Monday, November 12, 2018

Evolution and adaption: Trend-followers are constantly changing


Hedge funds styles, strategies, and firms evolve over time. The behavior of a hedge fund today is not the same as yesterday. These behavior changes are not because a manager has changed his style but because the environment, the tools, the regulations, and the ideas surrounding finance are different. 

Of course, many of the guiding principles for a hedge fund are the same. Value managers attempt to find cheap assets. Trend-followers attempt to find trends. The evolution is generated through changes in technology, the structure of markets, and the implementation of strategy ideas.  Technology implementation allows for cutting costs, enhancing skill, and gaining scale. The structure of markets requires managers to evolve. New ideas and knowledge help generate or improve skills.  

Managers who are not thinking how to adapt will see their returns deteriorate and ultimately fail. Technology not used is a lost opportunity or places a manager at a competitive disadvantage. Ideas not employed hampers skills. Avoiding adaption reduces opportunities and increases costs. Adaption is core to good hedge fund management, yet being adaptive to changes may not lead to higher returns. Competitors are doing the same thing. Adaption may be required just to stay even with other managers in the sector. Some structural changes may lead to lower returns

We have discussed the evolution of trend-following in the past, (see The evolution of trend-following firms to alternative risk premiums and quant shops), but the current environment requires more thought about firm evolution. Limited crises over the last ten years and greater competition means trend-following has seen a more constrained opportunity set. Managers have to adapt to exploit or enhance these existing opportunities and prepare for changes as we move further away from the last crisis.

The evolution of trend-following has followed a set of three intersecting paths: technology, structure, and idea generation. These intersect because some ideas can only be implemented if the technology is available, and some technology can only be used if there is a structural environment that allowed implementation. Investors need to ask how firms are addressing these intersections.

Technological improvements from computing power have been especially relevant for quant strategies like trend-following. The computing power has been directed into four areas: 
  • The use of speed to test new strategies through back-testing - any idea is quickly testable.
  • The low price for storage - any amount of data is easily stored and available for review.
  • The ability to electronically trade and parse orders to reduce transaction costs.
  • The ability to process operations and manage risk - process and overhead can be reduced and position knowledge is readily available.
Idea evolution has allowed for better return opportunities and risk management; however, the broad use of new ideas diminishes the marginal edge given to any one manager. The major ideas that have affected trend-following include:
  • The risk management revolution including VaR modeling.
  • The improvement of portfolio management including volatility targeting and equal risk contribution.
  • The use of new statistical tools for teasing out time series behavior.
  • The ability to engage in complex non-linear thinking like machine learning.   
Structural changes have reshaped the set of opportunities for managers:
  • The introduction of new derivatives markets - Trend-followers have more markets to trade than before.
  • Changes regulation that allows new products  - Regulation has allowed for new fund structures that change the investor base.
  • Changes in policy affect price behavior - Changes in monetary policy has affected the behavior of rates which impacts trends. Changes in capital regulation opens new markets. 
  • Changes in industrial structure - The concentration and behavior within industries affect price opportunities.
An assessment of how firms are dealing with change may provide answers on which firms will be able to come out on top in a changing investment world. Unfortunately, firm adaptation is not a guarantee for higher returns when we are in a competitive environment.

Sunday, November 11, 2018

Blending risk premia and generating craftsman alpha


Alpha generation will fall when it is measured correctly through an appropriate benchmark. Alpha shrinkage over the last ten years is a measurement problem. Returns for hedge funds are a combination of the underlying risk premia styles employed and the skill of the manager. 

This shrinkage seems to suggest that managers generate little extra return through their skill, but there are other forms of alpha associated with forming a portfolio. This has been called "Craftsman Alpha" (See Craftsman Alpha: An Application to Style Investing). The crafting or forming of a hedge fund portfolio is a unique skill and can provide value no different than security selection.

The definition of craftsman alpha is still somewhat vague, but it will include all of the activities associated with portfolio management after a style choice is determined. Craftsman alpha will be the value-added from bundling and managing a portfolio of risk premia. 

We break craftsman alpha into four categories or parts:
  • Risk premia style choices -
    • Implementation of styles: Given any risk premia style, there are a number of implementation choices, asset restrictions, inclusions and exclusions, and definition differences, which determine how a style is generated. For example, a FX carry strategy has to determine the currencies to include, the rate to determine carry, weight constraints, and rebalancing to name a view. These choices, all under the name FX carry, can have appreciable return differs and can be classified as skill. 
  • Portfolio management choices -
    • Volatility targets; determining the overall risk of the portfolio 
    • Rebalancing timing; determining when to reset the weights of the portfolio. 
    • Sector and name constraints; determining maximum allowable exposures.
    • Weighting scheme; determining the weights of exposure such as equal volatility weights vs equal risk contributions. 
  • Execution choices -
    • Mechanisms for minimizing transaction and trading impact.
  • Dynamic Adjustment choices -
    • If there are multiple risk premia in the portfolio, the decision process or mechanism for adjusting the portfolio weights.
A craftsman alpha discussion changes the focus from picking securities or risk premia to the process of managing a portfolio of risks; the strategy and tactical decisions of running a portfolio. Given there are no well-defined rules on how to create craftsman alpha, there can be significant variation across managers. These construction choices are the decisions of a craftsman and not scientist. 



  

Friday, November 9, 2018

Alpha production - As we get better at beta measurement, alpha will decline


A growing investment management theme over the last few years has been the incredible shrinking alpha. As investors gain more information on risk premia, there seems to be less alpha produced by managers. In reality, alpha production has likely not changed, but our measure of alpha has gotten more sophisticated so the skill associated with any manager seems to have declined or at least changed over the last decade. We can now tie what was previously thought of as alpha to specific risk factors. If alpha is tied to systematic risk factors, then it really is not alpha.

The measurement of shrinking alpha can be described in an alpha pyramid. As investors better define portfolio risks, the skill of the manager will shrink. It is harder to prove skill when we account for risk premia correctly. (See our previous post, The incredible shrinking alpha - Falling skill or alternative definitions?) It is hard to get to the rarified place of skill after accounting for risk factor premia. 

It can be affective to think through the process of moving up the alpha/beta pyramid. A stronger beta filter will reduce the number of managers who are special alpha producers. As we enhance our skill at measurement, the manager alpha skill has been more difficult to find.

If the majority of returns are associated with risk factors, there will be less room for alpha. If this is true, then building risk premia portfolios may better serve the diversification and return needs of investors. Find the betas that will enhance the return to risk of the portfolio and don't worry about finding that special skill manager. Of course, if you can find the high level manager hold onto him; that manager can indeed be very special. However, the search should be on finding the best risk premia mix.


Thursday, November 8, 2018

25 years after Jegadeesh and Titman - The Momentum Revolution

Trend-following and momentum has always been an important part of hedge funds and alternative investing but it would be hard to say that trend-following was mainstream thinking prior to the early 90's. This was the high water period of the of market efficiency, but that thinking started to take a major change with the "Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency", published in the leading Journal of Finance. There were other papers that discussed similar topics and the behavioral finance paradigm shift had already begun, but this was the one paper that many academics started to quote with increasing frequency about momentum effects. 

The paper was elegance was its simplicity. It showed that if you followed a strategy of buying past positive performers against a portfolio of losers, there would be significant excess returns. The tests were done using past returns not trend but the results transfer. It was not a paper about technical signals or moving averages. It just looked at past returns with significant look back.

We are now 25 years away from this path breaking work albeit it is unlikely that most academics or traders will hoist a beer to the authors in celebration. For core followers of trend, this is no big deal. The celebration of trend-following is much older, but for the mainstream this is as good as any place to mark a change in investment paradigm.

25 years later, momentum and trend-following are a core part of investment thinking. These are not just considered investment strategies but fundamental risk premia. The discussion has moved from thinking about ways to dismiss these risk premia to offering reasons for their existence. What trend-followers knew but may mot have clearly articulated is that behavior creates slower reaction to news. Biases drive trends. 

Nevertheless, some of the terms and language has changed to suit academics over older practitioners. Academics like to use the word momentum and askew trend. There is now a clear distinction between cross-sectional and times series momentum. The classic trend-follower will think the cross sectional approach is a form of ordering trend preference. No one uses the words technical analysis. Preference is for quantitative analysis and algos.

25 years into momentum and trend style acceptance, researchers have looked at an ever-increasing set of data over markets and time to show that trends exist. Of course, there is an ebb and flow with returns associated with these strategies but over the long run, you can go to the bank that the risk premia is present.

Can there be too many following this risk premia? That is, can popularity kill the golden goose of momentum? From a theoretical level, as long as there are trends in fundamentals, behavioral biases, limits to arbitrage, and uncertainty, there will be frictions that allow for trends. From a practical side, trend behavior and returns will change with the time length of trends, speed of reaction, differences in crowding, congestion, and liquidity. Momentum/trend returns will grow and decline which will force some to leave the strategy and others to jump in. There is a dynamic environment which will ensure that everyone cannot be a winner at all times, but for the patient, momentum/trend should work over the next 25 years.