Chris Brightman, chief investment officer of Research Affiliates, provides insight into the potential impact on inflation arising from policymakers embracing Modern Monetary Theory (MMT), while PIMCO asset allocation strategist John Cavalieri discusses how the firm applies machine learning to contribute to alpha potential within the underlying PIMCO funds. As always, their insights are in the context of the PIMCO All Asset and All Asset All Authority funds.
Q: Does the embrace of Modern Monetary Theory (MMT) by some progressive politicians increase the risk of inflation?
Brightman: If progressive politicians were to put their ideas into practice through a massive expansion in government spending funded by printing money, then yes, such an embrace of MMT would raise inflation risks. The consequence of such a turn of events, in Research Affiliates’ view, would likely resemble a rerun of the Great Stagflation of the 1970s and its dismal capital market returns.
Little about Modern Monetary Theory (MMT) is published in the traditional manner (i.e., in peer-reviewed journals). Its proponents tend to rely on blogs and podcasts to communicate their ideas, and even progressive economists struggle to explain the theory. Nonetheless, MMT isn’t just a fringe idea. Prominent economists, including Kenneth Rogoff, Paul Krugman, and Larry Summers, are sufficiently concerned by the growing prominence of MMT that they have publicly cautioned about its potential dangers.
So, what is MMT? In a nutshell, MMT argues that governments should use fiscal policy to ensure full employment, in line with mainstream Keynesian prescriptions for fiscal policy. A more controversial claim by MMT proponents is that governments with a fiat currency can essentially fund any amount of government spending by creating new money. Released from the constraint to fund government spending with taxes, some promoters of MMT advocate a massive increase in government control of the economy – from universal healthcare and free college education to an immediate transition to clean energy, as well as government jobs for all of the unemployed.
Advocates of MMT acknowledge that such high levels of spending might cause inflation, but posit that taxes and regulation can and will prevent it. We believe relying on Congress to manage inflation through tax policy seems naïve and even reckless, regardless of whether it would be theoretically possible.
The fiscal and monetary policies of the 1960s in many ways foreshadowed MMT. A 20% cut in the income tax rate, paired with much higher levels of government spending to pay for the moon landing, the War on Poverty, and the war in Vietnam, fueled an economic boom and a rising rate of inflation. In 1968, Congress reversed course by enacting a large tax increase in an explicit attempt to control the rapid upsurge in inflation. It didn’t work.
In the early 1970s, the already bad economic situation worsened. Nixon ended the convertibility of the U.S. dollar into gold, imposed wage and price controls, and raised trade tariffs. Federal Reserve Chairman Arthur Burns acquiesced to Nixon’s pressure to ease monetary policy to boost the economy ahead of the 1972 presidential election. The economy responded positively, and Nixon was re-elected in a landslide. By 1974, inflation hit 11%, the result of easy monetary policy and the oil price shock of 1973–1974.
The latter half of the 1970s and the early 1980s, a period known as the Great Stagflation, was characterized by high and volatile inflation and rising unemployment. By 1982, unemployment had jumped to 11%, from 4% in 1969. The Misery Index (the sum of the inflation and unemployment rates), invented by economist Arthur Okun, quantified the toll on consumers, investors, and workers, peaking at over 22% in the early 1980s.
The cure would prove to be no less painful. Fed Chairman Paul Volcker, putting monetarism into practice, intentionally slowed the growth of the money supply, allowing market interest rates to rise above the double-digit level of inflation. After two nasty recessions in the early 1980s, the growth rate of money, inflation, and nominal interest rates all began their decades-long decline to the lows of recent years.
If MMT were put into practice today, we believe the result could easily be a return of such stagflation, and U.S. stocks, bonds, and cash would likely face return headwinds. Soaring interest rates could cause bond prices to plummet and possibly push real returns on cash into negative territory as well. U.S. stocks would likely provide little to no protection from high and volatile inflation either. In fact, they provided a real return barely above zero for the decade of the 1970s, with the Shiller price/earnings (P/E) ratio dropping from an average of 17x at the start of the decade to below 10x by 1977, and remaining in the 6x–10x range until 1984. From today’s Shiller P/E of 31x, above the 95th percentile, a return to valuations observed during the stagflation of the 1970s would imply a 70% plunge in U.S. stock prices – and that’s before considering the damage to corporate profits.
How might investors position for the low-probability but consequential risk of stagflation? The obvious answer is to add allocations to traditional inflation-sensitive asset classes like Treasury Inflation-Protected Securities (TIPS), commodities, and real estate investment trusts (REITs). Unfortunately, all three provide paltry starting yields from today's prices, in our view. The All Asset strategies, with access to a broader set of inflation-fighting asset classes, a tactical emphasis on what we view as the most attractively priced constituents, and alpha potential provided by the underlying PIMCO funds, represent a potentially higher-yielding and ultimately higher-returning alternative, all while preserving important inflation-protection characteristics.
Q: What is PIMCO’s perspective on the use of machine learning techniques in investing, whether to enhance asset allocation decisions or to identify alpha opportunities?
Cavalieri: We at PIMCO agree with Rob Arnott’s view, articulated in the May 2019 issue of All Asset All Access, that “as is true of most quantitative applications in finance, machine learning offers both great promise and potential danger.” We view machine learning (ML), which we define here as including deep learning and artificial intelligence (AI), as an important part of our toolkit in seeking superior returns for clients. However, it is not a panacea, and if used indiscriminately or for the wrong applications, it can be counterproductive.
For a problem to be well-suited for ML, especially deep learning and AI, we believe the following elements are necessary: 1) a truly vast (if not effectively unlimited) pool of data; 2) a small and consistent set of inputs and outputs; and 3) a well-defined way to measure success. Imagine the algorithm for a self-driving car, for example. Additional (unlimited) data can be generated simply by driving the car around; it relies on standard sensor data and steering controls (inputs); and the measure of success, avoidance of a crash, is well-defined. That’s a problem well-suited for ML to address.
Now consider the same ML elements in the context of building a trading strategy using daily returns. A data set going back 20 years would yield only about 5,000 observations, and the only way to generate additional data would be to wait. Given how much the financial markets have changed in the last 20 years, the inputs and outputs would not be constant (e.g., data inconsistencies could arise from different and evolving regulatory frameworks, changing monetary policy regimes, technological advancements, etc.). Moreover, measuring absolute success is difficult because finance is often a relative, zero-sum game: Everyone can have an effective car-driving algorithm, but not everyone can beat the market. So at PIMCO, we're focusing our machine learning efforts on areas where we believe the data are truly powerful.
Some of the key areas include:
- Macro forecasting – using dynamic factor models based on reams of high-frequency economic data to make short-horizon estimates of likely changes in GDP growth or inflation; using neural networks to estimate changing probabilities of recession risk; and applying natural language processing to central bank releases to infer changes in hawkish or dovish sentiments
- Fundamental analysis – using machine learning (i.e., gradient-boosted trees) to predict changing loan-level default and prepayment rates in agency mortgages, or using natural language processing to parse corporate filings for language changes that could affect our assessment of a company’s stock or bond prices
These are just some of the machine learning applications we use today, and we are evaluating still more for future use.
These applications are not “silver bullet” predictors, of course, and there are many caveats, limitations, and strict preconditions to their use. This is why we use ML applications as complementary tools to enhance our time-tested investment process, not as replacements for it. Nevertheless, we think it’s important to stay at the forefront of innovations in ML, and to that end we have recently partnered with Caltech's Computing and Mathematical Sciences department and with Chicago Booth's Center for Decision Research. This allows us to remain in regular contact with academic researchers at the leading edge of new techniques.
In sum, we agree with Research Affiliates’ view that ML applications for longer-horizon asset allocation decisions are currently limited, given insufficient and inconsistent data sets relating to long-term asset class returns. However, we think they can be useful in seeking certain targeted and higher-frequency alpha opportunities within the suite of PIMCO funds underlying the All Asset strategies. We believe this approach reinforces the partnership between Research Affiliates and PIMCO in managing the All Asset funds, with each firm contributing complementary value-enhancing strategies and techniques in different parts of the same portfolio.
John Cavalieri would like to acknowledge and thank PIMCO’s Emmanuel Sharef for his contributions to the machine learning discussion.
Recent editions of All Asset All Access offer in-depth insights from Research Affiliates on these key topics:
- The role of machine learning in quantitative investment models and factors driving current risk tolerance within the All Asset strategies (May 2019)
- Differences between the All Asset and All Authority strategies, and their underlying allocation infrastructure (April 2019)
- Assessing investment risk in terms of the likelihood of meeting long-term wealth accumulation goals (March 2019)
- The potential impact of a bear market in U.S. stocks on emerging markets (February 2019)
- Research Affiliates’ outlook and asset allocation views for 2019 (January 2019)
- Market impact of the U.S. midterm elections and what differentiates All Asset’s positioning versus peers (December 2018)
The All Asset strategies represent a joint effort between PIMCO and Research Affiliates. PIMCO provides the broad range of underlying strategies – spanning global stocks, global bonds, commodities, real estate and liquid alternative strategies – each actively managed to maximize potential alpha. Research Affiliates, an investment advisory firm founded in 2002 by Rob Arnott and a global leader in asset allocation, serves as the sub-advisor responsible for the asset allocation decisions. Research Affiliates uses their deep research focus to develop a series of value-oriented, contrarian models that determine the appropriate mix of underlying PIMCO strategies in seeking All Asset’s return and risk goals.