When 'buy on dips' works

To see this, we of course need a test. Mine is to define a dip as occurring when the All-Share index falls more than 5 per cent below a peak reached in the previous six months: I'm considering only end-week data, and when there's a sequence of weeks in which prices are 5 per cent or more below their six-month average, I consider only the first such week. Since January 1988, there have been 53 dips on this definition. I then ask: what did the All-share index do in the following six months?

The answer is uninspiring. On average, the market rose by just over 1 per cent on those 53 occasions. This is less than the average rise of 2.5 per cent in all six-month periods since 1988. On those 53 occasions, the index fell 45 per cent of the time, which is slightly more than the average for all six-month periods.

Yes, buying on dips has worked sometimes. The dips in October 2005 and June 2009 led to rises of over 20 per cent in the following six months. But on the other hand, dips in June 2002 and May 2008 led to heavy losses.

This shouldn't surprise us. Sometimes a dip is just a dip. At other times, though, it's the start of a serious bear market. Buying on dips failed during the tech burst of 2000-02 and during the financial crisis of 2007-08.

This poses the question. Are there times when buying on dips works better than at other times, and can we identify these without hindsight?

Yes - to some extent. One useful guide here is whether prices are above or below their ten month average at the time of the dip (or their 44-week average since we are using weekly data). A dip that leaves prices above their 10-month average has often been a buying opportunity. The 23 occasions when this happened led to average rises of 5.3 per cent in the following six months. Even this, though, is not infallible. Prices fell on seven of these occasions.

Where buying on dips fails on average is when it causes us to buy when prices are below their 10-month average. On the 30 occasions when this has happened, prices subsequently fell more often than not, and by an average of 2.2 per cent across all 30 occasions.

What matters, then, is momentum. Buying on dips works when upward momentum is intact, but fails when there's downwards momentum.

It's touch-and-go whether this will be true of the next dip. If prices were to fall 5 per cent this week they'd stay just above their 44-week average, but a bigger or later drop would push them below it.

We can put this another way. Buying on dips works when value investing is dominant - when a price fall brings out the bargain-hunters. But it fails when momentum investors dominate - when investors regard others' selling as a reason to be bearish themselves.

This brings us to another criterion for when buying on dips works - uncertainty. When this is high, value investors will be less active, because they will have less confidence in the reliability of their valuation methods - it's dangerous to buy a share that's 10 per cent undervalued if market volatility is high it could soon become 20 per cent undervalued. And momentum will be stronger because if we don't know what we're doing, we're more likely to take our cues from others.

It's no accident therefore that the biggest failures of buying on dips occurred when volatility was high - in 2000-02 and in 2007-08.

You might think there's another criterion for when buying on dips works - monetary policy. Buying on dips should work better when the concerns about the economy that often trigger a dip also cause central banks to cut interest rates. At near-zero rates, however, this cushion for share prices is absent, so we'd expect buying on dips to fail.

It's a good theory. But there's not much evidence for it. There have been 15 dips since March 2009 (when Bank rate was cut to 0.5 per cent), and on average buying on them led to gains of 4.4 per cent. While respectable, this is less than the average of all six-month gains since then, of 7.6 per cent. Yes, the last attempt at buying on dips (in July 2015) would have failed my test. But it worked well in November 2012 and October 2014. It's not clear, then, that buying on dips fails at the zero bound.

Overall, then, there might be a case for buying on dips, but only as long as prices are above their 10-month average.