14: Where the Figawi in 2024

It’s January 2024 and time to see “where the figawi”, like the eponymous Hyannis-Nantucket sailboat race.  As a sailor and investor, it seems to me investing is not dissimilar from sailing. Both are attempting to successfully find a strategy to reach a destination, while dealing with uncertain weather.  We’ll review the current conditions and plan for the weather ahead. 

AT THE STARTING LINE

It began over 15 years ago when I chose to design and build a set of learning machines to make investment decisions for myself and useful for any long term investor, InvestEngines.  Learning machines fall within the broad category of artificial intelligence, AI.

2023 was a year when AI became “infamous” with the launches of readily available large language models (LLMs) like ChatGPT, Bard (now Gemini) and many others.  Although the complexity, scale and intent of those efforts dwarf the simplicity and goals of InvestEngines they share a common genesis, using lots of “training” data to uncover relationships in the data. 

The models of InvestEngines have evolved significantly and have grown in complexity, but the result is just a framework and simple guidance for asset allocation.  The “training data” for portfolios spans a minimum of 20 years.  There are 100 years of data for the long term market models.  The training period ended in 2021.  So applying those models for the past 2 plus years is providing some real-time performance results.  Here is the current 1 year and 30 month performance summary (end of 2023 at this writing).  Keep in mind that these are short term (updated daily) results relative to the longer term focus of the InvestEngines’ learning machines.

FIGURE 1: RECENT PERFORMANCE STATISTICS

The learning machine performance uses the asset allocation guidance of the learning machines.  The 60/40 benchmark is the rebalanced allocation (60% equities and 40% bonds) of those same ETFs.  The leveraged machine (see article 13 titled “Running with Scissors”) is the more recent exploration that includes some limited use of leverage.  The weighted machine combines both the levered and non-levered machines.

Currently the learning machines of InvestEngines are suggesting a broadly diversified set of ETFs (matching the benchmark) with added emphasis on US equities if one includes the current levered learning machine weightings.  Peering under the hood of the learning machines, I would expect some changes in 2024.  Especially if the FED begins to ease interest rates aggressively or the market has a hard landing or both.  

AT THE FIRST MARK

One of the data models I explored to get a sense of what may lie at the next mark is worth a revisit.  The article titled “Climbing Further Out on a Limb” attempted to identify shorter term trends.  The data is noisy, but it indicated an accelerating cyclicality and long term decline in average returns.  The training data began in the late 1990’s and ended mid 2021.  The trend showed a broad portfolio return low was likely to occur toward the end of 2022.  The learning machine portfolio 1 year return did in fact hit a low at the backend of 2022 and has seen a bumpy ride up since then.  The model implies the next return peak may have just occurred and the next low year will be seen toward the end of 2024.  (The blue line is the 1 year actual portfolio performance, updated daily, and the red line is the smoothed average predictor.)

FIGURE 2: PORTFOLIO PERFORMANCE SINCE 2000

This data is not used to influence the quantitative guidance of InvestEngines’ asset allocation models, nor as an attempt to time the market, but rather to provide a general sense of what to expect going forward.  Things now are very unclear with heightened uncertainty in geo-politics, market expectations and interest rates.  This metric is short term and noisy.  Hopefully there is some more upside over the next 6 months or more.  If it does provide a rough sense of conditions, things could start to soften during 2024 and 2025 could be a downer. 

So what’s the best InvestEngines can do to provide some semblance of the course ahead?  Not much better than a probability distribution for the short term.  And the best data we have is from all the training over the past 25 years and how the learning machine found an optimal set of parameters.  The charts below show the probability distribution of the 1 year return for the learning machines performance.  The x axis is the band of return ranges and the y axis is the probability over time of returns falling into those bands.

FIGURE 3: PORTFOLIO RETURN PROBABILITY DISTRIBUTION

So it looks like 2023 fell in the middle, most probable, range.  (The distribution on the left is the 4 engine learning machine, while the chart on the right includes some help from the levered learning machine.) The fact that the longer term average seems to be declining one would expect the most probable outcome is sliding south.  But any one year is still a crap shoot and with all the current uncertainties it’s a big crap shoot.

AT THE FINISH LINE

So what can the learning machines conclude over the longer term?  In the article titled “Details” the learning machine models used valuation and interest rates over the past 100 years to establish a predictor.  There was a reasonable 84% correlation between the learning machine’s predictor and future 10 year returns of the DOW.  Below is a chart of 100 years of performance for the predictor relative to the actual returns that occurred. 

FIGURE 4: MARKET PREDICTOR FOR 10 YEAR RETURNS SINCE 1920

The dark blue line is the total range of the predictor value sorted from low to high.  The red star represents the predictor’s current value, around -3.5%.  The blue star is where we were 10 years ago. The bright blue line is the best fit to the actual average future returns.  The faint shaded area is the rolling monthly update to the 10 year actual future market returns correlating to that moment in time and thus, noisy.  Although the predictor begins to depart at low values, the future returns do fall into a meaningful range to set expectations.  The next 10 years of average market returns may be in the low to mid single digits, both positive or negative.

No question that we are well into a period of heightened uncertainty about what lies ahead and below average long term returns.  So let’s look at the more details about portfolio allocation.  At the current moment the learning machines are diversified (equities and bonds) close to the old 60/40 rule of thumb.  If one includes the allocation suggest by leverage engine, it implies an added weight to US equities at the moment.    

Exploring the performance of the allocation engines relative to the benchmark, there is a significant potential benefit.  Below is the 1 year returns of the benchmark portfolio over a specific range of interest sorted from -10% to +10% (red line) which occurred over the past 20+ years.  The learning machine allocation raised that average return as shown by the blue line which correlates to the same moment in time associated with the benchmark performance.

FIGURE 5: BENCHMARK RETURN VERSUS ALLOCATION ENGINE

The light blue shadow is the daily possible variation at anytime, which highlights the fact that short term (1 year return sampled daily) data is noisy.

The levered machines do provide an added average boost as well, but at dramatically increased volatility.  As shown below. 

FIGURE 6:  ALLOCATION ENGINE VERSUS LEVERED ENGINE

The blue line is the annual return of the learning machines allocation model.  The green line is the return of the levered learning machine, with the light green daily noise superimposed. In all cases, any 1 year return can vary significantly from a long term average, but the average benefit of the levered allocation machine is substantial. 

It is easy to believe the next ten years are not for the faint of heart.  A lot hinges on what happens in the world and what is done about it.  The expected market numbers are not very comforting, although the learning machines will hopefully help.  Over the past few years of applying those models have improved matters.  Only time will tell if that continues. 

Here is a quick way to think about all the data in the above charts.  Assume likely market returns over the next 10 years will average around 3%.  Assume that’s the best you can hope for with a diversified portfolio.  The learning machine allocation engine could raise that average to around 10%.  The levered engine could deliver almost 30% in comparison.  But using leverage is very risky in general and more so in a volatile or declining market .  The optimal weight for the levered machine relative to the non-levered learning machines, appears to be .25.  So the expected return range for that combination would imply an average annual return range of roughly 15%.  There will always be moments in time when the engines may show no benefit or even a deficit.  That is best seen by those shaded areas in the charts, which are possible extremes at any moment in time. Caveat Emptor. 

All the models and projections are simply a learned combination of metrics that emerge from looking at history and are subject to a changing future.  They could be wrong and do not offer any narrative as to why we are where we are.  

THE WHY’S

So why are we where we are?  There is an interesting symbiosis of using a simple quantitative learning machine to extract an identifiable and higher probability trend, and then prompting an large language model AI to explore possible causes of that trend.  The learning machines of InvestEngines simply explore the quantitative interaction of selected metrics to identify correlations. Current market valuations, the growing impact of technology that accelerates change and global geo-political uncertainty imply greater volatility in the future.  Any explanation beyond that is beyond me.  So I queried one of those new LLMs to suggest what qualitative dynamics may be at play.  

Before presenting those results, I want to make a quantitative (I do prefer numbers to words) observation that may address a key metric of why we find ourselves where we are in 2024 with such social polarization and market volatility. The metric I am referring to is the ratio of median family income to GDP ratio. As long as that ratio holds constant it implies that all boats are lifted equally as the economy grows. If it starts to decline it means a large portion of our population is not benefiting proportionately. Below are graphs of the data.

While the economy may be doing well, not everyone is doing as well.

Below are the unedited responses to my prompts that add a narrative to why we are where we are and are food for thought.  Keep in mind that these new LLMs are subject to errors and making things up, euphemistically referred to as hallucinations.

PROMPT: What are the plausible qualitative reasons that it appears over the past 20 plus years the broad US market has experienced an accelerating series of valuations cycles with an accompanying decline in long term average annual return?

RESPONSE: Several plausible qualitative reasons could explain the seemingly paradoxical phenomena of accelerated valuation cycles coupled with declining long-term average annual returns in the US stock market over the past 20 years:

Shifting Market Psychology and Behavior:

·        Increased focus on short-term gains: The rise of high-frequency trading, algorithmic strategies, and retail investor participation focused on quick profits might prioritize short-term volatility over long-term value, exacerbating cycles and contributing to irrational exuberance during booms.

·        Greater risk tolerance amidst low interest rates: With limited returns from alternative investments like bonds, investors might be seeking higher yields in equities, even at potentially inflated valuations, leading to higher risk tolerance and potentially unsustainable bubbles.

·        Herd mentality and emotional trading: Increased access to online platforms and financial news may amplify groupthink and emotional decision-making, leading to rapid buy-ins and sell-offs, contributing to valuation swings that outpace underlying fundamentals.

Changes in Market Structure and Regulation:

·        Financialization and focus on short-term performance: The growing influence of the financial sector, with its emphasis on quarterly earnings and shareholder value, may prioritize short-term gains over long-term investments, contributing to market volatility and inflated valuations in certain sectors.

·        Declining transparency and potential market manipulation: Concerns about dark pools, short-selling strategies, and algorithmic trading raise questions about market integrity and fair pricing, creating uncertainty and potentially impacting long-term returns.

·        Loosening of financial regulations: Easing of certain regulations, particularly in the lead-up to the 2008 crisis, might have encouraged excessive risk-taking and leverage, contributing to asset bubbles and subsequent crashes, impacting long-term returns.

Economic and Technological Factors:

·        Low interest rate environment: The extended period of near-zero interest rates may have artificially inflated asset prices, particularly in stocks, by driving investors towards riskier assets for higher returns, potentially pushing valuations beyond sustainable levels.

·        Disruptive technologies and innovation: While innovative technologies and rapid growth in certain sectors can justify higher valuations, over-optimism and hype can lead to bubble formation and subsequent crashes, impacting long-term returns for early investors.

·        Globalization and interdependence: Increased interconnectedness of global markets exposes the US market to external shocks and uncertainties, leading to amplified volatility and potentially faster corrections in valuation cycles, impacting average annual returns.

There are others that attempt to offer socio-economic dynamics that may partially explain this moment in time.  Two such authors and books are Peter Turchin’s, “End Times” and Neil Howe’s, “The 4th Turning is Here”.   I queried Bard again about Turchin and Howe, who I’ve enjoyed reading.  Below are the unedited responses to my prompts and represent more food for thought.

PROMPT:  What did Turchin and Howe have to say about the 2020’s in their respective books?

RESPONSE:

Turchin's "End Times" explores cliodynamics applied to our current circumstances.

1. Political Instability and Potential for Conflict:

Cliodynamic models often predict heightened political volatility and potential for conflict during certain phases of the sociopolitical cycle. The 2020s may fall within such a phase, especially considering growing inequality, polarization, and resource scarcity.

Potential flashpoints could include trade wars, geopolitical tensions, and internal struggles within states.

However, cliodynamics also emphasizes the role of leadership and effective institutions in mitigating conflict and navigating these volatile periods.

2. Technological Innovation and its Disruptive Impact:

Technological advancements, particularly in areas like artificial intelligence, automation, and biotechnology, are likely to accelerate in the 2020s.

These advancements will have profound economic and social consequences, potentially exacerbating inequality and job displacement.

The 2020s could be a crucial decade for determining how societies will adapt to and manage these technological disruptions.

3. Environmental Challenges and Potential Resource Scarcity:

Climate change and other environmental challenges are expected to intensify in the 2020s, potentially leading to resource scarcity and mass migrations.

The ability of societies to mitigate climate change and adapt to its consequences will be critical in determining the overall trajectory of the decade.

Cliodynamics suggests that societies facing environmental pressures may experience increased internal social conflict and competition for resources.

4. The Rise of New Global Powers and Potential Realignments:

The current global order, dominated by the United States, is potentially entering a phase of decline.

Cliodynamics suggests that new powers may rise and challenge the status quo, potentially leading to geopolitical and economic realignments.

The 2020s could see increased competition and diplomatic maneuvering between various countries and blocs.

5. The Role of Collective Action and Social Movements:

Cliodynamics emphasizes the importance of collective action and social movements in shaping historical outcomes.

The 2020s are likely to see continued activism and social movements addressing issues like climate change, inequality, and political reform.

The success of these movements in influencing policy and societal change could significantly impact the trajectory of the decade.

It's important to reiterate that these are just broad possibilities, and the actual course of the 2020s will depend on a multitude of complex and unpredictable factors.

However, understanding the general trends and potential challenges identified by cliodynamics can help us better prepare for and navigate the possibilities that lie ahead.

Remember, cliodynamics is just one tool among many for understanding the complexities of history and potential future trajectories. It's always crucial to consider other perspectives, engage in critical thinking, and stay informed about current events to develop a nuanced understanding of our world.

In Neil Howe's "The Fourth Turning Is Here," he draws on his generational theory to predict a period of intense crisis and potential transformation for the United States in the 2020s, culminating in a climax around the early 2030s. Here are some of the broad conclusions regarding possible scenarios for the decade:

Crisis and Transformation:

The 2020s are identified as the Fourth Turning, a recurring historical period marked by societal upheaval, national unity, and a new moral order. The book predicts a significant crisis, similar to the Great Depression or World War II, that will challenge American institutions and values.

This crisis could manifest in various forms, including political and economic turmoil, environmental disasters, or even external threats.

Role of Generations:

Each generation plays a distinct role in the Fourth Turning. Howe predicts that Generation X, accustomed to hardship and resilience, will lead the crisis effort. Millennials, with their idealistic and technocratic tendencies, will shape the new order that emerges after the crisis.

The book emphasizes the importance of intergenerational cooperation and understanding to navigate the challenges of the Fourth Turning.

Possible Outcomes:

Howe describes four potential outcomes for the Fourth Turning, depending on the collective response to the crisis:

High: National unity, revival of civic spirit, and a new American era.

Adulthood: Gradual adaptation and recovery without a dramatic climax.

Prophet's Warning: A dystopian future of increased conflict and societal breakdown.

Fourth Revolution: A radical transformation of American society and institutions.

The book argues that proactive leadership and commitment to shared values can steer the country towards a brighter High scenario.

Criticisms and Uncertainties:

While Howe's predictions have gained attention, they also face criticism for being deterministic and overly reliant on historical patterns. The book downplays the influence of external factors and technological disruptions, which could significantly impact the trajectory of the 2020s.

Ultimately, the specific scenarios outlined in "The Fourth Turning Is Here" remain possibilities, not certainties. The actual course of the 2020s will be shaped by the choices and actions of individuals, communities, and leaders within the complex interplay of historical, social, and technological forces.

It's important to remember that "The Fourth Turning Is Here" offers a provocative framework for understanding potential trajectories of the 2020s, but it's not a definitive prediction. Engaging in critical analysis, considering other perspectives, and staying informed about current events will be crucial in navigating the complex reality of our times.