Vanguard: A framework for the active-passive decision

Some light bedtime reading:

https://static.vgcontent.info/crp/i...king-the-implicit-explicit-research-paper.pdf

 

Good read. I think this paragraph from the conclusion sums it up nicely:

 

Good to know. I haven’t had a chance to properly read it yet as doing some painting today. More paint on me as usual than the area being painted. But a quick skim through the paper suggested it would be useful..

Yes patience is definately important with good Mgrs. Many investors do badly with active Mgrs due to performance chasing. We’ve held PMC for a long time through their good and bad times and I believe our patience has been rewarded overall. If you believe in the LIC Mgr then when they go through a rough patch it usually presents a great NTA discount buying opportunity. I still remember grabbing all the PMC I could afford when they missed a dividend some years ago and traded down around $1. I’ll take that anytime thank you.

Good to see a balanced view for a change rather than the all or nothing Index vs active view. I think both active and indexing have a role to play.

 

The all-or-nothing view appears to be widespread (definitely not just limited to PC chatter) and they emphasise as such:

 

That was a fun time - not. I do recall at one stage buying when it was under a $1 and wondering why TF am I doing this?

 

. Perhaps increase the Gin intake next time it happens.

I was a keen buyer I must admit. But younger then though and still had one salary coming in.

During some stages of the GFC my nerves got tested somewhat. But as a long term investor what can do but take a deep breath and keep buying. If I gave in to fear I think I’d end up doing all sorts of stupid things. Wasn’t bloody easy at the time though.

 

This is IMO one of the great things about DRPs as it takes away the decision making process/ doubts at the worst possible time.

 

Agree with the sentiment that all or nothing attitude is short sighted.

The problem with the studies based on 'median' performance is that they throw out all the IP of the FP industry. In the real world not everything is Vanilla.

Good quality, well selected active managers DO produce alpha over extended timeframes. Tends to come back to 80/20 rule, you just need to know how to identify the 20 or find someone who can. The worst approach to this is to use past performance and brand names, portfolio construction runs much deeper than this and relies on researching the track records of the actual managers (as in the person, not brand name) then finding what we call the 'efficient frontier' that optimises low beta/high alpha.

Another mistake is to focus too much on absolute return. Yes it's important however the greatest reason of all to use active managers is their ability to manage risk and reduce portfolio downside in times of bear markets when index based portfolios have no choice but to ride the bear as he jumps out the window. Warren Buffett's no 1 rule of investing: Don't lose money

 

I think their is a case for active. However if you do go for an active strategy you need to remain convicted to that strategy.

If you combine different active strategies for the same medium term period the investment styles work against each other and you gravitate to the index.

It takes guts.

 

Agree Ross, important also to acknowledge that an 'active' approach encompasses a heck of a lot of variance in portfolios and how they are managed. Could be anything from a 'hands off no brain required' wrap account with active managers as the chosen allocation in any flavour or blend of MF, Multi manager, SMA, LIC, ETF, direct stocks..... through to a specific self managed diversified blue chip stock picking portfolio by a skilled individual capable of hours and hours and hours of research. Personally I rule out LIC's because I don't enjoy paying $1.30 for my $1 notes when I can get them for 70 cents elsewhere.

For anyone who really wants to retire and have free time the answer is inevitably the easy route that does not require thinking on their part.

 

Those investing in LIC's seem happy to get that $1.30 for $1 when the time is right though Alex

Getting it for 70c is impressive

I agree, anyone who really wants to retire and have free time the answer is inevitably the easy route that does not require thinking on their part. What is the easy route though?

Not having a crack, this area is your forte' so genuinely interested in the 'how to'

 

It also needs to work, after fees and taxes, then consider Manager (business) risk which is significant.
Who wants to be having to chop and change products (CGT) while sitting on 7 figure capital gains..... Pass!!

Heaps of strategies look great in hindsight and in a test tube. Cost matters, tax matters, diversification across markets, currency, geography all important. Index all the way for listed stuff I reckon...there are 132k CFA charterholders in circulation and more each year. The rocks have all been turned over.

As you raise, there is also the issue of the mindset that goes with expecting to outperform and not.....will someone sit for a decade on an underperforming value strategy for example with the hope that it will come good? Unlikely me thinks.

Of course unlisted / direct investment / vc a different game and where I think those that have an inclination and interest should look for more interesting opportunities than the public markets.

 

@Redwing Thanks I appreciate the question and your respectful tone. Also understand your POV re LIC's, you post a lot of good stuff here for sure. Don't get me wrong I don't think they are terrible but personally I just prefer not to have to deal with 2 different potential 'true value' displacements especially for a regular dollar cost averaging plan. Plus added liquidity risk is a negative IMO. I am also personally of the view that a lot of the value displacement relating to the premium often paid for 'confidence in the managers' has more to do with those products being flavour of the month right now. Ok go ahead and lynch me now

Prefer SMA's as a cleaner and more direct arrangement, just as much 'confidence in the managers' but don't ever pay a premium for it. They have recently become a lot more accessible to smaller players. Happy if we all make a decent cracker in the end sorry to drift lets leave the LIC debate for those threads

To put this 'portfolio specific' question in context, I advocate multiple slight edges.

Paradoxically, although you see me post a lot on market info and TA I regard personal psychology, cash flow discipline and debt structure as even more important and fundamental to success than the specific investment vehicle itself. For the masses, it's a safe road being invested in well managed, common sense diversified growth asset holdings and following the basic rules of accumulation under a decent tax structure. When using an active approach or passive active core satellite it is a good approach to develop the portfolio in stages adding incremental alpha at each stage with the goal reducing or making a minimum sacrifice of beta.

As you know, "how to extract maximum alpha for minimum beta using other peoples brains instead of your own" is a deep question and would require a very detailed (and personalised) answer to do it justice. I am also not in the habit of just giving away the keys to my kingdom, as many here know I am involved in the software development side of these achieving these things we speak of and that's a heck of a lot of IP you asked me for

I tell you one thing for sure there is no 'one size fits all' approach and it is very difficult to generalise in relation to 'how to'. And that is one of the points of taking the 'active' route, or even a core-satellite portfolio that combines the best of passive with the best of active characteristics.

What I would give as an answer to a beginner investor with $5k would be completely different to what I give to a successful sophisticated investor with $5M in investable net assets.

I will aim to post more info on the principles of 'how to' next week when time allows. It's complex. As much as I would love give the good folks here all the specifics, fund names, portfolio construction etc that would completely devalue the advice given to my clients many of whom are members and guests here. Again anything I post here is only for education and not advice.

Happy to take this to a private discussion of course.

 

I agree.

Many people talk about index funds not enjoying market outperformance. The greatest benefit of index funds is that they do not enjoy market underperformance.

Most alpha chasing funds are fluffed up nonsense. I get two factor theory and everything - I just don’t want to be a guinea pig and live through two decades of underperformance to ultimately be proven right.

The more you chase alpha the more likely you are to get negative alpha.

Keep it simple and keep it cheap. Alpha does exist but it is random. I think I will never see it - or it will be a momentary fluke.

Unlisted businesses and unlisted assets with direct ownership operating in imperfect markets is my choice. And index the rest.

 

@Redwing again thanks for the question that has come up about portfolio construction theory. Just a warning this will be a very long post however what I am about to run through will barely scratch the surface of this topic. If you read on and get annoyed at how long the answer is please don't blame me, @Redwing asked for this. Bear with me and I will aim to give some decent value

What I will attempt to do is cover off some of the basic research elements, traps to be aware of, then get in to the heavy stuff finishing at correlation coefficient for now in this post. The when time permits I will come back and do a second post to see how far we can get with portfolio theory using the capital asset pricing model (CAPM).

I'm sure there is much you will already know so again bear with me. I will try to keep things clear (used to be industry lecturer in FP subjects including this one so will pulling out a few old textbook examples to help), but please understand this will go in to some heavy maths later on.

Research Elements

The first thing to be mindful of is are we combining a portfolio of individual stocks, other listed vehicles or a portfolio of managed vehicles or a combination of all elements because it changes the way we go about the exercise.

In all cases an accurate source of historical market data is essential.

For stocks, market data from a good source will serve us fine. For managed vehicles of all descriptions, it's not so simple and the type of data and method of data basing is also critical. Understanding how to piece it together to achieve our objective is another whole topic.

Here is why I say that... in many cases it a is USELESS exercise to evaluate and rely on the historical performance of a funds brand name as in 'Perpetual XYZ Fund' or 'Colonial JKL Fund'. Why, because the people who's skill is responsible for the funds performance (ie: the managers themselves) are moving around behind the scenes from fund to fund.

Example: 'Perpetual XYZ Fund' under manager A performs at 15% p/a after fee rolling average return over 10 years when the index averages 11%. This manager then moves funds and manager B steps in with a track record over those same 10 years of 8.5% p/a (working for a more conservative fund). It is yet to be established whether manager B will replicate a similar track record and outperformance as manager A for the next 10 years and the track record for that fund is effectively invalidated as a basis for future expectations. I know many of you guys are aware of this but just wanted to mention the main trap in using fund data straight form the research house and relying on 'brand names' unless they have 100% stable management team over the period you are examining and no plans for changes.

So how to get around that? Industry multi-managers (one example of analysts that do this kind of portfolio construction) go out and interview each individual manager or team they are interested in and examine and piece together the data for that managers personal record and that becomes the historical performance data set needed to do long term data deep dive needed. As well as that industry players are first to know in advance of the rest of the market when, for example, high skills manager A is leaving 'Perpertual XYZ Fund' and where they are going next. They are able to sell down and exit if need be before the general market gets wind of this news and avoid the stampede exit that sometimes follows.

While on the subject there are many other incremental benefits to the way multi manager portfolios are constructed and managed including all the assets being help under a single trust structure even if they are using multiple fund managers. This means each time a manager leaves or needs to be replaced the assets all remain in the single trust while a new manager is selected to take over that same segment of the trust - minimising costs, CGT, time out of market, paperwork and hassles of this change for investors. In contrast if you had not used a multi manager and done what most people do - divided up and invested separate amounts of capital in each of the 'brand name' managed funds individually instead of using a single lump in a multi manager portfolio, to make the same change of manager you would need to sell down that entire amount, choose new fund, then buy back in again triggering a host of transaction costs and CGT issues. Little off topic but worth a mention because only looking at absolute return can be a fooling metric at times, it's also about the structure under which the assets are held and how efficient it is without causing costs and headaches.

Portfolio Construction Theory

Now we venture in to an introduction to some of the maths used, it runs much deeper of course especially in the modern algo based approaches

Now predicting probabilities of future returns form historical data is no easy task given all the variants that impact returns such as systematic risks - market risk, interest rate risk, purchasing power risk as well an unsystematic risks such as business risk and financial risk.

Clearly it is difficult for an analyst to estimate future earnings of a company (stock - value is based on assets, IP and business model of the company) or future earnings of a manager (individual or team - skilled people we are seeking to employ to make good decisions based on facts).

However using our data set that we now know is fully relevant to the task (see above) we can work on generating a standard deviation model using a normal distribution that gives us a very high probability as a range of returns. Let me give a simple numbers example to demonstrate.

Example: In the past fund manager A (or substitute stock ABC) has achieved the following results...

• 5% p/a in 10 of the past 20 years
• 10% p/a in 5 of the past 20 years
• 20% p/a in 2 of the past 20 years
• -8% p/a in 3 of the past 20 years

So in other words probability of each return look like this...

Probability Return
0.50 5%
0.25 10%
0.1 20%
0.15 -8%

Naturally the sum of probabilities is 1.

Using an expected return formula we can calculate expected risk and return.

E(R) = E WiRi

where E is summation (sigma), Wi is probability weighting and Ri is expected return for that probability weighting

Plugging in our numbers...

E(R) = (0.5 x 5) + (0.25 x 10) + (0.1 x 20) + (0.15 x (-8))
E(R) = 2 + 2.5 + 2.5 - 1.2
E(R) = 5.8

E(R) = 5.8% represents our mean of the expected return however we need to account for uncertainty (risk!) and this is modelled using probability distributions either side of the mean using a standard deviation of the expected return formula.

SD = Sqr Rt { E Wi (Ri - E(R))^2 }

(Sorry bit hard to type these formula properly without the right symbols). Percentages need to be expressed as fraction of 1 for this.


SD = Sqr Rt { 0.5(0.05 - 0.058)^2 + 0.25(0.1 - 0.058)^2 + 0.1(0.2 - 0.058)^2 +0.15(-0.08 - 0.058)^2 }
SD = 0.0731 (or 7.31%)

Now we are in a position to make some firm probability statements about likelihood of future returns by making use of the normal distribution curve which as you know resembles a 'bell' sitting on a table.

1. The majority of returns are concentrated around the mean.
2. As you move further away from the mean the probability of the return becomes less likely.
3. Half the returns are above the mean and half are below.
4. 95% of returns are within 2 standard deviations of the mean (5.8 - (2 x 7.31) to 5.8 + (2 x7.31))

Therefore the majority of expected returns would be expected to be 5.8% and 95% of expected returns are expected to be within -8.82% to 20.42%

Combining Securities into a Portfolio

The maths really gets fun now so I hope you are ready

Assume we have data for 3 securities (managers or listed equities) we are looking at combining in to a portfolio and we have calculated the following expected returns and standard deviations...

Wi Security x Security y Security z
0.15 25% 40% 40%
0.2 15% 20% 20%
0.3 10% 0% 10%
0.2 5% -5% 0%
0.15 -25% -15% -25%


E(Rx) = 7%
E(Ry) = 6.75%
E(Rz) = 9.25%

SDx = 14.78%
SDy = 17.7%
SDz = 18.93%

Now to understand the effects of combining these into a portfolio next we need to allow for correlation or relative movement between 2 securities. Reduction in risk relies on choosing stocks with negative correlation.

The way to do this is use a covariance formula. For example the covariance between securities x and y is represented by....

COVxy = E Wi (Rix - E(Rx))(Riy - E(Ry))

where Rix are the individual returns for the probability weightings on security x and similarly for Riy

Without wanting to bore you any further lets cut to the chase and here are the results...

COVxy = 214
COVxz = 269
COVyz = 313.8

Now we need to convert these to something more useful by giving it a range of values from -1 through to +1. To achieve this we convert it to a correlation co-efficient using this formula...

CCxy = COVxy / (SDx x SDy)

These turn out to be...

CCxy = 0.82
CCxz = 0.96
CCyz = 0.94

Using these results we can calculate the expected future return for a portfolio made up of various weightings of each security. To keep it ultra simple we will stick with only 2 securities being blended...

Example: Portfolio made up of 50% of security x and 50% of security y

Expected return of portfolio

E(Rp) = 0.5 x 7 + 0.5 x 6.75
= 6.875%

Standard deviation of portfolio

SDp = Sqr Rt { Wx^2 x SDx^2 + Wy^2 x SDy^2 + 2 x Wx x Wy x SDx x SDy x CCxy }
= 15.48%

We now know the majority of future returns will be expected to be around 6.875% and 95% of future returns are expected to be in the range from -24.085% up to 37.835%

Anyone still awake? Hope you are following this there's a test later

Might leave it there for now, more to follow when time permits.

 

Good on yay @Redwing you owe me 2 Bex

 

Great post thanks @Alex Straker. Had I read the second half of your last post earlier I would have been reaching for the home brew (multiple times). I expect @Redwing to pay for rehab if required.

This post triggered my memory in regard to an article by Chris Cuffe who’s obviously no stranger to the world of funds management. It’s in regard to his charity focused “fund of funds” Third Link Fund. I found this comment of interest:

My 10 biggest investment management lessons - Cuffelinks





Alex, unlike me, as a professional what’s your take on Cuffe’s approach which appears to be far removed from a formal scientific process?

Secondly it would be great to get your opinion on LIC fund of funds FGX/G. Annual charity donation of 1%, no performance fee. Most service providers also providing services for free. Some excellent Mgrs on board but does the blend of styles end up in Index like performance? If it achieved this with lower volatily however there may still be value in owning these products?

FGX (asx):
http://www.asx.com.au/asxpdf/20171114/pdf/43p77lc19fl4zn.pdf

FGG (International):
http://www.asx.com.au/asxpdf/20171114/pdf/43p77rv0qlfq66.pdf

Thanks

 

Forgot to add this Research Report to previous post:

https://cuffelinks.com.au/wp-content/uploads/Zenith-Third-Link-Growth-Fund-Final-30-June-2016.pdf