Thursday, May 31, 2018

The Undoing Project

I recently read "The Undoing Project" a 2017 book by Michael Lewis.  It primarily tells the story of two Israelis, Daniel Kahneman and Amos Tversky, who worked together for many years studying how people make decisions especially bad decisions.  Kahneman received the 2002 Nobel economics prize for this work (shared with Vernon Smith).  Tversky would probably have shared in the prize as well if he had not died in 1996 (at the relatively young age of 59) making him ineligible.

I was disappointed in this book.  It is quite long (352 pages plus notes) and unlike most of Lewis's work I didn't find it to be particularly entertaining.  It contains a lot of biographical material about Kahneman and Tversky which (while intermittently interesting) isn't especially relevant to their professional work.  It is repetitive in places (for example colleagues and students describing how brilliant they were).  It abruptly introduces other characters and then drops them without really integrating them into the narrative.  It contains chapter notes at the end of the book but no index.  It suggests that their work was very important but doesn't really explain why.

Nor did I think it was especially instructive.  It isn't technical enough to be a good introduction to Kahneman's and Tversky's professional work.  I have previously reviewed books by Ariely, "Predictably Irrational" and Thaler, "The Winner's Curse" which discuss related work on decision making.  While I didn't recommend them either they would provide a better introduction to the field.

In short I would skip this book.  It isn't a good introduction to this subject area and I didn't find it compelling as entertainment.

Sunday, May 13, 2018

Chasing Hillary

I recently read "Chasing Hillary" by Amy Chozick.  Amy Chozick covered Hillary Clinton's 2008 campaign for the Wall Street Journal and her 2016 campaign for the New York Times.  In this 2018 book she gives a long (375 page) first person account of her experiences.

Chozick mentions several times Hillary's inability to give an inspiring explanation of why she was running for President.  One could similarly ask what was the point of this book.  It is pretty useless as history as it assumes you already know the big picture of what happened.  A college student assigned to read this book 20 years from now would not learn much about American politics between 2006 and 2016. 

The book contains a lot of autobiographical material.  In many ways it is more about Chozick than Hillary.  But it does not really succeed as autobiography either.  Chozick doesn't seem that important or interesting.  And she presents herself as such a caricature of the insanely ambitious (and insecure) career woman who only cares about getting her byline on the front page of the paper that you begin to wonder if she is writing self-parody.

And while the book is amusing in places it is far too long to succeed purely as entertainment.

The book describes a toxic relationship between Hillary and her press corps which is largely Hillary's fault as she (and her campaign aides) make no effort to hide their utter contempt for the reporters covering her.  While somewhat understandable this was a bad mistake on Hillary's part as just a little judicious flattery would probably have paid off in better coverage. 

Chozick criticizes the way the New York Times covered the material hacked from the Clinton campaign and released but doesn't explain what she thinks they should have done instead.  Hillary's speeches to Goldman Sachs were a campaign issue, once the transcripts were obtained (albeit illegally) and published on the internet I don't see how the New York Times could avoid discussing them (especially since authenticity doesn't seem to have been an issue).  Of course if Hillary wasn't  willing to release the transcripts herself she probably shouldn't have given the speeches.  It used to be fairly easy to tell different things to different audiences but it is a lot riskier now.

In general I think the Russian meddling wasn't a big deal.  It was well known during the campaign that Putin preferred Trump to Hillary.  Prior to this election being known as the candidate preferred by the Russians wouldn't have been considered an advantage (and in fact there is little reason to believe it helped Trump more than it hurt him). 

So in summary unless you are a real politics junkie you can safely skip this book, it doesn't contain much of lasting value and isn't that entertaining.

Saturday, February 24, 2018

Pension Guaranteed

The fourth post I made to this blog (which I started in 2009 after getting laid off from IBM) was about the Pension Benefit Guarantee Corporation (PBGC) which is an US government agency which insures private pensions up to certain limits.  These limits were of interest to me as I was planning to start taking an early retirement pension from IBM later that year.  For people who have started collecting a pension before their plan fails the insurance limit is determined by the year in which their plan fails and their age at that time.  This means the limit increases from calendar year to calendar year (due to an annual inflation adjustment which is sometimes 0) and on birthdays (due to a formula which guarantees larger amounts as your age increases from 45 to 75). 

The upshot was that when I started taking my pension later in 2009 as planned less than half of it was guaranteed but as time went by the guaranteed amount increased in a somewhat irregular way until on my birthday in 2017 it exceeded the amount of my pension (which is fixed).  My pension plan was (and remains) in pretty good shape so I wasn't all that worried that it wasn't completely guaranteed.  Still you never know what the future will bring so it is nice that it is now 100% covered.

The PBGC maximum monthly guarantee tables can be found here.  Note there are some additional limitations (none of which apply to me) that can reduce these amounts.

Wednesday, February 21, 2018

XIV Implodes

XIV was the ticker symbol for an exchange traded note (ETN) designed to move inversely to the VIX index (a measure of volatility in the S&P 500 index) on a daily basis.  So if the VIX index were to go up 10% XIV would ideally go down 10%.  Or if the VIX index were to go down 10% XIV would ideally go up 10%.  This is a simplification, an exact description with numerous warning and caveats can be found in the lengthy prospectus.   One of the warnings (see page PS-16) was:

... The long term expected value of your ETNs is zero. If you hold your ETNs as a long term investment, it is likely that you will lose all or a substantial portion of your investment. 

Nevertheless with historically low volatility holding XIV would have worked well in 2016 and 2017.  At the end of 2015 XIV closed at 25.8, at the end of 2016 XIV closed at 46.75 and at the end 2017 XIV closed at 134.44.  So for 2016 XIV returned 81.2% and for 2017 XIV did even better returning 187.6%.  This didn't go unnoticed and despite the above warning XIV began to attract long term holders.  Some them thought they had found the road to riches and invested most or all of their available funds. 

Then in February 2018 volatility returned to the market.  On Friday February 2 XIV opened at 126.5.  After a bad Friday for the S&P 500 XIV opened Monday February 5 at 109.57.  After a worse Monday for the S&P 500 XIV opened Tuesday February 6 at 10.49 losing all the gains for 2016 and 2017 and more in one day.  These losses caused the ETN sponsor to terminate the fund eliminating even the theoretical possibility of regaining the lost ground over time.

This of course came as an unexpected and costly shock to those investors who had invested heavily in XIV.  It seems to me that they had basically made 2 serious and avoidable mistakes.

First the fact that XIV had done well in 2016 and 2017 was no guarantee that it would continue to do well.  There are thousands of securities trading in the US markets and an infinite number of strategies for buying and selling them.  So at any given time there are bound to be many securities and strategies that have done well recently out of pure luck.  Also the more popular a strategy becomes the harder it is for it to achieve extraordinary returns as its popularity will move prices against it.  If for example someone noticed that stock prices tended to be low at 10 AM and high at 11 AM and lots of people tried to buy at 10 AM and sell at 11 AM this would drive up prices at 10 AM and reduce them at 11 AM until any excess profits were eliminated.  For this reason no widely known strategy should be expected to reliably produce outsize returns.  I should note here that this argument doesn't apply to index funds because they are trying for average returns not outsize returns.

Second if you have a strategy that has positive expected return but will occasionally suffer substantial losses it is unwise to invest all your money in it as it is very difficult to recover from losing most (or worse all) of your money.  If you lose 50% of your money it takes a 100% gain to get back to even but if you lose 90% of your money 3 consecutive 100% gains still won't get you back to even. 

Sunday, February 18, 2018

2017 Portfolio Review

After outperforming the market in 2016 my brokerage account returned to normal in 2017 and underperformed the market.  The market as represented by the Vanguard S&P 500 ETF, VOO, was up 21.60% (19.47% capital gain, 2.13% income).  My brokerage account was up 19.17% (16.60% capital gain, 2.57% income).  So I lagged by about 2.43%.

Since I didn't make any transactions during the year it is relatively easy to determine the source of my underperformance.  At the start of the year my account was 30.44% invested in individual stocks, 46.46% invested in VOO, 14.72% invested in other Vanguard ETFs and 8.38% invested in cash.  My individual stocks actually outperformed returning 26.32% (23.19% capital gains, 3.12% income).   VOO of course matched the market.  However my other ETFs lagged badly returning 6.99% (3.31% capital gains, 3.68% income) as did cash returning about 1.01% all income.  So weighting by position size my individual stocks contributed 1.44% of outperformance, my ETFs contributed 2.15% of underperformance and my cash position contributed 1.72% of underperformance.  Which sums to 2.43% of underperformance.

My individual stock outperformers (beating the market by at least 10%) were CAT, AET, ALL, SOUHY, NSC and BBL.  My market performers (within 10% of the market return) were INTC, JPM, CM, PEG, BNS, ED, WFC and WBK.  My underperformers (lagging the market by at least 10%) were XOM, IBM and TGT.  Among my ETFs VYM and VPU lagged the market but were within 10%.  VNQ and VDE underperformed.   

Monday, January 1, 2018

Quicken Workaround Again

I use the 2010 Quicken Premier program to keep track of my personal finances including some stocks, ETFs and mutual funds that I own.  In 2016 the update function which downloaded current prices from the internet stopped working.  However you can still input prices from a csv (comma separated values) file in the appropriate format and as explained in this post I figured out a way to produce such a file (without entering each price manually) from the Google finance portfolio feature.  Unfortunately in November 2017 Google revamped their finance pages and eliminated the portfolio feature.  It is unclear to me why they did this since this seems like a useful feature that one would not think would be difficult to support.  The new pages (which seem to be designed to be viewed on a phone) seem much less useful.  Anyway I needed a new workaround.

I tried Yahoo again but I still can't create an account.  Apparently this is because Yahoo requires new accounts to be associated with a cell phone (which I don't have).  Annoyingly Yahoo doesn't clearly explain this instead providing a registration procedure for people without cell phones that doesn't work.

Fortunately I was able to use Morningstar.  This requires a basic account which is free but you do have to register.  Morningstar's portfolio feature allows you to enter a list of securities.  It demands purchase dates and share numbers for each but you don't have to enter real values.  Once you have your portfolio set up you can periodically update the security values with current prices.  You can then download them into an Excel spreadsheet.  Then you can use the free Open Office version of Excel to export the spreadsheet as a csv file (ignoring warnings about format incompatibilities).  Finally I wrote another little Fortran program to extract the prices from the csv file and put them in the format Quicken wants.  The main difficulty here was that the security names (like "International Business Machines") were stored instead of the symbols ("IBM") which is what Quicken wants so I also have the program read a list of symbols which I created (this list will have to be updated if I buy or sell securities).  A bit complicated but still better than trying to enter many prices manually.

Wednesday, December 6, 2017


The Deep Mind division of Google just released a paper in which they describe how they applied the methods they used to develop a Go program that achieved super human strength to develop programs to play chess and shogi.  The best previously existing chess and shogi programs are based on alpha/beta search and have been incrementally improved over many years.  They have been stronger than the best human chess players for about 20 years (Deep Blue beat Gary Kasparov in a 6 game match in 1997) and recently surpassed human shogi players as well.  However the Google programs (named AlphaZero in each case) appear to be stronger still, defeating Stockfish, a strong chess program, and Elmo, a strong shogi program by wide margins in 100 game matches.  10 example games (all wins for AlphaZero) in the match against Stockfish were included in an appendix to the paper and can be played over here.  They are pretty convincing (with the caveat that there doesn't seem to be much opening variation).  AlphaZero wins several games with long term positional sacrifices where it is not initially apparent that it has sufficient compensation for the material loss.

This is kind of a big deal.  People had tried applying Monte Carlo search with neural net evaluation functions to chess before but were unable to match the performance of highly tuned alpha/beta search programs developed over many years with lots of domain specific knowledge hardwired in.   Using a general algorithm with domain specific knowledge limited to the rules of the game to quickly develop apparently superior programs is impressive and a bit scary.