Misinterpreting Trading-Volume Statistics

It seems that every summer there are articles published about how price movements in the stock markets are less than meaningful since many investors, both retail and commercial, are on vacation. The contention is that the resulting low volume has two significant effects:

1. Trends in equities and other securities and commodities are untrustworthy since market action is not fully representative of investor sentiment.
2. Lower volume means higher volatility in individual issues.

Supposedly, so the story goes, that post-Labour Day these effects may be reversed as the totality of the investor community is back on the job. I believe this belief is completely wrong. I will demonstrate this by qualitatively countering the arguments of the purveyors of this wrong-headed idea with some elementary statistics.

Let's start with the first point by quoting from this Reuters article that exemplifies the argument:

Reuters data shows volumes for the Dow Jones industrial average for the full day on Tuesday were only 56 percent of the index's 90-day average daily volume. They were higher for the Nasdaq, but still only 73 percent of the norm.

In Europe, the FTSEurofirst 300 volume was 69 percent of its daily average on Tuesday, while on Wednesday it looked likely to be even thinner -- it was at less than 40 percent with a little more than two hours to go.

As for China, where the Shanghai Composite Index has fallen more than 18 percent this month, volumes have mostly been less than half what they were in July, especially in the past eight trading sessions.

It is harder to gauge volumes for other assets -- foreign exchange trading data, for example, tends to come with a lag, from central banks -- but anecdotally, participants say it has also been thin.
...

Given this, investors might be better off waiting until September to see what is really happening to financial assets.

These percentages can be rigourously treated as a sample over the total population of investors. As anyone will know from various political polls, as the sample size grows in proportion to the population the expected error in the results, the variance and standard deviation decline. In this manner we typically see variances (typically stated with a 5% confidence, or 19 times out of 20) of several percent when the sample size is several thousand out of a voting population of over 20 million citizens, or on the order of 0.02% of the population. Variances are higher when the polling results are broken down further by province, sex or other factors since the sample size is a smaller percentage of the population (or a similar percentage of a smaller population).

Yet for the case of the stock market we have effective population samples of 40% to 75% of peak season trading volumes. The variance in these instances is well under 1%; that is, summertime equities quotes should accurately reflect the thinking of the entire investor community, including those at the beach. The only way around this is to hypothesize that bulls or bears, but not both, are more (or less) active in the markets during the summer. I have seen no evidence to indicate that this is true, and I would be suspicious of any evidence that would be presented since it seems so unlikely.

There is one psychological factor that may strike the market in September that could give a false indication. People who return from vacation tend to be mentally refreshed and therefore somewhat more optimistic. I wonder whether it is possible that these individuals are more likely to press the 'buy' button when they first get back to work. Even if this conjecture is true (and it could actually be the opposite if the news flow is persistently negative over the summer) their belief structure should quickly revert to the norm once they find that their new-found optimism isn't confirmed by their trading returns.

Regarding the second point about volatility, there is validity to the idea. Illiquid issues do show higher volatility (increased short or medium-term variations from longer duration time-lines such as 200-day moving averages), and that volatility does show a modest inverse correlation to trading volume. However, we are talking about volatility and not longer-term trends. This means that if the price of a stock makes a sharp swing up or down, there will be investors that will quickly take advantage of any "mis-pricing". We then see a reversion to the mean. It may not happen the same day, but it will almost always occur.

Canadian investors should be well-versed on volatility since the majority of stocks listed on the TSX have low average daily volumes. This is not surprising since there is a comparable number of public stocks in Canada as compared to the US, yet our economy is less than one-tenth the size. Our investors have to spread themselves thin if they are to avoid piling onto the relatively small number of liquid stocks with a large market cap. The point here is that low-volumes, while resulting in higher volatility, are irrelevant to long holding times. Traders, however, do care since their holding times are typically measured in hours or even minutes.

Expect to see more of the same sort of story in mid-December when the Christmas-New Year's doldrums hit trading volumes. The only difference is the shorter duration: two weeks rather than two months.