The Fed Depends on Undependable Data

In its January 31, 2024 FOMC statement, the Fed said: “In assessing the appropriate stance of monetary policy, the Committee will continue to monitor the implications of incoming information for the economic outlook.” The translation of this Fedspeak is that the Fed’s target level of the federal funds going forward would depend on the forthcoming data as they relate to the Fed’s dual mandates of promoting price stability and full employment. But what if the data upon which the Fed were depending to determine the level of the federal funds rate were undependable? In what follows, I will provide examples of “undependable” data and recommend a solution for how the Fed might conduct monetary policy in the face of undependable data.

In the Bureau of Labor Statistics (BLS) February 2024 Employment Situation, it was reported that the level of January 2024 nonfarm business establishment payrolls was 157,533 thousand, revised down from its preliminary estimate of 157,700 thousand. Mind you, this is just the first revision of January 2024 nonfarm payrolls. When the BLS releases its March 2024 Employment Situation report, there will be a second revision to January 2024 nonfarm payrolls. And then in 2025, there will be annual “benchmark” revisions to 2024 nonfarm payrolls, including those of January 2024. The level of February 2024 nonfarm payrolls reported on March 8, 2024, 157,808 thousand, was said up 275 thousand compared to the first-revised January 2024 level of nonfarm payrolls. However, compared to the first-reported level of January 2024 nonfarm payrolls, the level of February 2024 nonfarm payrolls was up only 108 thousand. And, of course, in the next two 2024 BLS Employment Situation reports, the February 2024 level of nonfarm payrolls will be revised twice. Because of monthly and annual revisions, the monthly reports of nonfarm payrolls would seem to be undependable data upon which the Federal Reserve might use to determine monetary policy.

On March 14, 2024, the Census Bureau reported that the level of February 2024 retail sales increased 0.6% compared to the revised level of January 2024 retail sales. However, the level of January 2024 had been revised down by $3,581 million or 0.5% from the originally-reported level. So, the level of February 2024 retail sales was up only 0.06%, not 0.6% from the originally-reported level of January 2024 retail sales. Based on revised data in the February 2024 retail sales report, in the three months ended January 2024, retail sales contracted at an annualized rate of 3.8%. Based on the data reported in the January 2024 retail sales report, in the three months ended January 2024, retail sales contracted at an annualized rate of only 1.8%, less than half the rate of contraction exhibited by the data revised in the February 2024 retail sales report. Again, monthly revisions to retail sales data would suggest that these data are undependable for the purposes of guiding monetary policy.

The next problematic economic report I will discuss is the Consumer Price Index (CPI), more specifically, the Owners’ Equivalent Rent (OER) component of the CPI. At 26.7% of the CPI, OER has the “heaviest” weight in the CPI. That OER has such a high weight in the CPI is understandable given that the US homeownership rate is about 66%. My quarrel is not with the weight of OER but how it is estimated. From what I have read about this estimation process is that a sample of homeowners are asked by the BLS what the respondents think their detached dwelling/condo/townhouse would rent for. How many homeowners, especially owners of detached houses, have a reasonably accurate estimate of what their abode would rent for?

OER was reported to have increased month-to-month annualized 6.94% in January 2024 compared to a 5.22% annualized increase in December 2023. The CPI excluding OER monthly increase was 2.66% annualized in January 2024 compared to 2.03% in December 2023. The month-to-month annualized change in the CPI-All Items was 3.73% in January 2024 compared to 2.83% in December 2023. The BLS received queries as to why there was such a relatively large percent increase in the January 2024 OER compared to December 2023. On February 29, 2024, the BLS issued a statement saying that there are now annual updates effective in January of a year in the weighting of the OER in terms of owner-occupied detached dwellings versus condos/townhouses. The BLS said that “[i]n January 2024, the proportion of OER weighted toward single-family-detached homes increased by approximately 5 percentage points.” My point, again, is not that OER is unimportant, but that its measurement is, for lack of a better term, “flaky”. Given the difficulty in accurately measuring OER, the European Union excludes OER from its calculation of EU consumer price inflation. Plotted in Chart 1 are the year-over-year percent changes in the All-Items CPI (the blue bars) and the CPI excluding OER (the red line). The year-over-year change in the CPI excluding OER in February 2024 was 2.27%, close enough to 2% for Federal Reserve work.

So, how should the Fed operate in the face of “undependable” data? Taking a page from Milton Friedman’s writings, the Federal Reserve should try to stabilize the growth rate of a monetary quantity. The monetary quantity of whose growth rate Friedman thought the Fed should stabilize is the M2 money supply. I would prefer that the Fed attempt to stabilize the growth rate of the sum of the monetary base (currency plus reserves held by depository institutions at the Fed) and loans and securities held by depository institutions. The Austrian school of economics refers to this monetary quantity as created credit, credit created, figuratively, out of thin air. If it can be shown that growth in a monetary quantity has a relatively close relationship to the growth in the real aggregate demand and the inflation rate, then stabilizing the growth in this monetary could be compared to an element of the Hippocratic Oath, “First, do no harm.” Given the lags in the reporting of revised economic data, given that an equilibrium federal funds rate level is not a constant through time, given the variable lags between a change in the level of the federal funds rate and the behavior of real aggregate demand and the inflation rate and given the randomness of exogenous shocks to an economy, such as a pandemic, there are a lot of opportunities for federal-funds rate targeting to do a lot of harm.

Let’s examine the historical relationship between the growth in “thin-air” credit, real aggregate demand and an inflation rate. Plotted in Chart 2 are the year-over-year percent changes of annual averages of Thin-Air credit deflated by the chain-price index of Gross Domestic Purchases (blue line) and real Gross Domestic Purchases (red bars). In the upper left-hand corner of the chart is the correlation coefficient, +0.61. The “plus” sign in front of the numerical correlation coefficient indicates that the two series are positively correlated. If the two series were perfectly correlated, the correlation coefficient would be 1.00. So, a correlation coefficient of 0.61 indicates that there is a relatively high association between the two series – not perfect, but relatively high. The data run from 1959 through 2019, before the exogenous pandemic shock. If the period is extended through 2023, the correlation coefficient falls to 0.43. Notice that contemporaneous values of each variable are being compared. If in the 1959-2019 period real Thin-Air credit is advanced by one year, meaning that the percent change in Thin-Air credit in year X is associated with the percent change in real Gross Domestic Purchases in year X+1, that is, percentage changes in real Thin-Air credit “cause” percentage changes in real Gross Domestic Purchases, the correlation coefficient declines. But the correlation coefficient declines even more if we test for percentage changes in real Gross Domestic Purchases “cause” percentage changes in real Thin-Air credit. So, the evidence suggests that percentage changes in real Thin-Air credit “causes” percentage changes in real Gross Domestic Purchases contemporaneously.

Now let’s look at the relationship between year-over-year percentage changes in Thin-Air credit (the blue line in Chart 3) and the chain-price index for Gross Domestic Purchases (the red bars in Chart 3). The data start in 1959 and end in 2023. In Chart 3, percentage changes in Thin-Air credit is advanced by two years, suggesting that the percentage change in Thin-Air credit in year X is most highly associated with Gross Domestic Purchases inflation in year X+2. The correlation coefficient is +0.60. The correlation coefficient also is 0.60 if the data period is truncated at 2019. Notice that the percentage changes in Thin-Air credit plotted in 2024 and 2025 (actually the percentage changes in 2022 and 2023 because the variable is advanced by two years) declined to 2.64% and 1.60%, respectively. This suggests that the rate of inflation would be expected to slow in 2024 and 2025 based on the behavior of Thin-Air credit.

In sum, due to numerous and relatively large revisions to government-estimated economic data, the Fed depends on undependable data with which to make its decisions in relation to adjusting the level of the federal funds rate. As an alternative, I would suggest that the Fed attempt to stabilize the rate of change in nominal Thin-Air credit, which historically has demonstrated a relatively stable relationship with the rate of inflation and, when adjusted for prices, has demonstrated a relatively stable relationship with the rate of change in domestic aggregate demand. What rate of change in Thin-Air credit should the Fed target? I would task the largest institutional assembly of economists in the World, the Fed’s economic research staff, to answer this question.