iFlow > Short Thoughts
Published on Tuesdays, Short Thoughts offers perspectives on US funding markets, short-term Treasuries, bank reserves and deposits, and the Federal Reserve's policy and facilities.
John Velis
Time to Read: 5 minutes
EXHIBIT #1: BILL SUPPLY USUALLY MEETS BILL DEMAND
Source: BNY Markets, U.S. Department of the Treasury, Bloomberg, iFlow
The Supreme Court's (SCOTUS) ruling against the administration on the IEEPA tariffs has so far failed to produce meaningful or lasting market reactions. Currency markets largely took the announcement in stride. Bond markets also had little response. We think this has been the case because it’s still quite uncertain what a new trade policy regime will look like. While the President has announced new 15% across-the-board tariffs under a different authority, it’s not clear to us whether this new authority will be applicable or remain unchallenged. In any case, another reason for the muted reaction in Treasury markets is that future revenue lost from the nullification of the tariffs will likely be made up with short-end T-bill issuance, which we think will meet ready demand.
While many are discussing refunds that might be required after the SCOTUS ruling, we think this process will indeed be messy. We don’t feel strongly convicted one way or the other on the issue. For us, the lost customs duty going forward is more important to analyze. Much of it depends on how successful the administration is in imposing the new “Section 122” tariffs of 15% across the board. Under the 1974 Trade Act, these tariffs can be imposed for 150 days, after which Congress must approve them. We aren’t so sure this will be an easy task, especially given that any vote on approval would occur in late July, just as the midterm Congressional elections are heating up. Nevertheless, we expect the attempt to reset tariffs under additional authorities will continue.
The fiscal revenue impact is thus difficult for us to quantify precisely, given renewed uncertainty, but it is likely to be substantial. A back-of-the-envelope calculation suggests that about half of the approximately $20bn per month in tariff revenue would go missing, or around $120bn per year. We think that can be offset with T-bill issuance going forward, without placing significant upward pressure on long-term yields due to fiscal backsliding. Exhibit #1 shows that bill demand, as measured with our iFlow data on 0-1y U.S. Treasurys, responds nearly inelastically to bill supply. Our chart shows 10 years of data going back several years before the pandemic. Bill supply is robust to economic circumstances and more a function of steady demand for short-dated paper.
EXHIBIT #2: HIGHER PRODUCTIVITY IN THE PAST USUALLY MEANS HIGHER POLICY RATES
Source: BNY Markets, Federal Reserve Bank of San Francisco, Federal Reserve Bank of New York
GDP data last week was slightly disappointing, with the headline measure of national activity coming in at just 1.4%, half as strong as expectations. This suggests that Q4 labor productivity data will not show a massive quarterly jump in output per worker. We care about productivity because much of the discussion on the impact of artificial intelligence on efficiency has touted its revolutionary impact. This belief has given central bank doves comfort that higher productivity resulting from the AI boom will allow equilibrium policy rates to fall. We’re not as convinced. We argue that higher productivity from AI adoption and innovation could lead to higher rates once it finally shows up in the numbers, which we do think will eventually occur, although not immediately.
We begin by looking at a five-year rolling average of quarterly data on annualized U.S. total factor productivity. We wish to identify periods in the past several decades in which it was elevated as evidence of a productivity boom. We prefer total factor productivity (TFP) over the more common measure of simple labor productivity. TFP adjusts the contribution to output by adding one more unit of labor with technological changes (both the quantity of capital input and the changes in composition that improve productivity). Workers can be made more productive on the margin without technological improvements merely by increasing existing capital stock. Adjusting the output per unit of input measurement by changing capital stock accounts for this process and provides a more comprehensive picture of productivity improvements from technological change.
We use a widely accepted historical data series from the San Francisco Fed and plot its five-year moving average against the New York Fed’s estimate of the real neutral policy rate over time since 1967 (nearly 60 years). Two episodes stand out. The first is in the second half of the 1980s, when TFP rose markedly from its earlier growth rate. This period was accompanied by relatively steady and somewhat high estimated neutral rates. This may be a disingenuous result, given the decade still witnessed relatively high real rates, a hangover of the inflationary 1970s and the Volcker Fed’s reaction to that episode.
More recently, the late 1990s were characterized by very strong and sustained TFP growth. This period is often cited as an analogy to the current AI boom. Note that until the recession of 2001, real rates of r* were similarly elevated. It’s true that then-Fed Chairman Greenspan saw that rapid U.S. economic growth could coexist with a similarly rapid increase in inflation. Nominal rates between 1995 and 1999 were still between 5 and 5.5%, and real rates were rarely below 2% and as high as 4% or more during the period.
We think that the expectation of lower policy rates due to increasing productivity could be off-base. Higher productivity likely increases the demand for capital, raising interest rates in markets that could be crowded out by AI (and other technology) investment. Furthermore, in equilibrium, higher productivity should raise real wages, as workers are compensated for the increased output they can make with the same hourly input. This can raise consumption and lower savings, requiring higher equilibrium rates. Finally, higher growth expectations due to increased technological adoption should raise asset values, making them more impervious to higher interest rates.
In short, we’re not sure that higher productivity axiomatically leads to lower policy rates. We show empirically that this wasn’t clearly the case in the late 1900s, and we don’t think it will be going forward. This debate should prove lively going forward as we think about what a new Fed Chair and the productivity boom likely to be unleashed will mean for the real economy and financial markets.
The market has barely budged from its view of two rate cuts this year, with the DEC26 Overnight Indexed Swap (OIS) implying around 60bp in looser policy. This defies a set of minutes widely reckoned to be slightly hawkish, a sticky PCE inflation report last Friday, and a less-than-inspiring GDP report that was counter to expectations. We note that one must look all the way out to the July meeting before a full cut is priced. This is not too far from where we stand, still expecting three 25bp cuts in the second half of the year due to weakening labor markets.
The minutes had a slightly hawkish tone, not least because it was reported that “several” members indicated that they would have supported a two-sided description of the Committee’s future interest rate decisions, reflecting the possibility that upward adjustments to the target range for the federal funds rate could be appropriate if inflation remains at above-target levels.” This is the first time that rate hikes have been discussed in quite some time, although we’re not sure this is a realistic concern. We believe that if our view on U.S. labor markets materializes later this year, even if inflation remains sticky, the Fed will move rates in line with the full-employment side of its mandate.
