Who’s Talking Now?

The Days Ahead: Jobs report, job openings, and durable goods  

This Week

• People don’t want to do surveys.
• Response rates are falling.
• Some of the economic data may be late or incomplete.
• It may take longer to know what’s really going on in the economy.
• Meanwhile we should view monthly data as very noisy.
• The Fed paid banks very well during the last two years.
• That helped banks’ profits and margins.
• But the era of high payments to banks is over.

Who’s Talking Now?

A few months ago, the University of Michigan Consumer Confidence Survey (UMich) changed its survey system from phone calls, which it had used since 1946, to web-based questions. The UMich researchers typically called around 600 household to gauge consumer sentiment for things like inflation expectations, home values,  job prospects, and buying conditions for houses or expensive household durables like washing machines.

The Federal Reserve (Fed) uses the inflation expectations part of the survey as an important input for its inflation outlook. The Fed’s concern was always that if consumers felt inflation was going to run high, they would cut back on purchases or demand higher wages and an inflation-wage cycle could kick off quickly.

The UMich economists started a partial switch from telephone to web responses over four months ago. By July, the survey was 100% web based and the number of respondents increased to 1,000. Why the change? Cost. The interviewers had to call 90 separate phone numbers for every one completed survey and the questions took over three hours to complete. Now, three hours is a big  ask and, with the best will in the world, I’m not sure I would want to spend the time.

However, the longer time requirement had advantages. Interviewers were good at keeping people on track and making sure they understood the question. For example, ex-Fed economist and inventor of the Sahm Rule, Claudia Sahm, recalled an interviewer asking,

“Do you expect…?”

and the response came back,

“Well, I hope….”

The interviewer kept correcting,

“It’s not about hope. What do you expect?”

No wonder it took three hours, but the data was solid.

The UMich analysts were good enough to run a side-by-side comparison with the phone and web-based responses and found a 0.8 correlation, meaning the responses overlapped 80% of the time. That doesn’t sound too bad but it degraded the data. Ideally, how and where you ask the same question to the same person should not create differing responses at all.

But it did.

What they found was that web-based responses were consistently gloomier than the phone respondents. For example, since 2017, between 22% to 29% of phone respondents expected their incomes to “stay the same” in the subsequent year. For the web-based group it was 18% to 62% with the average decidedly less optimistic than the phone-based group. Similarly, the web-based group expected higher current and future inflation than the phone interview group.

The result is that in the last four months consumer sentiment from the UMich surveys has been surprisingly low. At a time when headline inflation fell from 3.5% to 2.5%, survey respondents expected inflation for one year ahead to rise from 2.9% to 3.1% and for five years ahead to rise from 2.8% to 3.0%. In the prior three months before the full conversion from the phone survey, inflation expectations fell every month. In July, a quarter of respondents saw inflation accelerating to 7% in the next year compared to expectations of 4.5% inflation back in March 2024.

It was the same on the job front. Headline unemployment ticked up from 3.8% to 4.2% in the three months to August, but much of that unemployment was because of new entrants to the workforce. The level of job losses was 76,000 lower in August than April and the number of weekly claims fell from a four-week average of 230,000 to 227,000. But the UMich survey showed people were more fearful of losing their job than back in April and saw the probability of losing their job in the next five years rise from 17% to 20%.

Much of the difference in the two surveys is because people are just more upbeat when they talk to another person and more downbeat when responding surveys. That means we must look at the UMich survey, especially for the next year or so, as more somber and gloomier than normal.

This exercise got us looking at other surveys we rely on. The Bureau of Labor Statistics (BLS) has pointed out for a while that response rates for surveys on things like inflation, employment, and job openings have steadily fallen for years. Here’s the full list but the chart below shows three of the most important surveys.

For the inflation data, which samples 138,000 price quotes, response rates fell from 80% to 67% during Covid-19 and have only recovered to 70%. The peak was over 90% eight years ago. The employment survey and the job openings survey, which sample 651,000 businesses and 21,000 businesses respectively, saw response rates fall from 60% to less than 30% in the last five years from a peak of nearly 80%.

In the case of the two business surveys, responses increased in the follow-up months, but that just means we see ever-bigger revisions to the data, usually to the downside. That’s because a business with declining employees or openings is facing some sort of crunch and will push a government survey to one side for as long as it can. Who can blame them?

What’s going on?

• People don’t like surveys. As in the change in the UMich surveys, fewer people have the time for lengthy questionnaires.

• Survey Fatigue. By the time you’ve responded to “how was your Amazon delivery of dental floss” or “please rate your stay at the Scranton Airport Radisson,” the next well-meaning survey from the BLS may leave you exhausted.

• Privacy concerns. Data breaches, once rare, are all too frequent. However safe the source, people are more concerned about data theft, especially when asked about employment and job prospects.

• Answering the phone. The UMich survey only switched from landlines to a mix of cellular and land lines in 2014. The number of landlines in the U.S. peaked at 192 million in 2001 and is now less than 68 million. Meanwhile the number of cell phones grew from 290 million to 360 million.
  • Half of all households aged over 65 have a landline but 90% of adults under 40 only use cell phones. People with cell phones are just not as willing to pick up calls from numbers they don’t recognize. There are 33 million spam calls made every day and 60 million Americans lost $29 billion to scam callers in 2022.
• Survey designs. While the BLS uses a robust and insightful survey model, most surveys are generic, badly designed and free to use. That’s why we see so many surveys every day. But it also means surveys often give misleading information.

What does all this mean for the economic data?

First, revisions will be greater. We now expect data like new jobs, GDP, unemployment, and the number of employees to undergo several revisions from first to final print.

• Between 1999 and 2022,  initial average quarterly GDP growth was 0.4%. Through six subsequent revisions, it was 0.6%.

• In the same period, personal spending started at an average of 1.1% and ended up at 0.8%

• From 2013 to 2022, employee compensation started at 3.4% and ended at 0.4%.

• In its annual revision of the number of people employed, the BLS revised down the size of the workforce by 818,000 or 0.5%. The average revision for the last 10 years was less than 0.1%

Markets react to the initial print and sometimes to the second. By the time the final calculation is ready, relevance plummets.

Second, the BLS is changing its data collection methods. It started using video data collection during Covid-19 and found it was more accurate than traditional phone and paper surveys. The BLS also uses web-scraping tools and secondary source data, which means gathering data from more than a one-on-one data enquiry. The BLS gasoline price index, for example, now uses daily price observation to capture six million price points a month. The former survey was monthly with only 4,000 price quotes.

There’s nothing misleading or inaccurate about the BLS data and we don’t subscribe to the view that “falling survey rates undermine economic data.” But the data may take longer to gather while the BLS improves its methods.

It’s another reason we tend to look at trends and longer time series. The monthly data can be all too noisy.

Bank Reserves

We wrote about the new era of bank reserves last year when the Silicon Valley Bank crisis erupted and again in January of this year when the Fed started to change the way it was running off assets bought during the second quantitative easing (QE2) period in 2020.

The short version is that when the Fed started QE2 in early 2020, they bought Treasuries and mortgage-backed securities from banks. This increased bank liquidity from $1,500 billion to $4,250 billion by mid-2021. Banks deposited the cash back with the Fed.

Banks are in the business of making loans. When a bank makes a loan it credits the borrower with money, which reduces its cash holdings, and receives a loan asset in return. The interest on the loan will pay more than cash held at the Fed. But in 2022, there wasn’t enough loan demand in the economy to absorb the large cash balances.

Banks had a similar problem in 2008 when QE first started. Back then, the Fed paid interest on required reserves at a rate of fed funds less 0.1% and on excess balances (interest on excess reserves, or IOER) at a rate of fed funds less 0.35%. Banks had little incentive to hold excess balances but, as in 2020, there was little or no loan demand. So, the Fed then started an “abundant” reserve system, which did not require banks to hold minimum reserve levels. It also started to pay interest on all balances held at the Fed, and gave up splitting them into required and excess reserves.

This system worked during the Fed’s balance sheet run-off period from 2010 to 2020 but when QE2 started, bank reserves shot up again and loan demand collapsed by 28%. Banks were again sitting on huge cash reserves yielding zero putting a large dent in profits. In March 2020, the Fed scrapped the concept of minimum reserves and started to pay Interest on reserve balances, or IORB, at a discretionary rate but which has been fed funds plus 0.5%. In the zero-rate environment, this wasn’t a lot of money but as the Fed increased rates from 2022 to mid-2023, the numbers became very large.

Here’s a very simple model of what happened.

In the graph below we show the banks’ reserve balances in blue. There’s a clear jump to $4,250 billion in the QE2 period. The IORB started in 2020 at only 0.17% but ratcheted up to 5.4% in mid-2023 where it stayed until last week, when it fell to 4.9%. The IORB was set at a 0.25% premium to the fed funds rate and remained well above short-term bills. It was, in effect, a very nice subsidy for banks to keep reserves at the Fed and not have to worry about credit or duration risk.

How much did banks earn?

We show the old IOER in the black line. From 2010 to 2020, we calculated the Fed paid out a total of around $633 billion. Since the switch IORB in 2021, shown in the green line, the Fed paid out a total of around $1,500 billion. The $167 billion is the current annualized rate, after falling from $182 billion following last week’s rate cut.

What does this mean for banks? Well, if we take the $1,500 billion earned by banks for depositing their funds at the Fed and subtract the average interest paid to customers of 0.4%, banks enjoyed a windfall of around $1,200 billion from 2022 to now.

At the quarter ending in June 2024, banks had raised the average deposit rate to 2.2% but that was still way below the 5.5% earned at the Fed.

The pattern of banks capturing the upside of high rates is not new. Most people experience it when reading about rates increasing but see little change in bank deposit rates. This time, banks enjoyed especially strong earnings because of the Fed’s special rate. It doesn’t mean an end to bank earnings growth. As rates fall, banks should see more loan and mortgage demand and refinancings. But the era of high rates paid by the Fed may be over.

Bottom Line

After last week’s big Fed news, this week was quieter. Economic data was thin on the ground, but what there was positive. The Purchasing Manager’s Index was unchanged at 54. Anything above 50 is “expansionary” and it hasn’t dipped below 50 since January 2023. Jobs claims were at their lowest since May of this year and the four-week moving average has fallen every week since last July.

The People’s Bank of China (PBOC) announced a set of policy steps to support the economy. These include policy rate cuts, lower mortgage rates, lower down payments for homebuyers, a very generous swap program for asset management companies and insurers to use securities as collateral for cash and a funding program to help companies buy back their own shares. This gave an immediate boost to Hong Kong stocks, up 16% in a week, and China mainland shares, up 11%.

China is 23% of the MSCI Emerging Markets index which was up 7%. More to come. These monetary bazookas work in the short-term but can unwind in the long-term.

The S&P 500 reached another all-time high on Thursday and is now up 20% for the year. Investors don’t seem worried about any fears of a slowing economy.

Subscribe here to receive weekly updates.

---

Art of the week: Clark Voorhees (1871-1933)