Goldman Sachs' Jose Ursua assesses puzzling seasonal patterns of volatility by month, day, and type of data period
There are a few “market anomalies” affecting the seasonality of stock returns that have captured some investor attention, like the day-of-the-week effect or the January effect, for example. They are called anomalies because – according to financial theory – the market should arbitrage away the regularity of such patterns. But in reality, it does not. We ask whether similar patterns exist with respect to market volatility. And the answer is yes: they exist and are equally puzzling.
Monthly patterns: Fall is the season of vol
We start by looking at average volatility (of S&P 500 daily returns) over the course of the year, split by month. If the arbitraging away argument were right, we would not expect to see major differences across months. Yet we find the opposite – volatility tends to be steady in the spring and summer, considerably higher in the fall, and relatively lower in the winter. Indeed, the averages for those periods since 1928 are as follows: March through August (14.8%), September through November (17.5%), and December through February (13.9%). For the whole year, average volatility stands at 15.2%, so in effect there are substantial intra-year “Vol Seasons,” which do not change much depending on which historical period we take.
[7]
The famous adage, “Sell in May and go away; don't come back till St Leger Day,” is based precisely on the notion that investors would go away for the summer, volumes would come down, and volatility would rise to uncomfortable levels. The St. Leger Stakes (an English horseracing classic) usually takes place in mid-September, so the pattern would be somewhat off when it comes to volatility – at least with respect to US markets. Our results show that a more proper recipe for volatility traders would be to “Buy in Independence Day and go away; don’t come back till Halloween” – or something like that.
Daily patterns: Not so smooth coming back from the weekend
We then look at average annualized volatility by day of the week. We find that volatility is highest at the beginning of the week, and then declines towards Friday. For the three historical periods, Monday through Friday volatility goes from: 19.4% to 17.4% (since 1928), 18.0% to 13.9% (since 1946), and 21.3% to 16.3% (since 1980) One could argue that some historical events that occurred at the beginning of the weak are to blame for these patterns – like Tuesday 9/11 2001, Black Monday of October 1987, and Black Tuesday of October 1929. But excluding them does not materially alter the patterns.
We also find that volatility is highest at the turn of the month, with two intra-month spikes: one around the 11th and another one around the 22nd-23rd. The bottom line appears to be that markets get relatively more nervous at the turn of the week and at the turn of the month. At least partially, these patterns could reflect some Monday blues on the one hand and portfolio-rebalancing sprees on the other.
[8]
Data patterns: Bad news makes markets nervous
Finally, we look at how volatility behaves around data surprises. We find that volatility is substantially higher around data misses (when our MAP indicator – which is higher, equal, or lower than zero when prints exceed, match, or undershoot consensus, respectively – is negative). Moreover, that spike happens almost exclusively in the midst of what we have called “Active” data periods of the month (running from Philly Fed to Non-farm Payrolls, the most intense period in terms of data releases). In contrast, “Lull” periods (when there are fewer data releases) show more boring patterns, regardless of the sign of MAP.
[9]
Keep “Vol Seasons” in mind, but trust fundamentals
For an infinitely-lived investor, exploiting market anomalies can be profitable. But for regular investors with shorter horizons, the strategy is somewhat riskier – since patterns tend to materialize, but do not always. In the end, “Vol Seasons” are something to keep in mind, but fundamental analysis should stay in the driver’s seat.
