Search This Blog

Monday, August 31, 2015

How Did Those Robo Advisors Do Last Week...We Will Likely Never Now

Robo advisors are to a large extent a black box. While the client can know his allocation he really has no idea how the transactions are executed in the markets. The past week trading was very erratic with many trading halts and large discrepancies between the trading prices of ETFs and the "intrinsic value" (value of the underlying securities) and wide swings in individual stock prices many transactions executed through any automated basis had a high likelihood at being executed at unattractive prices that didn’t represent "fair value" and were soon reversed.

August 24 was the most notable example with multiple trading halts and wide swings of ETF values. Any human being (like me) could look at a screen and know something was very wrong creating a trading environment that was best avoided. For example several of the dividend ETFs were trading at prices far lower than the drop in the S+P 500 Etf even though there was a large overlap in the holdings of the two. The dividend etf price simply couldn't represent fair value. And many stocks traded down close to 10% at the open and quickly recovered a significant part of those losses.

Investors should have learned long ago not to place stop loss" orders in the market as they are likely to be executed during "flash crash" type events when markets take unusual short term moves. A more sophisticated investor would also know that with the exception of a few of the largest ETFs it is bad trading practice or place market as opposed to limit orders. And in market conditions that prevailed most of the day on August 24 and at other times during the past week, the volatility and numerous trading halts for stocks made it best practice to simply refrain from trading.

Many of the algorithmic traders and high frequency traders that provide liquidity in normal markets simply did the same. Turning off their computers or placing very wide bid ask spreads. Market makers in ETFs unable to execute trades in many of the underlying stocks did the same. All of this further complicated market conditions and left those with market or stop loss orders literally paying the consequences.

It will be impossible to know how the robo advisors performed in these markets. At some point a researcher far more capable than me will measure the actual performance of the ETF portfolios of the "robos" vs the performance of the ETFs in their allocations based on market closing values. That would give some measure of any "value added" or "slippage" because of the transactions of these robo advisors.

The most interesting/complex test case of the impact of robo advisors execution is the case of Wealthfront.

Unlike other robo advisors which make use of ETFs. Wealthfront uses what it calls a" third generation tax loss harvesting strategy" which makes daily transactions which according to their description optimizes tax savings.


Wealthfront makes use of what it calls a unique tax optimizing strategy. There is a well-known strategy of tax loss harvesting of replacing one etf with a short term loss with another essentially identical ETF (for example two total US stock market ETFs) to realize a tax deductible loss without changing strategy. This is a widely used tax management tactic used by many individual investors and "non robo: advisors. 

In order to avoid "slippage" a loss in performance due to the trades in tax harvesting it is best to do this with tow essentially identical ETFs in quiet markets .An example would be a sale of a position in the Vanguard total stock market  Etf when there was an imbedded loss and when the ETF was down .5% for the day immediately executing the buy part of the tax harvesting with a purchase of the ishares total stock market etf when it is also down .5%. That would create the tax loss with no impact on portfolio performance. Any slippage between the two trades would affect portfolio performance. For instance in a volatile market the sale might be done with the one ETF down .5% and the purchase made when the other ETF in the trade was down only .2% the net "slippage" of .3% on the trade was the difference between the sell price and the buy price.

Wealthfront claims it is unique in two ways. Not only do they execute tax harvesting transactions on a daily basis they also do it with their own portfolios of individual stocks rather than with listed ETFs, generating individual stock trades. They call this "direct indexing"

 As Wealthfront explains it:
Instead of using a single ETF or Index Fund to invest in U.S. stocks, Wealth front’s Tax-Optimized Direct Indexing directly purchases up to 1,001 individual securities on your behalf — up to 1,000 stocks from the S&;P 500® and S&;P 1500® Indices and an ETF of much smaller companies.
This allows us to take advantage of the countless opportunities for tax-loss harvesting presented by the movement of individual stocks, to further improve your investment performance. Combined with our Daily Tax-Loss Harvesting service, we believe this could add as much as 2.03% to your annual investment performance.
Wealthfront uses a similar strategy in which it replicates the total stock market index with a smaller number of individual stocks.

Wealthfront does these transactions on a daily basis. The assumption is that this generates greater tax savings than for instance doing the tax loss harvesting periodically based on the status of the individual account simply swapping between two ETFs.

I am sure this "optimized" strategy works in back testing on a computer, although Wealthfront claim of a single number for client tax savings is questionable given the different tax status of individual clients. Wealthfront notes http://www.quora.com/What-are-the-main-differences-between-Wealthfront-and-Betterment-3


But two other major questions remain:

There can be two drags on performance:

Does Wealthfront's replication strategy work? Since Wealthfront doesn’t actually buy the ETF does its replication match that of the corresponding ETF?

While the tax harvesting may create tax losses what is the impact on portfolio performance?

The large number of transactions increases the possibilities of slippage. The buys and sells in the accounts may not be executed at zero cost. The transactions may not be executed such that the net result of the buys and sells does not create an outcome lower than the movement in the S+P 500. So the tax savings may potentially be higher...but the large number of tax harvesting strategies may actually be a negative for portfolio performance.

 Wealthfront uses this strategy on accounts with balances as low as $10, 0000 meaning daily tax harvesting numbering possibly several hundred trades a year of tiny size.  Is it necessarily true that the net result of 100 individual stock trades generating $1000 a year in tax losses in a $10,000 account (a high number since it would be 10% of the account value) in tax harvesting losses produces a better outcome than one trade between ETFs generating that same $1000 in tax losses And of course the more trades particularly since they entail purchases and sales of individual stocks. The more opportunity for slippage. In fact with that number of trades the cumulative impact of the bid ask spread could add up significantly. And as Wealthfront grows. In assets its tax harvesting trades will have greater impact on the market, making it even more difficult to avoid slippage

Did the strategy that worked in computer research work in highly volatile markets? The history of markets is full or automated strategies that worked on research based on years of data. Until the markets hit a highly volatile period. When these strategies run into illiquid markets. Those outcomes predicted by computer modelling disappear. A notable example was "portfolio insurance" in 1987. As noted many algorithmic traders and high frequency traders have their own "circuit breakers" when markets get too volatile and/or valuations across securities seem to be far away from fair value they turn off the computers and watch the market chaos. Humans both professional and some nonprofessionals have learned the same thing: to sit back and wait under such market conditions.

But Wealthfront was active in the markets on August 24.

Wealthfront CEO   proudly told Bloomberg that Wealthfront's tax loss harvesting did over $200 million of trades on August 24...the most volatile trading day in years.


And this anecdotal article reports on a client notes

The account had been easy enough to set up—skim a few questions, assess a few options, and voilĂ —all within a few minutes from the comfort of our couch. But, on Monday, as the Dow tumbled more than 1000 points, he watched as the service rapidly rearranged his account on the fly. “I’ve been constantly refreshing,” he texted, “and it’s a roller coaster.”
He was startled, concerned, and a little bit confused. But the service was doing exactly what it was supposed to do. Betterment, Wealthfront, and FutureAdvisor say their services not only take the headache out of investing, but offer real opportunities when the market dips. The question is how well this will work—and how well the services can retain their clients—in the long-term, especially when the market takes a turn for the worst, or, one day, heads into a recession.
Wealthfront may have generated tax losses and as seen in the above quote rapidly adjusted portfolios (with account size as small as $10,000).but was what the impact on the performance. Wealthfront’s tax strategy entails buying and selling individual stocks. For the strategy to work there should be minimal "slippage" the down movement on the stock should be equal to the decline on the stock purchased. If not the net gain or loss will cause the portfolio return to differ significantly from the index performance.

Call me skeptical that Wealthfront was able to execute its tax loss harvesting strategy which involves buying and selling individual securities without significant slippage and thus drags on performance. On August 24 over 1200 stocks and ETFs had trading halts which are put in place when there are swings of 5% or more in the stock. Numerous stocks had volatile swings particularly at the open of trading. How did Wealthfront's optimized tax harvesting manage in these markets executing $200 million worth of trades ?

It certainly wouldn’t have been easy given the market environment. To give just two examples of the market environment on August 24 that would have been faced by Wealthfront's direct indexing consider he performance of two widely traded SP 500 stocks. Citigroup opened trading on August 24 at 48 traded 6% higher an hour later the spread between high and low for GE was even larger trading at the open at 19.87 before recovering to 23.87.

As Wealthfront actively traded on August 24 it isn’t hard to imagine that their trading got them caught taking "tax losses" at extreme lows and buying back other stocks that didn’t have similar losses. Buying a stock down 10% and replacing it with one that had fallen only 8% would result in a loss vs the index far in excess of any tax savings.

August 27 was another roller coaster as the Dow opened 200 points higher dropped 300 points mid-day and recovered 300 points at the close. It is reasonable to assume that Wealthfront’s tax harvesting trading was active throughout and also found it difficult to avoid slippage.

Analyzing Wealthfront’s tax management trading and its impact on performance vs the index it is trying to replicate would be extremely difficult if not impossible.. Most likely Wealthfront will keep that information proprietary and it will be difficult to ever discover.

But it is clear that tax harvesting strategy through replication is not as simple as it sounds. And in fact may produce little if any benefit compared to a far more simple strategy of periodically swapping between ETFs when there are "harvestable" tax losses that are significant. And to wait until markets are orderly to execute the trades. 


No comments: