Friday, 13 March 2009

Sell Low, Buy High?

Last week I decided to sell off around half the stocks remaining in my Comsec and Leveraged Equities margin lending accounts. They were mostly smaller companies in which I had small positions, adding up to around 25% of the value of the stocks in my geared equities portfolio (I haven't sold off any of my managed fund investments yet). The proceeds for the LE sales will sit in the linked Cash Management account until the end of June, when my fixed-rate loan period ends. At that time the total margin loan amount for this account will drop from around $170K to around $70K. The extra cash will reduce the margin utilisation from 98% back down to around 93%, and as the cash value will remain constant, any further market declines shouldn't generate a margin call on this account. The stocks sold off from the Comsec account reduced the net loan amount, as the proceeds generate a negative variable loan amount which partially offsets the fixed loan amount (interest pre-paid until June). I probably should do the paperwork to setup a cash management account linked to the Comsec margin loan account, as otherwise I don't get any interest for the negative variable loan debt.

The stocks that I sold were the ones that would be most affected by a recession in Australia - entertainment stocks, retailers, one steel maker and a construction company - so it may be just as well to off-load them now. Of course, it would have been much, much better to liquidate my Leveraged Equities account back in 2007 when I was seriously thinking about reducing my margin lending accounts and investing via my SMSF instead.

Overall, my 20-year experiment with gearing has been a disaster. As happened last time (in the 90s), the use of gearing has meant that I am unable to buy into the market at the bottom (indeed, I'm again being forced to sell stocks when the market is down). I had thought that this time around my gearing was reasonably conservative (only using 50%-60% LVR, rather than the maximum 80% as the market rose), but the 2008-9 bear market has exceeded my "worst case" scenarios of the usual 20%-40% drop during a bear market. When the market does eventually start to rise again, I won't be maintaining the current 90%+ LVR levels by increasing my borrowings, so even if the market regained it's previous peak my equity wouldn't fully recover.

The scope of my stock losses during the 'Great Recession' shows that my strategy of keeping gearing levels at around 60% LVR during bull markets and then simply holding on and riding out a "typical" 20%-30% bear market was flawed. It didn't have a contingency for a severe recession (well, it did in the form of my Index Put options, but letting them expire in Dec 07 negated that strategy). So I'm working on a revised strategy that will employ dynamic asset allocation to reduce my gearing levels during future bear markets. The trick is to get the timing of such reallocations correct. Many pundits say that 'timing the market' is impossible (assuming random walk/efficient market theory holds), and that examples of successful timing can be attributed to luck.

A book I read a few years ago suggested selling off stocks that had broken their up-trends and dropped by 10%, thereby limiting losses during bear markets, could produce superior returns to simple 'buy and hold' strategy. I'm trying to adapt that theory to help avoid losing too much in any future severe bear markets. My current model is to sell off my geared stock portfolio when the 60-day moving average closing price falls more than 10% below the high attained since the last 'buy' signal. The 'buy' signal (to initially invest, or to reinvest after exiting the market) would arise when the market index daily close had risen 10% above the low reached since the previous 'sell' signal, and stayed there for at least 5 consecutive days. So far back-testing with daily data for the All Ords since 1984 suggests that using these signals to time strategic asset reallocations would have moved you out of the market relatively early in a bear market phase (and without too many false sell signals), and got you back into the market when it had bottomed out. The model did fail with the current 'Great Recession', as it signalled a 'buy' in May 2008 (when the market appeared to have bottomed out), only to signal another 'sell' in June after the market had continued to slide. But mostly using the model to time exits and reentries into the market would have restricted losses during bear markets to around 20%, which would have produced superior returns to 'buy and hold' in about 75% of bear markets. Overall it would have improved my ROI considerably since I started investing in stocks in the early 80s. I could tweak the parameters of this model to eliminate the false 'sell' signal in 2008, but optimising for historic data doesn't mean the model would perform any better in future periods. Keeping models as simple as possible, and retaining some common sense justification for what the parameters are, is usually more likely to result in a reliable model.

If the market doesn't drop too much further I'll retain my remaining current direct stock investments, and aim to slowly pay off my margin loan balances over time. I'll also keep our future superannuation contributions sitting in cash until 'the model' gives a buy signal - currently that would be the All Ords Index closing above 3422 for five days in a row. I may then start to use some leverage within our Self-Managed Superannuation Fund investments - either via buying Comsec CFDs for the market index (if there's a suitable CFD available), or else by investing in a geared stock fund such as the one run by Colonial First State (providing the fees aren't too high). Otherwise I'll just start moving the accumulated cash into the Vanguard High Growth Index Fund.

Subscribe to Enough Wealth. Copyright 2006-2008


Chris said...

with the recent rally, are you getting close to a buy trigger?

mbp said...

I'm continually amazed by your frankness. Well done on admitting to yourself that such a long and expensive project has not turned out well.

The down/up 10% model will do badly in a volatile market where stocks drop and then quickly rebound, such as happened this month. It's possible it will happen again.

Have you thought about taking a Value Averaging approach? said...

As the sell trigger is a 10% drop in the 60-day moving average (not the daily close), the model shouldn't be too sensitive to short term +/-10% movements. Since 1984 the model only generated a handful of sell signals, pretty much matching the significant bear markets.

Unfortunately using the 60-day moving average means that there is considerable 'lag' in the indicator, so you won't be selling anywhere near the peak. Still, the idea is to reduce the risk of massive losses (>>25%), not to completely remove market risk (otherwise I'd be invested 100% in cash).

This month the model generated a 'buy' signal (ie. a bottom has been reached), so I managed to sell off 25% of my stock holding almost exactly at the bottom! D'Oh!

If the market does trend sideways for a while and then recovers in the second half of 2009 I'll probably start buying stocks again (eg. more CDF index fund shares) once my gearing drops well below 80%. I'll be pretty cautious about starting to buy again, as the model gave a false 'bottom' signal in March 2008 before the market started sliding again...