Posts published by "Orpheus":

Why is CAPM not CRAP?

What is CAPM?

Capital Asset Price Model of 1961 says that expected return on any portfolio (or stock) should earn a premium above the risk free rate. In simpler words it said, low risk meant lower return and vice versa. The CAPM was introduced by Jack Treynor, William Sharpe, John Lintner and Jan Mossin, building on the work of Harry Markowitz. Sharpe, Markowitz and Merton Miller got the 1990 Nobel Memorial Prize in Economics.

Why CAPM is, was and will be important?

CAPM is an asset pricing model. Before we invest in the assets, we need to price them. Pricing models are beneficial in asset selection and in modelling market behavior. We can create financial innovations only if we can map market behavior.

Who called it Crap?

James Montier (1990) rechristened CAPM as CRAP (Completely Redundant Asset Pricing) in a research paper. No doubt behavioral experts had insights into the market behavior, but somewhere there is an “Academic Bias” that creeps in, making academicians be more positively biased about their body of work. History is full of literature where new academic theorists have not been very objective about the previous body of work. Mandelbrot called the bell curved non-sense; Fama said how this stuff (Behavioural Finance) ever got published etc.

Why did Montier say that?

Apart from the fact that Montier wanted to justify the “Academic Bias”, the author strengthens his case against CAPM assumptions by illustrating the low beta and high beta portfolio behavior. He illustrates Fama and French’s 2004 review of CAPM.

“Each December from 1923 to 2003 they estimate a beta for every stock on the NYSE, AMEX and NASDAQ using 2-5 years of prior monthly returns. Ten portfolios are then formed based on Beta and the returns and tracked over the next 12 months. The figure plots the average return for each decile against its average Beta. The straight line shows the predictions from the CAPM. The model predictions are clearly violated. CAPM woefully under predicts the returns in the low beta stocks and massively overestimates the returns in high beta stocks.”

This might suggest that investors might be well advised to consider a strategic tilt towards low beta and against high beta - a strategy first suggested by Fishcher Black in 1993. Suggesting simply that low-risk could deliver higher – return and vice versa.

Is everything wrong with CAPM assumptions?

According to Montier, investors prefer to benchmark to market returns rather than to risk free rate. This is why investing community is more focused on tracking error than on risk free interest rate. Which is correct, but having the risk free rate as a benchmark may not make it realistic, but it does not make it redundant.

Do Fama and French make CAPM redundant?

Fama and French improved the model by adding value, size (capitalization) variables to the CAPM variables. Though testing suggested that the new variables enhanced the understanding of the market behavior, the model was still offering better guidelines to understand asset prices but was still not unequivocal in its findings. Even newer models with momentum as a variable failed to establish rules and relegate CAPM into redundancy. The model still worked in a few cases and was still valid.

Is CAPM failure, a mean reversion failure?

So if CAPM was still valid and far from completely redundant, why was it crap? Did we need to look at CAPM beyond the “academic bias”? The history of financial markets has limited data compared to many other natural systems. A lot of our financial models are still looking at snapshots of data, rather than studying any dynamic evolution in market behavior. A lot of data interpretation focuses on causally explaining mean reversion failures, or simply putting divergence from idealized cases. This is why when a divergence from CAPM, made CAPM a poor idealization. We continue to seek better idealized scenarios, but somewhere we forget that markets are not made of one idealization, but a set of idealizations. In this case both CAPM and Fama and French being two sets of idealizations.

Is it not all about divergence?

If it’s about mean reversion failure, it’s all about models failing to explain divergence. Could it be that simple? This is what we explained in our re-take on Thaler’s “End of Behavioural Finance” that how behavioral finance was a psychological explanation of cases of mean reversion failure.

The power of proxy

In our paper on data universality we explained the power of proxy, and how data behavior is universal irrespective of the variables be it financial or non-financial. A simple performance ranking can be a good proxy to explain value, growth, momentum, reversion, low beta, high beta, volatility etc. in a certain universe of assets prices or simply any natural data set. We took a proxy percentile performance ranking of worst (bottom fifth) and best in a group of assets (top fifth) for the S&P 100 components. The test was made for 20 day to 1200 days. And even after 1200 days of holding nearly 20% of the worst losers and best winners continued to remain worst and best respectively.

This proved that though there was a tendency for the worst to outperform and for the best to underperform, this was not a rule. This could be extended to the idea of low beta stocks could continue to stagnate, which in other words meant that low-risk could continue to deliver low-return or simply suggesting that CAPM was not crap, but a relevant case of market behavior.


RMI India Active 10 Feb

After you select the RMI model which fits your risk preference, the key question investors ask is “How should one come on model?” RMI Models are customizable for any risk preference and do not suffer from starting point bias. So it can be customized from wherever you begin. The RMI India Active 10 initiated in February has gone 70% invested. And considering the average holding period of any RMI Active 10 style for most global regions is around 200 days, we might just have our leading bullish signal for the Indian market pre-election. RMI Active is absolute capital conserving models which are designed to outperform their respective universe. RMI Active India 10 back-tested model has delivered 23% annualized at a 6% lower volatility than the respective top 100 India universe benchmark.

Enjoy the latest RMI Active India 10 February


IBB Nasdaq Biotech up 100%

 

There are various ways to select a big sector winner. Any sector ETF investment which delivers more than 30% over 12 to 24 months can be considered a big winner. We can adopt various tools and techniques to do so; inter-market analysis, macro economic analysis; sentiment analysis or we can look at the RMI ETF Active 10.

The model selected the NASDAQ Biotech Index back in 13 April 2012. The sector is a big winner with 118% upmove since the point of entry. Now ofcourse we can ask how did the model manage to select the respective winner? How did it know? The RMI Active Indexing solutions use the best of momentum and reversion approach to select out of the globally top traded ETF. And IBB was not the only big winner. PJP Powershares Pharmaceutic was the other above 100% winning trade. Above this the model had more than 5 above 30% movers.

The universe contains North American ETFs, which is why we have benchmarked the group to a composite of TSX and S&P500. The benchmark delivered 12% annualized with a standard deviaion at 13%. The Active 10 ETF outperformed the benchmark delivering 14% annualized with a standard deviation at 12%.

Enjoy the latest RMI Active ETF.


BRIC Model from a Japanese Perspective - II

This was a paper first presented at the EAECS Asian workshop in Kyoto Sustainability and Future Perspectives of Emerging Markets,  Kyoto Institute of Economic Research in 2009 by Ioan Alin Nistor and Mukul Pal.

In the 2009 paper we illustrated that relative performance was cyclical. We used a simple indicator to illustrate a relative performance cycle. We took relative performance RP (ratios) between BRIC countries and Nikkei and then did a Rate of Change (ROC) on the relative performance. This long term ROC like any other indicator moved from one extreme to the other. Using an average period between two ROC lows, we projected the movement of the ROC indicator into the future, where will it top and when will it hit a new low again. This gave us an estimate of peak out-performance and under-performance low. In the presentation we made seven forecasts till date (the eight one is due in September 2015). All the seven forecasts turned right and could predict the relative performance dynamics between BRIC countries and Nikkei objectively.

Unlike any other indicator, a Rate of Change of Relative Performance over longer periods of time exhibit objective seasonality, which can be used for anticipating out-performance and under-performance between the pairs under study. We believe that such a simple time cycle indicator could be employed to understand other data pairs series like Gold vs Commodities (OIL, GAS, Wheat, Corn etc.); DOW 30 vs DOW Key Sectors; USD vs Other currencies; US Bonds vs Other Bonds; 30 year Bonds vs other term yields.

What does ROC on RP workability prove? It showcases that performance is cyclical and even if the average time frame of performance turning from out-performance to under-performance may wary and differ for different pairs, the fact that performance can be measured just with such a simple indicator suggests that long term seasonality is a phenomenon which may not change with short term information or drivers. It also suggests that fundamental, behavioral, or any other external information may drive performance seasonality between pairs at a shorter time frame but not influence the long term direction of the trend. The long term trend can not be causally explained and is more ordered than believed.

Excerpts from the Presentation Paper in 2009

Starting with the fundamental idea of an “emerging market economy”, it’s role, utility and dynamics in the current global set up as a balancing economic block, the paper analysis Goldman Sach’s emerging BRIC’s countries model in context of the pre and post 2008 financial crisis. The paper looks at micro and macroeconomic valuations, currency and the economic cycles to illustrate changes in the four economies. Using Japan as a developed economy, the paper also makes a comparative approach and tries to forecast the economic development of the block and respective relation among these countries.

In the following analysis the authors make a comparative approach of the BRIC block with Japan, using the Stock Exchange Index of each country as a base for comparison and forecast. The indices used are: - Japan: The Nikkei-225 (N225) Stock Average is a price-weighted average of 225 top rated Japanese companies listed in the First Section of the Tokyo Stock Exchange.- Brazil: The Bovespa (BVSP) Index is a total return index weighted by traded volume and is comprised of the most liquid stocks traded on the Sao Paulo Stock Exchange - India: The Bombay Stock Exchange Sensitive Index (BSESN - Sensex) made is a market capitalization weighted index. The selection of the index members has been made on the basis of liquidity, depth, and floating-stock-adjustment depth and industry representation. - China: The Shenzhen Stock Exchange Component (SSEC) Stock Index is a Capitalization Weighted Index. - Russia: The Russian Trading System Index (IRTS) is a capitalization-weighted index. The index is comprised of stocks traded on the Russian Trading System and uses free float adjusted shares.

We believe that unlike the last time a rise in commodity prices now may not see a similar action with both Brazil and Russia rising together. Rising commodities are not good beyond a certain level for the underlying growth. We think a reprieve in commodities this time around will benefit Russia more than Brazil and China being the world’s manufacturer will suffer more than India till 2015. We compare the entire BRIC region with N225 and observed that owing to these underlying structural problems Nikkei should outperform both China and Brazil. India and Russia on the other hand should relatively outperform Nikkei. One on side this may look strange, but as we know that tough times bring out real outperformance. Russia owing to its large energy basket and sizeable correction from historical high levels in 2008 will also benefit owing to base effect just like it benefited in 1998 after the Rouble crisis.

In conclusion the authors believe that there is an intricate balance between the world order and relative outperformance against China and Brazil will keep the Japanese growth engine sustain and grow contrary to popular belief. This should also lead to both actual and relative growth in Nikkei and even the underlying GDP growth for the Japanese economic zone.

Table– Performance cycle time projections and results

Fig 1 – N225 vs. BVSP ratio performance cycles on time projections

 

Fig 2 – IRTS VS. N225 ratio performance cycles on time projections

Fig 3 – Sensex VS. N225 ratio performance cycles on time projections

 

Fig 4 – N225 VS. SSEC ratio performance cycles on time projections.

 

You can read the complete previous paper on SSRN.


The Predictive Bias

The Predictive Bias

In case you got the annual forecast wrong, don’t punish yourself. The recent publication by Jason Hsu suggests, “2013 was not the year for Nobel worthy investment ideas”. If the Nobel prize winners can have a poor annual forecast, does forecasting accuracy need to be re-understood and re-defined?

CAPE Failure

Index dividend yields and cyclically adjusted P/E ratios (CAPE’s), among the other aggregate variables can predict future equity returns. According to Hsu, “Though Fama and Shiller are on the other side of market efficiency; both conclude that market valuations ratio forecast five year returns with satisfactory accuracy. High dividend yields and low CAPEs tend to predict above-average future returns. Conversely low yields and high CAPE’s signal below average returns; “at some point of time”.

According to the paper, “The Shiller CAPE spectacularly forecasted the carnage of the 2000 tech.” One of the goodwill drivers for Shiller; a forecast. “The CAPE was at 21 for the S&P 500 Index in Jan 2013. And given that it had a recent trend (high) at 22 and a long term average at 16.5, the US equities appeared to be neutral to extremely overvalued. The US equity’s market returns of nearly 30% in 2013 had pushed the CAPE toward 25. This appears decidedly expensive relative to the recent and long term levels.” Emerging equities on the other hand were CAPE cheap. “Their -1% performance in 2013 has been one of the big disappointments for global equity investors.”

The paper questions if we should shrug the 33% out-performance of US equities over EM equities as a fluke or an outlier? Should we double down for 2014 by further re-balancing away from US equities towards EM stocks? ”

Technical analysis 101?

What does the Hsu’s paper tell us regarding the Shiller CAPE indicator? First; The indicator was at an extreme. Second; Extreme suggested overvalued markets. Third; because the extreme worked last time in 2000 making Shiller a household name, it should have worked again. The indicator was supposed to revert and bring the markets down with it (or underperform), which did not happen.

Now there are assumptions here in CAPE and in it’s interpretation; First; CAPE as a measure is good and will predict; Second; Extremes in CAPE are bound to reverse; Third; Extremes though subjective should work objectively, and work as a timing indicator annually.

Are the Fundamentalists and behavioral experts behaving like longer trend anticipating technicians. The indicator has reached an extreme, and it should turn now, bringing in under-performance. This is technical analysis 101; so prone to failure.

The Predictive Bias

Give us any measure, be it Value, Growth, Beta, Momentum, Reversion, CAPE or Dividend etc. and we can show you cycles of underperformance and outperformance and failures of the same across multiple time frames (3 months to 60 months or more).

The fundamentalists or behavioral experts are trying to prove, how their respective variable (indicator) reverts back to the mean. There is a clear coverage bias. This is my coverage, my discipline, my skill, my theory; which works. This solo-discipline work keeps us away from a holistic interdisciplinary approach to risk and keeps us glued to our genetic predictive bias, which is inherently rooted in mean reversion. If it has gone up, it’s going to come down.

End of Behavioral finance

In our take on Thaler’s, “End of behavioral finance” paper we illustrated how Behavioral finance is biased towards explaining all mean reversion failures as behavioral anomalies, while mean reversion failures are happening all over nature.

The Total forecast

There is no “total forecast” and just because mean reversion did not work for Shiller CAPE in 2013 and worked in Fixed Income for the fundamental Indexing team does not give mean reversion a predictiveness. Mean reversion is prone to failure. This is why we see a maze of agreement and disagreement between conventionally different disciplines of study; convergence and divergence.

Markets are more about risk management than about predicting value vs. growth; momentum vs. reversion; high beta vs. low beta; CAPM vs. Fama and French etc. We are no more in a single discipline; forecasting (risk) is an interdisciplinary science and till the time we reach there, mean reversion failure will keep us lost in our biased prediction models.


Orpheus Global Webcast

The Educational Web Series is a webcast seminar held at least 3 times a month featuring recognized industry professionals in a one hour long presentation – free to our membership! In the past, we’ve had such noted technicians as Martin Pring, Ned Davis, Dennis Gartman, Thomas Dorsey, Charles Kirkpatrick, CMT, and Ralph Acampora, CMT. From this page, you’ll get a glimpse of our immediate upcoming schedule and even have the opportunity to register yourself for one of these live webcasts.

5 Feb 2014

Momentum and Reversion

Momentum and Reversion are considered two different strategies, styles of investing. A few even believe that 25 years of research has failed to marry Momentum with Reversion. It’s the same way a few researchers feel about value and growth, or low or high beta etc. What if a framework could explain the divergence between M&R and how they can be redefined and understood? This could open up a new approach to signal identification and classification. The webcast will build a case for explaining M&R using Intermarket Analysis, Performance Cycles and other Behavioural Finance cases. The talk will showcase a new framework based on data universality and how M&R can be redefined, comprehended and applied for trading and investment.

Speaker Profile

Time is a social construct and we see time through the life and nature around us. Understanding Time could give a unifying theory to research of a few thousand years and also bring more than a conventional thought down. It’s a revolution. Mukul has written and spoken globally on the geometry of TIME, patterns, risk and investing; has a data innovation patent filed in his name; is an enthusiastic R data scientist and runs Orpheus Risk Management Indices, a global Indexing company. The company has built and manages the multi strategy and multi styled Indices. Mukul is the author of “Risk Management Indexing”, a book on the new Indexing approach. He is also a ranked author on the Social Science Research SSRN network. As a speaker he has been invited to speak at various platforms like the Bombay Stock Exchange, Prague Stock Exchange, Bucharest Stock Exchange, Market Technicians Association New York, Canadian Society of Technical Analysts, Saxo Bank, Thomson Reuters, TED, Princeton University, University of Chicago etc.

Register: http://www.mta.org/eweb/dynamicpage.aspx?webcode=educational-web-series

 


The Customized Active

Lot of times we get queries regarding, the paper model looks good, but there are few fresh entries. How do I go on model?

We did an extensive white paper on start point bias (Market Scenario Analysis), to explain how starting point bias does not affect the performance of the RMI. The starting point case was built to show our ability to customize a solution and explain to a large institutional client who wanted to run the RMI models for any starting point bias. We understand the RMI paper model and starting point convergence issues, which any investor may have.

How long will the RMI model performance converge or showcase in my portfolio? The convergence may happen in a few months, half a year or more. We have just tested that the starting bias does not affect performance, we have not tested for convergence or how long it may take for a fresh RMI model to converge with a back tested model. The results have shown that over few months or year, the model performances with different starting points are marginal.

The performance differential between real and model may be more for Active 10, considering it has just 10 stocks. If owing to starting customization you have slightly different 10, the performance differential may not be marginal. But that does not change the working of Active 10. The starting point could be just a good feeling of buying fresh. Because at this stage RMI Models do not know how long a winner would persist and when it will revert. Active entries don’t exit till there is a price confirmation. RMI models are designed as benchmarks and need no discretion. We believe all running entries should be bought irrespective, as exits are clearly defined.

Here is the running UK customized active performance from 1 Jan 2014.


RMI ® Active IFTY 5

What if you would have a choice to simulate the NIFTY 50 by 5 stocks instead of 50. That means using just five stock allocations you can not only simulate NIFTY 50, but do it at 9% lower standard deviation compared to NIFY 50, with 70% average invested cash and more than 180 days average holding for each component. Above this your RMI portfolio just fell 30% in 2008 compared to 60% for Nifty, you have an average worst case drawdown near 15% and about 50% chance of outperforming Nifty any running year. All comparisons to NIFTY are owing to lack of a disruptive active benchmark accomplishing an Active 5 selection out of top 50 in India. In real terms an Active model cannot be compared to NIFTY. The IFTY 5 delivered an annualized 10% (2% above NIFTY) from July 2006.

Well this is not stellar performance, as we are used with a much better ACTIVE 10 performance, but considering capital protection, lower volatility, ease of execution and a stock selection mechanism highlighting the top 5 India blue chips, Active IFTY 5 caters well to the Indian investors and global participants looking at the best in India.

We created ACTIVE IFTY 5 (updated for version 3) for the signal driven Indian market, where traders crave for stock picks. It’s still hard to simulate this portfolio on Options owing to low liquidity even in the top 50 stocks. So this remains a spot only portfolio for us. All simulations for 2 times and 3 times are just simulations and should not be attempted using derivatives.


The Dollarama Trend

At 6 billion dollar market capitalization, this is not just any discount store. The store sells Cleaning supplies, Toys, Candy, Grocery, Gifts, Healthcare products, Kitchenware, Stationery, Party Supplies, Hardware. The game is sourcing, pricing and distribution. The competency comes from history, size and real estate. And then consumption does the rest.

This was our last exit from the RMI Toronto 10, up 75% in 475 holding days. BCE was another top exit at 65%.

The RMI Toronto Active 10; holds components for an average 313 days, delivered 40% annualized (over 10 years), at 18% Standard deviation; and outperformed secularly in all the 12 years. This compared to the TSX; at 6.5% annualized, 18% Standard Deviation and secular under-performance against RMI over the last 12 years.


The Priceless MasterCard

MasterCard was a one way road from 2009 lows. On Jan 30 the stock was 135. Now it’s at 800. This means a return of near 500%. This means more than 100% returns annually. Even if for a moment we accept that it was humanly possible for you to pick up MA there in Jan 2009, because you are a contrarian, because you trained under Warren, or maybe you are just too good. Now if we tell you to repeat this by picking not one but three stocks from Jan 2009. We know it’s going to get harder and nearly impossible if we tell you to pick a portfolio of 10. We don’t think any available artificial intelligence could accomplish this feat. This is why if we can capture a part of this ALPHA, it’s a hallmark of a good system. If we can repeat this performance for multiple markets we have the RMI Active approach.

The RMI US Active identified MA more than 900 days back delivering 200% from Jun 2011. The other two near 200% winners were VISA and Gilead. Now one may question what the Credit Card companies knew which everybody who questioned the uncertain pit did not? Well these are all questions that accompany hindsight bias, maybe VISA and MA themselves did not know what rockets will power them. At the end of day what matters are results not cause; affect and explanations.

The RMI US Active 10; holds components for an average 267 days, delivered 28% annualized (over 10 years), at 16% Standard deviation; and out of 12 years outperformed all 12. This compared to the US 100; at 4% annualized, 20% Standard Deviation and secular under-performance against RMI in the last 12 years.