by the Asset Allocation Committee | PDF

It’s probably no surprise to anyone that information processing has become a bigger and bigger part of the economy over the last several decades.  By now, we’ve all gotten used to seeing new kinds of computers and telephones every year, and we’ve struggled with learning new software programs at home or at work.  Whether you call it the “digital economy” or the “knowledge economy,” what few people realize is that it’s about to reach an important milestone in the data on U.S. gross domestic product (GDP).  Therefore, this may be a good time to explore some of the investment ramifications of the trend.

In the U.S., private investment spending currently accounts for about 17.5% of total GDP, but the category naturally includes a lot of subcomponents.  Housing and corporate inventory investment are only a small part of the total.  The biggest component of the category is “nonresidential fixed investment,” which is often referred to as “corporate” fixed investment and makes up roughly 13.5% of GDP.  As late as the 1990s, corporate fixed investment was dominated by spending on machinery, equipment, and structures.  Investment in intellectual property—mostly spending on research and development—was a much smaller part of the category.  The difference now is that spending on intellectual property has exploded, mostly because of a surge in software investment.  Overall, corporate fixed investment has grown at an average annual rate of 7.6% over the last four decades (including price changes), with software investment rising at a rate of 10.9% and R&D investment rising at a rate of 7.1%.  As a result, intellectual property is on the verge of becoming the biggest component of corporate fixed investment.

The implications of this evolution are enormous.  In economic terms, the marginal cost of selling an additional software package or leveraging the R&D behind it can be minimal.  In addition, digital products and software can have big spillover effects, in the sense that they can give value to customers beyond the core service the software is designed for.  A good example of this is the “network effect” when the customer gets greater value from joining a social media network when the network has more people.  The low cost of selling additional units and the incentive to capture spillover value means that digital businesses have a huge incentive to grow their network and become as big as possible.  The growing dominance of intellectual property in the economy is probably one reason why corporate profits as a share of GDP have risen starkly over the last several decades (see chart below).  Note, however, that those businesses’ enormous size is now generating concerns that they are stifling competition, which could lead to tougher antitrust regulation around the world.

Naturally, investors are attracted to the increasing size and profitability of firms that produce software or can efficiently leverage their digital R&D.  In addition, as other businesses continue learning how to boost their productivity using big data, data management, data analysis, and artificial intelligence, they are probably less likely to cut their intellectual property investment in times of recession.  Indeed, intellectual property investment continued to rise throughout the Great Financial Crisis of 2008-2009 and the recession in 2020 due to the coronavirus, as shown in the first chart above.  From another perspective, the growth rate of inflation-adjusted intellectual property investment has had a standard deviation of just 5.2% since 2000, which is modestly lower than the standard deviation of 5.8% for overall GDP and dramatically lower than the average deviation of 13.9% for equipment and 14.4% for structures.  In sum, intellectual property investment is much more stable than investors realize, so there is a good economic reason why digital technology and software firms are increasingly being seen as “defensive.”  In many environments, we favor this sector for offering an attractive combination of good prospects with relatively less sales volatility than investors perceive.

View PDF

by the Asset Allocation Committee | PDF

In last week’s report, we updated our views on long-duration Treasuries, using our standard 10-year T-note model.  This week we are going to examine the impact of policy on long-duration Treasuries.

First, let’s start with the model.

The two most important variables in the model are fed funds and the 15-year average of the yearly change in CPI.  The latter acts as a proxy for inflation expectations.  We add the yen’s exchange rate, oil prices, German Bund yields, and the fiscal deficit scaled to GDP.  So, there are two policy variables in the model, fed funds and the fiscal deficit.

A truism that has developed in recent years is that divided government is good for financial markets because different parties in the White House and Congress means it is harder to enact major legislative changes.  To test this thesis, we created a binary variable that signaled if one party controlled the legislature and the executive branch, or not.  When we added the variable to the model, it was not statistically significant, nor did it change the model’s forecast.

Regression models are sensitive to initial conditions; simply put, where you start a model in time has a notable impact.  Jim Bullard, the president of the St. Louis FRB, likes to think of eras with certain conditions as “regimes.”  Applying Bullard’s notion of regimes, we next broke down the data by attitudes toward government.  In other words, the truism that divided government was good for financial markets may not have always been true.  To test this idea, we first ran the model from 1960 to 1982; attitudes toward the government were once positive.  For the generation that lived through WWII, the government was not seen as an obstacle to overcome but as a support.  The model did change; the government variable was statistically significant and showed that when the same party held both branches of government, yields were 50 bps lower.  From 1983 to the present, the government variable was also statistically significant, but the sign changed.  Unified government increased bond yields by 31 bps.  We propose that the Reagan/Thatcher revolution changed the views of government compared to the earlier era.

Another interesting factor is that in the earlier regime, deficits were bearish for bonds; in other words, a deficit led to higher yields.  In the current regime, deficits are less significant but the sign on the coefficient changed here as well.  Deficits now lead to lower yields, although the impact is quite modest.

Finally, the model from 1983 has a current fair value of 1.70%, which is a bit higher than the full model shown above.  Thus, its fair value is close enough to give us confidence that it is giving proper signals.  The other important takeaway from this analysis is that if the Democrats lose one of the houses of Congress in November 2022, caeteris paribus, the fair value yield would decline by nearly 50 bps.

View PDF

by the Asset Allocation Committee | PDF

One of the key developments in financial markets recently has been the quick rebound in bond prices and the associated drop in yields.  As investors started to sense faster economic growth and the prospect of rising inflation in the first quarter, they eagerly sold down their bond holdings, driving yields higher.  The yield on the benchmark 10-year Treasury note jumped from 0.92% at the end of 2020 to an intraday high of more than 1.75% in late March (see following chart).  Since then, however, bond buying has gradually strengthened again, and yields trended downward throughout the second quarter.  Even when the Federal Reserve surprised markets in mid-June by hinting that the next interest rate hike might come by 2023, the resulting bond sell-off and yield jump was quickly reversed.  In July, investors began buying bonds and driving down yields even faster, with the 10-year Treasury yield falling below 1.13% on July 19.  This report looks at why yields are falling again and whether the downtrend is likely to continue.

It’s one thing to know where yields have been; it’s quite another thing to understand where they should be and where they may be going.  To get at that issue, our bond model estimates where the 10-year Treasury yield should be based on several variables.  The most important variables are the Fed’s “fed funds” interest rate and, as a proxy for inflation expectations, the 15-year average of the yearly change in the consumer price index (CPI).  Other variables in our model include the yen/dollar exchange rate, oil prices, German Bund yields, and the U.S. fiscal deficit scaled to gross domestic product (GDP).  As shown in the chart below, the jump on bond yields early this year merely brought the 10-year Treasury yield up to the fair value estimated by our model.  With the recent rebound in bond prices, yields have again fallen below their fair value, which the model currently puts at 1.65%.

In other words, bonds once again look expensive―not as expensive as at the beginning of 2021, but enough to suggest bond investors see reason for caution regarding monetary policy and economic prospects.  Many investors think the Fed will eventually tighten monetary policy just enough to smoothly bring down today’s high inflation rate.  A less benign view among other investors is that the Fed might tighten policy too soon or too quickly.  Those investors are worried about a policy mistake that could trip up the economic recovery and produce another recession.  Still other investors think longstanding structural factors such as globalization and population aging will eventually reassert themselves and push down growth and inflation to the levels seen before the coronavirus pandemic.  Such an environment of rising interest rates in the near term coupled with an eventual moderation in rates is consistent with the recent flattening in the yield curve.  For example, the chart below shows how the yield curve, represented by the difference between the 10-year and the two-year Treasury rates, has changed recently.

In any case, we think our model’s call for higher interest rates should be respected, especially given the risk that inflation could stay high for longer than anticipated and gradually push up longer-term inflation expectations.  We note that the 15-year average of CPI inflation that serves as a proxy for inflation expectations in our model is just under 1.90%.  Other measures of inflation expectations, such as consumer surveys and the difference between nominal and inflation-protected bond yields, point to even higher future inflation.  Even if the 10-year Treasury yield rebounds in the near term, we still believe it probably wouldn’t go much past 2.00%, but the risk of rising yields (and falling bond prices) has prompted us to reduce our exposure to longer-term bonds in several of our strategies for the third quarter.

View PDF