Viewpoints

My key observation about the current angst about “high” bond yields: I much prefer high real interest rates associated with healthy economic growth and high returns to capital than low real interest rates associated with economic malaise and low expectations.

Also, while I have long argued that the persistently high government budget deficits and rising debt/GDP are a major problem that involve a misallocation of national resources and must be addressed, I’m not so sure that the recent rise in rates can be attributable to the government’s rising debt.

It’s a tossup between what was more illogical last week, the headlines lamenting “soaring bond yields” or U.S. Secretary of the Treasury Scott Bessent’s announcement that the Treasury had upped its bond buyback program with the hope of lowering yields on long-duration Treasury securities.

A bit of historical perspective is instructive. Current yields are not “soaring” or even “high”—they are in a range that is consistent with current economic and inflation conditions—and these levels are preferable to the low bond yields that characterized the troubled post-Great Financial Crisis period. And Bessent’s Wall Street career in fixed income presumably taught him that interventions aimed at manipulating the bond market would not work; clearly, his announced intervention was a political ploy to placate President Trump. It didn’t work. I strongly recommend Stanley Druckenmiller’s Wall Street Journal article “Let the Bond Market Speak”, August 24, 2026.

Currently, 10-year US Treasury yields are roughly 4.7%, up from 4.2% earlier in 2026. With inflationary expectations around 2.25%, real rates are in the 2.4%-2.5% range. It’s likely that real rates would be even higher if not for the Trump Administration’s tariffs, clampdown on immigration and other policies that have dampened real growth and threatened longer-run potential. Current real rates are above the low bond yields during 2010-2014 (they ranged between 2.5%-3.25%) that were associated with the disappointingly soft economic recovery from the Great Financial Crisis, lingering high unemployment and diminished expectations, the Fed’s zero interest rates and extended asset purchases and worries that inflation was too low, but significantly below the bond yields of the 1990s and modestly below those that prevailed during portions of the early 2000s.

A cursory comparison of recent decades suggests clearly that the post-GFC period was aberrant period. (Federal Reserve Bank of New York research “Measuring the Natural Rate of Interest”, 2026, estimates that the 2009-2014 period was the low point of the natural rate in modern history.) Yet many financial market participants and commentators use the low rates of that aberrant period as the benchmark for characterizing current yields as high and rising, when economic conditions are far different.

Chart 1 shows 10-year Treasury bond yields and core PCE inflation since 1990. Chart 2 shows the inflation-adjusted bond yields (a measure of inflationary expectations is not available back to 1990) and their correlation to productivity gains.

A surge in productivity growth could solve a lot of problems. It would mean higher government revenue and, combined with spending discipline, lower the budget deficits. Indeed, it is the centerpiece of Treasury Secretary Bessent’s deficit reduction plan. It could also mean easier Fed policy: if productivity increases are not matched by higher compensation growth, unit labor costs and inflation weaken, allowing the Fed to ease rather than hike. This was Warsh’s argument for rate cuts early this year (and he continues to talk about how great productivity is).

The case for and against a surge

Both Administration economists, and Warsh during his campaign for Fed chair, argue that growth-friendly Administration policies and the AI revolution mean higher trend growth. Tax cuts, they argue, stimulate investment and labor supply, while deregulation increases economic efficiency.

I’m skeptical for three reasons. First, we have seen this movie before, and it had a flat ending. Starting with President Carter and continuing with Reagan there was a major push for deregulation in the 1970s and 1980s. Indeed, it was much bigger than what Trump is doing. Reagan also implemented major tax cuts in marginal tax rates. Again, they were much bigger than what Trump has implemented.

And yet, history shows there was no pick-up in productivity or trend growth during this period. With the benefit of hindsight, the CBO has good estimates of what happened to trend growth during and following the “Reagan revolution.” They show productivity initially rebounding from the 1982 recession but then fading for the rest of the recovery (chart). A sustained pick-up did not happen until the “new paradigm” technology boost, starting in 1995.

In my youth, I co-authored a paper on The Supply-Side Consequences of U.S. Fiscal Policy in the 1980s at the NY Fed. We found early evidence that supply-side indicators, like investment, productivity and the labor supply did not respond to the policy changes. We then used a simple simulation model, to show that the crowding out of investment from surging budget deficits offset the benefits of lower marginal tax rates and deregulation. The CBO numbers confirmed those preliminary findings.

Starting on August 27, 2026, and running through August 29, 2026, the Federal Reserve Bank of Kansas City will host its annual “barbecue” in Jackson Hole, Wyoming. The best and the brightest of monetary policy gurus from around the world will be there and the highlight of every conference is the comments by the chairman of the Federal Reserve Board. (For some reason, my invitation has repeatedly been lost in the mail.) The theme of the 2026 conference is “Financial Innovation: Implications for Payments Policy”. Although this is an important topic, is it really the most pressing issue for the Federal Reserve to be considering? Of course, I am biased, but I believe that Fed should be discussing a superior way to conduct its monetary policy so as to more consistently achieve an inflation target, especially in an environment of random shocks to the aggregate supply of real goods and services. For whatever reason, these random shocks seem to have been occurring more frequently in recent years. So, I humbly suggest that Fed Chairman include in his remarks that the Federal Reserve operate in a manner such that the sum of depository institution reserves and the securities and loans on the books of depository institutions grow at some constant rate. (Depository institutions are commercial banks, saving institutions and credit unions. Commercial banks dominate depository institutions.) In what follows, I will provide empirical evidence demonstrating that growth in the nominal annual averages of this sum, let us call it “thin-air” credit, has a relatively high correlation with future growth in the nominal annual averages of domestic aggregate demand and the future rate of inflation associated with domestic aggregate demand. In addition to having relatively high correlations with growth in nominal domestic aggregate demand and the associated inflation rate, a constant rate of growth in “thin-air” credit will prevent cumulative increases or decreases in the rate of inflation. I will provide the rationale for why I believe 5-1/2 percent would be a reasonable target rate of growth in thin-air credit. I also will explain how the Federal Reserve can achieve, with precision, whatever target rate of thin-air credit growth it chooses without operating via a federal funds rate target. Not only would this modest proposal eliminate persistent overshoots and undershoots of the Fed’s inflation target, it would simplify the Fed’s communication challenges. In terms of forward guidance, all the Fed chairman would need to say at the end of every Federal Open Market Committee (FOMC) meeting is that the committee intends to have the sum of depository institution reserves plus securities and loans grow at a steady annual rate of 5-1/2 percent.

Plotted in Chart 1 are the year-over-year percent changes in the annual averages of the sum of depository institution reserves at the Federal Reserve, securities and loans (the blue bars) and the year-over-year percent changes in the annual averages of nominal Gross Domestic Purchases (the red line). (Monetary policy primarily influences domestic aggregate demand. That is why I have chosen nominal Gross Domestic Purchases as the “dependent” variable rather than Gross Domestic Product, because the latter involves exports, which is more of a function of foreign demand for US goods and services.) The gray shaded areas represent periods of recession. The highest correlation between these two series starting with 1955 data occurs when thin-air credit is advanced by one year. That correlation is 0.52 (shown in the upper left corner of the chart). This implies that growth in thin-air credit leads growth in nominal Gross Domestic Purchases by one year. So, what happens to thin-air credit this year has its largest effect on nominal Gross Domestic Purchases next year.

State labor markets were fairly stable in July. One state (Maryland) reported a statistically significant increase in jobs, and one state (New Jersey) clocked a significant decline. A full 10 states saw statistically significant drops in their unemployment rate, though none was larger than .2 percentage points. The highest unemployment rates were in DC (5.9%), Connecticut (5.2%), Oregon (5.2%), California (5.1%), Nevada (5.0%), and Washington (5.0%). Nebraska, New Hampshire, Hawaii, Vermont, North Dakota, and South Dakota had unemployment rates under 3.0%, while South Dakota’s 2.0% was the lowest in the nation.

Puerto Rico’s unemployment rate moved up to 6.0% and the island’s job count moved up by 2000.

The world is being pushed to invest more than it has in decades.

A world on autopilot Every few months the International Monetary Fund publishes a projection of the world economy, and the Blue Chip panel of forecasters, along with other bodies, does something similar each month; together they help to set the consensus against which everything else is judged. It describes a world returning gradually to normal: global growth a little above 3%, inflation drifting back to target, and US policy rates continuing to ease from their peaks. Its most consequential feature is one that is easy to miss, because it sits in an accounting identity rather than being stated as a view: investment in the advanced economies is projected to remain close to 22% of GDP for the rest of the decade, essentially flat, with national saving tracking it (chart below). The large external imbalances — the US deficit and the surpluses of China, Germany and the oil exporters — are expected to persist more or less unchanged, and the real interest rate that balances saving against investment drifts gently lower. It is, in short, a forecast of continuity, and this piece argues that continuity may well be the wrong assumption.

The world has become more fragmented politically in the past decade, which has heightened global risk and uncertainty. The diminished sense of security has caused many countries to increase military spending. While this may be a rational response for individual countries, collectively, the rise in military spending has been alarming and seems to be leading the world toward even more risk and uncertainty. This in turn may feed the need for more security. In addition, the devotion of so much of the world’s resources to military spending may be damaging for the global economy.

According to the Stockholm International Peace Research Institute (SIPRI), world annual military expenditure reached $2.8 trillion (in constant 2024 terms) in 2025. That figure represented a 41 percent increase in the past ten years. However, half of that rise occurred in the past three years. We looked at the SIPRI Military Expenditure dataset (available in Haver’s GLSECTOR database) to understand how and why military spending is changing across countries and regions.

Worldwide gains in military spending The rise in military spending has been widespread. Between the years 2022 and 2025, 45 of the top 50 countries in terms of military spending posted increases, with an average gain among that group of 41.8 percent. NATO spending (excluding the US) jumped by 44.7 percent, while non-NATO military spending increased by 29.2 percent.

Surprisingly, US military spending was essentially flat during this time. As a result, US military spending declined as a share of world spending by 6.6 percentage points to 33.5 percent. However, the Trump Administration has proposed a $1.5 trillion military budget for 2027, a roughly 60 percent increase from the 2025 spending listed in Figure 1. If that budget is enacted, then in 2027 the US's increase alone would be equivalent to 20 percent of 2025's entire world total. And we know that other countries both in NATO and Asia are ramping up spending as well.

It has been argued that AI could usher in "10x of the Industrial Revolution at 10x the speed." Yet, for AI to deliver on its growth potential for the economy, not just for a select group of companies, it needs to create jobs, lots of them- to generate worker income and spur consumption growth. So far, the records for growth, jobs, and productivity are unimpressive; however, to be fair, they are far from the final scorecard.

In the first half of 2026, even with double-digit annualized gains in business spending on technology equipment and intellectual property (AI), the economy saw an annual growth rate of 1.75%, with real consumer spending rising by 1.8%, the creation of 450,000 payroll jobs, and productivity growth of about 1%. Various factors impacted that growth and job performance, with AI emerging as the most significant influence.

AI is reportedly already managing or assisting with large volumes of white-collar tasks, allowing some companies to reduce their headcount. But the future impact could be more significant in scale and breadth. AI is designed to perform tasks traditionally done by humans, whereas the Industrial Revolution was a massive job creator.

Predicting the impact of AI on the total number of employed individuals is challenging. Nonetheless, if AI enables companies to downsize their workforce or eliminates the necessity to expand it, this could significantly harm overall economic performance.

This isn't a pessimistic view; it's simply basic arithmetic. Real consumer spending accounts for 70% of overall GDP growth (add another 3% for housing), while all types of business investment spending are less than 15%. If real consumer spending decreases by 100 basis points due to sluggish job growth, investments in AI models and infrastructure would need to increase by 3X or 4X above the current rate to maintain the same GDP growth rate.

Is it possible? A surge in AI investment of this magnitude would far exceed most companies' current cash flows, requiring unprecedented borrowing from the private sector. Yet, then the real question with AI is whether such an investment is essential, considering AI might impede the growth of its main market, the consumer, due to stagnant job growth.

It's crucial to find a balance where AI enhances GDP output and productivity without adversely affecting the job market to the point where consumer spending significantly decreases. However, it's uncertain where such a trade-off exists.

These issues will not be settled today, next month, or even next year. Historically, infrastructure booms and new technologies take years before their economic and financial benefits become evident. Nonetheless, all past innovations have ultimately created more jobs than they eliminated. For AI to be truly successful, it must create more jobs than it removes, and the responsibility to prove this lies with AI.

There continues to be talk about how high-income households are carrying the economy. Is this new? The Bureau of Labor Statistics (BLS) annually surveys households regarding their expenditures in the aptly-named Consumer Expenditure Survey (CES). The latest BLS Consumer Expenditure Survey is for 2024. So, we do not know what went on in 2025 and so far in 2026. But what we do know from the 2024 CES is that higher-income households continue to spend more than lower income households – duh. But what we also know is that in 2024, things were close to or at their 1984-2024 medians with respect to average annual nominal expenditures compared to aggregate expenditures. To wit, in 2024, average annual nominal expenditures for the bottom 20% income quintile for households compared to the total average expenditures of all income quintiles was 8.9% versus a median of 8.8%. The comparable figure for the top 20% income quintile for households was 38.3% versus its 38.3% median. These data as shown in the chart below.