Regime-based tactical asset allocation - it can add value

A simple paper that focuses on tactical asset allocation based the business cycle suggest that using macro top-down information will be helpful for forming a dynamic portfolio. See "Regime-Based Strategic Asset Allocation". 

The authors break up the macro environment into four regimes: overheating, goldilocks, stagflation, and downturn. Given these environments, different portfolios are formed using just five assets: equities, government bonds, credit, commodities, and REITs. Given these 5 assets, the authors form risk-based and equal-weighted portfolios focused on regime probabilities and compare with an optimized, equal-weighted, and risk budget portfolios. These portfolio constructs suggest that regime-based portfolios can support better risk-adjusted returns.

 

Perhaps macro announcements are not that important

 

Following earlier work on macro announcements effects, there is a paper that states that macroeconomics are associated with approximately half of the equity premium. Now this seems like a large number, but earlier works has argued that 100% of the equity risk premium is associated with selected macroeconomic days and over half the days in the sample may be announcement days. This may be due to sample selection. See "More than 100% of the equity premium: How much is really earned on macroeconomic announcement days?"

When all announcements are included, the Sharpe ratio between announcement and non-announcement days are approximately equal. They conclude that these days are not so special. 

This is a good piece of research, but we do know that some days are more important than others, so a selected sample of special announcement days seems reasonable. If we condition on the size of the announcement, the excess return may be even greater. Nevertheless, this research should temper any investor who thinks he can just buy announcement dates as the road to riches. 

Why focus on macro announcements? That is where all of the return action is

Macro announcements matter and impact returns. If we look at announcement days versus non-announcement days, there is a significant difference in excess returns. This is applicable at all levels of beta.  This work is nicely presented in the paper, "The Macroeconomic Announcement Premium", and shows that this excess return story is applicable for stocks and bonds.

Trade the macro announcements for profit. 

 

What are you playing - the pricing or the valuation game

Aswath Damodaran, the renowned professor of finance at the Stern School at NYU has done a good job of comparing two approaches to looking at asset returns, the pricing game, and the valuation game.

The pricing game is based on the belief that price is the only number that can be acted upon. You don't know the true value of an asset so follow the trend. If prices seem to be moving to an extreme relative to past behavior use that as basis for spread trading. This is the basis for technical trading. 

The value game is based on the belief that an asset can be given a fair value. Find the fair value and use that as the basis to trade. When value is low relative to current prices, buy and when value is high versus prices, sell.

I like this comparison because it is an easy guide to describe the type of traders in the market. Of course, there are hybrids that have the features of both. Similarly, this framework can be used to describe quants and discretionary traders. 

 

Some things don't change with respect to central banking

 

This cartoon is an old trope against central banking, yet it will not go away. The Fed does not have a dual mandate of growth and stable prices, but a third mandate to maintain financial stability which results in policies that protect large banks and other financial institutions. For the Main Street crowd, financial stability through some form of bailout to ensure that bankers don't face the consequences of their actions seems to point to our old cartoon. Cn this perception be avoided? The answer is not clear, but the Fed has generally erred in the direction of supporting banks which have gotten larger than pre-GFC levels. Of course, there is more regulation but that only helps the financial institutions that are too big to fail. The fixed cost is too high for smaller institutions.

Old tropes do not go away because old behavior does not change.

The "New Economic Order" - No more multilateralism

 

The liberal order is dead according to The Economist. The new economic order is completely different with sanctions, tariffs, and restrictions of trade and capital flows. No more world organization supporting free trade and the rule of law. It is more every nation for itself in this chaotic order. The multilateral thinking is dead as countries form bilateral links. 

Sanction have become the norm not the exception, yet these sanctions are often ineffective and create a chaotic system based on changing alliances and not economic efficiency. 

Tariffs have increased under the Biden administration after their use under Trump. Local politics drive decisions, not consumer welfare.

Capital controls and restrictions which limits potential buyers is also the norm, and regulations are placing barriers upon activity across borders. Costs go up and economic efficiency goes down.

The power of international organization has declined and the use of international courts to solve disputes have fallen. The WTO does not solve trade problem in a timely manner or at all.

Simply put, this world order driven by the United States and international organizations like the IMF and World Bank is a thing of the past. Now, some of this order needed reforming but a new world of bilateralism and regional coalitions will not support greater global trade nor raise global income. Delinking will reduce market price correlations.

False consensus effect - we are comforted by being with similar people

 

The false consensus effect is present in finance. You pick your friends, and you pick the people you talk to based on their willingness or their similarity to you. Who wants to be around people that don't agree with you. You have experienced it, "There is no one I know who thinks like that...". Hence, we have the problem of the false consensus effect. We often overestimate how much others share our beliefs. We project our view on others or at the least assume that our circle of beliefs is more widely held than reality. 

Why wouldn't this occur in finance. We hire for the team. We go to the same clubs and events. We are often educated at the same institutions. We remember the negative feelings projected on us when we don't follow our crowd. We will then make decisions based on the belief that there will be comfort being in the crowd with others. We get self-affirmation and validation by being with others that have similar opinions. 

The foundation of non-consensus investing is fighting the false consensus effect. However, it may be easier to just following a model. The validation for model only comes with being correct.

Running for the exits and liqudity spirals

 

What should we fear in the coming months? Liquidity spirals are a thing and if we have markets that move to extremes, there is a greater likelihood of seeing one. A spiral is different from a bubble although the two can be related. If there are some initial loses, especially if positions are highly levered, there will be funding problems. The funding problems can be caused by higher margins or the requirement to add to margin for existing positions. See "When Everyone Runs for the Exists".

If there is some funding issue even as simple of movement of margin, positions will be reduced which will add to any existing losses. There is a positive feedback loop. Feedback loops drive trends, and these trends will create further liquidity issues. This will only get worse if there are crowded trades. The simple issue is that the microstructure of markets can create endogenous risks which generate liquidity spirals. 

Co-momentum and arbitrage; another tool for improving momentum strategies

The returns for the momentum strategy can be measured through the activity of arbitrageurs who create co-movement of returns. When the co-momentum is relatively low, the momentum strategy is not crowded. When there is less crowdedness, the returns to momentum should be positive and not likely to revert. When co-movement is high, there is likely to be a higher probability of returns to mean revert and reduce overall returns. If there is no co-momentum, there is more stabilizing behavior from arbitrageurs. If there is too much arbitrage activity, that is higher co-momentum, there will be overshooting and destabilizing behavior. The behavior of momentum will be time varying which will mean that there is under and overreaction from momentum.  See "Co-momentum: Inferring Arbitrage Activity from Return Correlations"

What is notable is that this behavior with respect to the momentum strategy is not seen with the co-value behavior or the abnormal return correlation among value stocks. When there is not more co-value, there will be higher returns because arbitrageurs re pushing returns to fair value and increasing the returns from holding value stocks. Momentum does not have fundamental anchors, so more momentum trading will lead to destabilizing behavior.

So how is co-momentum measured?  The asset universe is sorted by the past returns as traditionally done with any momentum strategy. The partial correlations for the past year are then found for each momentum decile from the ranking period. The average partial correlation or co-momentum can be found for the loser and winner deciles for some rolling past period.  When the co-momentum is high there is likely to be lower momentum strategy returns.

Momentum crashes and market highs

 

The momentum risk factor will see crashes. This is a well-known fact. After a significant market decline, there is a high risk that those names that are held short will outperform, generate higher gains than the long positions. That is, there will be negative returns from holding shorts as they move quickly higher on a market rebound. It is also found that the crash risk is predictable and can be reduced through dynamic risk management. 

Another interesting feature with momentum is that those stocks that are far away from their 52-week highs are more likely to suffer from crash risk. See "Momentum Crashes and the 52-week High".  The distance away from 52-week highs can be a key tool for reducing momentum risk given this inverse relationship. This condition seems to be associated with market sentiment.  There are simple ways to improve the momentum factors return profile. 

Momentum and moments

Momentum is consistent factor risk and a great strategy for investors. It has proved to be applicable across all asset classes and across time, yet it is not perfect. In fact, it is problematic for many investors because has negative skew and fat tails. This has been described as the tendency for momentum crashes.  In a not so recent paper, "Momentum has its Moments" the other measure the size of the major momentum crashes and looks at ways to cut the risk from these crashes. 

The major momentum crashes will eliminate years of return in a matter of months; however, there is a hope given these crashes seem to follow systematic behavior. Condition on the market environment and the risk from a crash can be reduced.  Momentum has time-varying betas and is sensitive to measures of volatility.  While beta is a contributor, momentum risk is strategy specific so just managing the volatility is a simple way of reducing crash risk.

 

When is good enough - good enough

from The "age of good enough"

One of the ongoing problems with research as well as consumption is determining when is spending or analysis good enough. We always want more but is there a point of diminishing returns. The answer is yes, there is always diminishing returns, but at what point do you stop? It is easy to say that you can do a cost/benefit analysis, but for a good or a service, can you actually measure the benefit? If I can get an extra 10 bps of return out of new model feature as measured in a back-test, is that enough? Is it worth the added work? Is it work forming another condition? 

The question of good enough and how to measure the benefit is worth serious consideration. 

 

Can AI replace stock analysts? Yes, it can

 

The recent study "Can AI Replace Stock Analysts? evidence from Deep Learning Financial Statements" shows that an atheoretical neural network approach using financial statement information, stock prices, and interest rates can do as well or better that the predictions of stock analysts. Given a lot of information and without making any assumptions on how to use the information, the NN approach can beat the analysts. 

However, the story is a little more complex than just letting the computer do its work. When the analysts compete against the NN model which only uses fundamental company information, the results are in favor of the analysts. When the NN is able to use both interest rate and stock price information, it does much better. 

The AI model is able to capture the combination of stock fundamentals with some macro information and price behavior. The value-added comes when the computer is able to blend all these features together.

Momentum crashes - always a fear

Momentum has been found to exist oil all asset classes, yet there are risks with this core strategy. Momentum portfolios may be subject to crashes - sudden declines in return which makes holding momentum portfolios risky. See "Momentum Crashes". 

In a state of panic or during periods of high volatility, the prices of past losers will have a high premium. When the markets conditions change and there is a rebound, losers will see strong gains which will lead a crash for those assets that are held short. Down market betas of past losers re low and the up markets betas are very high. This type of optionality may not be priced in the past losers. Given these conditions the past losers now have high expected returns which is what an investor does not want to hold. Unfortunately, this behavior does not apply to the winners during good times. Hence there is an asymmetry in the exposure of winners and losers during extreme times. 

Given this pattern is predictable, it can lead investors to dynamically adjust their exposures during these periods to adjust the loser's portfolio and avoid these sharp reversals. It is notable that the crash risk story seems to serve as a good story for equities but is harder to apply to other asset classes such as bonds, currencies, and commodities. 

The presence of crashes with momentum is a good reason to manage downside with stop-losses and to manage volatility exposure. Momentum may not be a factor that follow set-it and forget-it strategy. 

So you think you know the repo market?

 

What is one of the greatest fears of the Fed? Financial instability and not inflation. They are afraid of a liquidity crisis. The decision to lower rates is relatively easy versus the decision on what to do about liquidity or the cut in liquidity from a change in the Fed's balance sheet. This is why the Fed is planning to cut the QT program. 

The Fed does not want a repo crisis and the likelihood of a crisis is associated with the plumbing. The graph above is hard to read but it traces all of the pipes with how money and lending moves through the repo market. Thanks to concoda for their fine work.

This structure will change with centralized clearing as mandated by the SEC; however, that does not change the fact that liquidity is a chief concern of the Fed.

Dynamic rebalancing works when dealing with alternative risk premium

 

Using six factors or smart betas (value, low beta, profitability, investment, momentum, and size) the folks at Research Affiliates tested different portfolio rebalancing approaches and found that a dynamic rebalancing method works best. (see "A Smoother Path to Outperformance with Multi-factor Smart Beta Investing". The research compares buy and hold with systematic rebalancing which moves allocation back to the equal weighted each quarter, and a dynamic rebalancing which allows for modest tilts based on short-term price momentum and long-term mean reversion signals at each quarter. 

Factors follow time series patterns which can be used to help with rebalancing. In the short run, follow the trend; however, if there is a large divergence from long-term averages, cut exposure. This will provide a smooth return path.

What do central banks think about inflation - look at gold

 

Global inflation debases currencies. Higher inflation in the US debases the reserve currency. If the inflation is transitory, central banks should not change their exposures to different currencies; however, if you believe that inflation will last for a longer time, it is time to buy hard assets. Even though we are off the gold standard, and it was described by Keynes as a "barbarous relic", central banks are buying a lot of gold - tonnes of it. 

2022 and 2023 were banner years for gold buying. We believe that this is also sanction related. There is less reason to hold dollars if there is chance of sanctions reducing your ability to use those dollars. Look at the world official reserve assets in gold. It has moved from about 950m oz to above 1150m which is close to the levels seen when the world went off the gold standard. Despite high real rates and falling inflation, central banks want gold.  Central banks are voting with their gold and they do not want to hold currencies that may lose their purchasing power.

VUCA and the investment world

 

from Jeremiah Genest 

We love the term VUCA - Volatility, Uncertainty, Complexity and Ambiguity which has been used for decades to describe the environment that decision-makers and leaders face. It encapsulated all the problems that a lease face in a simple framework of how much you know and how well you can predict outcomes. The graph below gives a nice visual of the problem. 

However, the VUCA framework has been applied to finance and investment problems. I am not sure what this is the case. Perhaps it is a just different and not the say that investors are taught how to look at the market environment. Perhaps more importantly, the definitions associated with volatility, uncertainty, complexity, and ambiguity are not the way that people in finance and statistics think about the environment and risk. Still, the VUCA issues are discussed in finance; however, there is a not a framework that ties these issues together. 

Volatility is easy for us to measure and discuss. The meaning of uncertainty gets to the heart of the definitional issues. What is uncertainty as separate from risk is not an easy issue to define. I am in the camp that risk is what is countable, and uncertainty is what is non-countable. This could be view as a distinction between objective and subjective risk.  Uncertainty, through Frank Knight, has been described as those events which are uninsurable. Again, the current VUCA definitions do not translate well into finance.

Complexity is another concept that can be discussed but does not have a good definition within finance. The same can be said for ambiguity. What is an ambiguous environment will lead to several different definitions. 

We think the VUCA framework if it can be adjusted to the finance and investment work will have significant usefulness for improving the discussions concerning risk.

Frank Knight and uncertainty

 

How should investors deal with uncertainty? Prior to dealing with uncertainty, there has to be a definition of uncertainty relative to other terms like risk. This is an old question but one of critical importance.  The controversy or discussion about the difference between risk and uncertainty goes back all the way to Keynes and Knight in the 1920's. 

Some have viewed risk as referring to known or knowable probabilities. These would be objective probabilities. Uncertainty is where the probabilities are not known or cannot be deduced or counted and are thus subjective. This distinction was addressed by Savage in the 1950's with the answer that if subjective probabilities follow the simple axioms for any expected utility framework, the problem can be solved. That said, Knight was accepting of the combination of objective and subjective probabilities. He focused on another problem associated with uncertainty.

Knight made the distinction that risk is what can be insurable while uncertainty cannot be insured. An insurable event generally would mean that we can have data to help us determine frequency, but insurance can also be provided for events that are not countable or have subjective probabilities as long as we can define correctly the likelihoods for an event. Knight argues that uncertainty is when there is no insurance market for the event. This would be a market failure.

In Knight's view profit comes as the reward for bearing uncertainty where there is no insurance. Innovation and technological change are uncertain because it is not insurable. Management action is often uninsurable because we cannot measure properly.

Profit is the residual between revenue and costs. The cost will include the risk premium or the cost of capital. In a normal market, profits will be zero because the residual is zero after accounting for all costs. Profit is a reward for bearing uninsurable events or hazards and that would be events that can be called uncertain. Entrepreneurship is uncertain because you cannot separate bad luck from bad decisions. The outcomes cannot be determined in a way to make them objective in form that can be insured. 

While I appreciate the distinction between risk and uncertainty and the difference between objective and subjective probabilities, they're still the problem of how these subjective and objective probabilities are formed. Objective probabilities are still a problem when trying to form proper sampling. Some events do not have a large sample so there is the issue of forming subjective probabilities. These problems are present event before getting to the distinction between risk and uncertainty.

Multiple models versus an ensemble

 

"A man with two watches is never sure (what time it is)." 

- Segal's Law 

There is the view that having competing models is a bad thing. Investors should form one model that incorporates all their thinking. If you have more than one model you will never know or believe what is the true reality.  Yet, we also know that models will fail. We will make mistakes. Most models will explain only a small portion of the variation in asset returns.  

Yes, having a single model is theoretically pure, but I will usually choose to form an ensemble. An average prediction from an ensemble will usually do better than a single model. The reality is that prediction is about getting it right and if there is a means that is not perfect but gets us closer to a better answer should always be preferred.

The problem with SEU subjective expected utility

 

How do investors deal with uncertainty? This is a critical question, perhaps the key question for investing, yet it is often overlooked. Of course, there are many answers to this question, but the fundamental problem is how do investors form expectations about uncertain events.  The answer to dealing with uncertainty is to form better expectations of the future.

The investor must first isolate what is the event, and second, the investor must form some probability associated with the likelihood of that event. In return space, the problem can be simplified through just giving different returns some probabilities and have the probabilities sum to one. Once an investor tries to describe specific events, the issues get much harder. 

The response from a technical perspective is to use some form of subjective expected utility (SEU) as a framework. The problem is just multiplying probabilities times events with a set of rules for behavior to generate the probability formation. We have all used this type of framework, yet there is something missing when trying to apply it in real life settings. 

Where do those probabilities come from? This is again easy if you have countable events, that is, if we use some probability distribution that comes from the sampling of past return. The problem is less tractable if we are trying to incorporate events that have not occurred in the past within our sample of return. 

The problem also becomes more difficult if we try and map events or shocks into the return space. If we have an event, say an earthquake or hurricane, we must both think about the likelihood of the event, the type of event, and the mapping of the event to specific asset returns. It may sound easy; however, the work involved is difficult. This is the process that is often not discussed in the classroom. 

So how are probabilities formed for non-countable or rare events? They are subjective and rely on the decision-maker. We assume that he will be rational and have a process that is consistent, but beyond consistency, there is little work on how these subjective likelihoods are formed.

 There is uncertainty over the process of forming likelihoods of uncertain events. Perhaps that is why rules of thumb are developed and we try our best to in the words of Herb Simon satisfice. 

The illusion of financial skill - most don't have any

 

One of the the big issue in asset management is the illusion of financial skill. We cannot truly assess if we are good managers if we have the feeling we are good managers. We cannot make good decisions, if we have more confidence in our abilities than exists. There is often described as the cognitive bias called the Dunning-Kruger effect whereby individuals who perceive themselves as experts will have the illusion of superiority concerning their cognitive abilities. 

In his book Thinking Fast and Slow Daniel Kahneman describe an experiment using 25 wealth advisors to see their correlation between their performance from one year to the next. That is, do the wealth advisors have any skill. The results were surprising. There was zero correlation between their performance rankings through time. We can not say whether these advisors thought they were better than others or that they deserved their bonuses, but we can say that they did not have nay persistent skill. 

Whether discretionary or quantitative, we need  measures to tell us whether we have nay skill. If you are going to play the investment game, you need performance measurement and a feedback mechanism to learn from mistakes.

The hazard of experts - How do you avoid this hazard effect with hedge fund managers?

The problem with quant models

The development of core factor investing

 

First there was the CAPM factor which failed as researchers found that low beta stocks had higher returns and high beta stocks had lower returns than what would be expected with the CAPM model. It is a good theoretical model but its ability to explain cross-sectional returns is limited.

This led to the development of the Fama-French three factor model which included market risk, size (SMB) and value (HML) factors. This was a significant improvement and changed the way investors thought about risk.

From this framework Carhart added the momentum factor (UMD) or (MOM) which created the four-factor model. This now caused a significant amount of finance confusion. How can past performance predict or tell is something about relative returns" There is acceptance hat momentum is present in all assets classes and that is a fundamental risk factor albeit there is a still a view that this is a behavioral problem that should not exist. 

However, there was a desire to find more factors based on economic theories which were developed through the papers of Zhang et al. who found what were called q-factors which included operating profitability, ROE, and investment or a real investment factor or asset growth. 

Fama and French extended their 3-factor model to include operating profitability robust mins weak (RMW) which is measured by revenues COGS - interest expense SGA scaled by book value, and investment, conservative minus aggressive or CMA which is just asset growth scaled by total assets. This led to the quality factor as a key addition or for some a better interpretation of value.  

Since these core factor developments, there has been a zoo of factors to describe many risk premia. Many of these factors have not stood the test of time, but it shows that the search for return drivers is a dynamic and ongoing process. However, the core work of market risk, size, value, momentum, and quality are now the key factors for any discussion of equity returns.