED for “house days on market San Francisco” and make sure to take the monthly series in days. From the “Edit Graph” panel, open the “Add Line” tab and search for the Denver and then Austin series.

Suggested by Christian Zimmermann.

The FRED Blog has discussed interest rates before, including those used as benchmarks of overnight borrowing costs in financial markets. Today, we revisit the topic of overnight financial transactions by comparing the interest rates of two types of loans: secured and unsecured.

The FRED graph above shows two different interest rates paid by financial institutions for borrowing cash at the end of the business day and paying it back at the start of the next business day:

• The blue line shows the overnight unsecured AMERIBOR benchmark interest rate. Reported by the American Financial Exchange, this is a volume-weighted average of interest rates applied to transactions where the borrower does not offer a security as collateral for repayment.
• The red line shows the secured overnight financing rate (SOFR). Reported by the Federal Reserve Bank of New York, this is a volume-weighted median of interest rates applied to transactions where the borrower offers Treasury securities as collateral for repayment.

If you look closely on any given date (we suggest zooming in by clicking and dragging on the graph itself), you will notice that the interest rates are very similar. Despite the fact that overnight loans are paid back very quickly, in less than 24 hours, secured transactions generally record lower interest rates than unsecured transactions: The collateral offered in secured borrowing reduces the amount of potential losses associated with lending, making it cheaper for both lender and borrower.

However, there are days—or even weeks—when unsecured borrowing is cheaper than secured borrowing. You can see here a FRED graph showing the difference between the AMERIBOR and SOFR interest rates. The largest spike of the SOFR series, on September 17, 2019, provides an example. On that date, as described by Sriya Anbil, Alyssa Anderson, and Zeynep Senyuz, a momentaneous shortage in liquidity resulted in a momentous increase in secured borrowing costs and a minimal increase in unsecured borrowing costs. The New York Fed quickly intervened to address this acute need for liquidity, and interest rates promptly returned to their pre-shortage levels.

How this graph was created: Search for and select “Overnight Unsecured AMERIBOR Benchmark Interest Rate.” From the “Edit Graph” panel, use the “Add Line” tab to search for and select “Secured Overnight Financing Rate.”

Suggested by Diego Mendez-Carbajo.

The unemployment rate is the highest-profile labor market data point, but there are plenty of other ways to gain additional insight into the job market. One such figure is the ratio of unemployed workers to job openings. It’s a straightforward statistic created by combining two key BLS series: unemployment level (via the Current Population Survey) and total nonfarm job openings (via the JOLTS survey).

Exits from the labor force during the COVID-19 pandemic have been analyzed, with research conducted into the reasons for exits and the likelihood of re-entry. It’s led to some debate as to the true tightness of the labor market— after all, total nonfarm payrolls are still 2 million below January 2020 numbers. This ratio sidesteps all that, giving us a measure of labor tightness that reflects firms’ attempts to hire at the present moment.

After reaching a high of 6.5 unemployed workers for every job opening in July 2009, the ratio fell over the next decade to 0.8 in February 2020. It rose to 4.9 unemployed per job opening during the COVID-19 recession, but has since fallen steadily as economic conditions have improved. As of January 2022, it reached the lowest ratio since the BLS began collecting JOLTS data in 2000: just 60 unemployed workers for every 100 job openings. This aligns with reports of widespread labor shortages across several industries; when looking only at the current labor force, there just aren’t enough job seekers for the current level of positions open.

How this graph was created: Search FRED for “Job Openings: Total Nonfarm” (JTSJOL). Set level in thousands, seasonally adjusted. Combine with “Unemployment Level,” thousands of persons, seasonally adjusted (UNEMPLOY). Divide UNEMPLOY by JTSJOL (formula b/a).

Suggested by Nathan Jefferson.