Turn of Month Returns: Positive but Test Fell Short
- Annualized return (ToM)
- %
- Result
- Holding period (days)
- 5 days
- You entered
- Days in month
- 30 days
- You entered
This article explains why the turn of the month (ToM) effect—historically higher stock returns around the first few and last few trading days of a month—can show positive average returns in a backtest while a statistical test still fails to confirm it. You’ll learn what the ToM effect is, how it’s measured, why a test might “fall short,” and how to interpret such results without overclaiming.
Formula
The turn of the month effect is often expressed as the average daily return during a window that includes the last few trading days of one month and the first few of the next (e.g., trading days –1 to +4 relative to month-end). The formula is:
Average daily ToM return = (Sum of daily returns in window) / (Number of trading days in window)
Where:
- Sum of daily returns – the total percentage change of the index or stock over the ToM days.
- Number of trading days – the count of days in the window (e.g., 5).
- Annualized return – converts the average daily return to a yearly figure using the formula (1 + daily return)^252 – 1, assuming 252 trading days in a year.
The calculator above lets you enter a hypothetical average daily ToM return (in %) and the holding period length to see the annualized impact. It doesn’t test significance—it just illustrates the magnitude.
Worked Example
Using the default values: average daily ToM return = 0.1%, holding period = 5 days.
- Annualized return = (1 + 0.001)^252 – 1 = (1.001)^252 – 1 ≈ 0.2861 – 1 = 0.2861, or 28.61%.
- The calculator rounds to 28.6%.
That’s a huge annualized number, but it’s unrealistic because the 0.1% daily return is only for a few days each month, not every day. The annualization assumes constant daily compounding, which overstates the effect. This example shows why raw averages can look impressive but still fail statistical tests.
Background: What Is the Turn of the Month Effect?
The ToM effect is a calendar anomaly first documented in academic literature in the 1980s. Researchers observed that stock returns tended to be higher on the last trading day of a month and the first few trading days of the next month, compared to the rest of the month. The effect has been reported in U.S. and international markets, though its magnitude and persistence vary by period and methodology.
The cause is debated. Some attribute it to cash flows (e.g., pension contributions, salary deposits), others to window dressing by fund managers, or simply to data mining. Regardless of cause, the effect is a statistical pattern, not a guaranteed profit.
How the Test Works
To test whether the ToM effect is real, researchers typically:
- Define the window – e.g., trading days –1 to +4 around month-end.
- Collect daily returns – for many years of index or stock data.
- Calculate the average ToM return – and compare it to the average return on non-ToM days.
- Run a t-test or regression – to see if the difference is statistically significant, usually at the 5% level.
A test “falls short” when the p-value is above 0.05, meaning the observed positive return could plausibly occur by random chance. Even if the average ToM return is positive, the variance across months is high, so the signal is weak relative to noise.
Why a Positive Average Can Fail the Test
Statistical significance depends on three things: the size of the effect, the variability of returns, and the number of observations. A positive average ToM return might fail a test because:
- High volatility – stock returns are noisy; a few extreme months can offset many modest ones.
- Small sample – even 30 years of data gives only 360 monthly windows, which may not be enough to detect a small effect.
- Outliers – a single crash or boom month can skew the average.
- Multiple testing – if you test many calendar effects, some will appear significant by chance; correcting for that weakens the result.
For example, if the average ToM return is 0.05% per day but the standard deviation is 1.2%, the t-statistic will be small, and the p-value will be high. The effect is economically real in the sample but not statistically robust.
Common Mistakes in Interpreting ToM Tests
Investors and analysts often misinterpret results:
- Confusing economic and statistical significance – a 0.03% daily edge can be profitable after trading costs, but if it’s not statistically significant, you can’t be confident it will persist.
- Ignoring transaction costs – the ToM effect is tiny; trading in and out each month eats the edge.
- Overfitting the window – trying different window lengths until one works is data mining.
- Not adjusting for risk – higher returns during ToM might simply reflect higher beta or volatility.
Always ask: Was the test pre-registered? Was the window chosen before looking at data? Is the p-value corrected for multiple comparisons?
How to Read a ToM Backtest Result
When you see a backtest showing positive ToM returns, check:
- The time period – recent data may differ from historical averages.
- Transaction costs and slippage – are they included?
- The benchmark – is the ToM return compared to a buy-and-hold or to non-ToM days?
- The test statistic – look for t-stat, p-value, or confidence interval.
If the test “fell short,” it means the positive return is not strong enough to rule out luck. That doesn’t mean the effect doesn’t exist—it means the evidence is inconclusive.
Alternatives and Related Anomalies
The ToM effect is one of many calendar anomalies. Others include the January effect (higher returns in January), the day-of-the-week effect (Mondays are often negative), and the Halloween effect (higher returns from November to April). These anomalies often weaken or disappear after being published, possibly because arbitrageurs trade them away. The ToM effect is also related to monthly cash flow patterns, which can be tested directly by looking at fund flows.
Frequently Asked Questions
What does “test fell short” mean exactly?
It means the statistical test (usually a t-test) did not reach the threshold for significance, typically p < 0.05. The observed positive return could be due to random variation.
Can I still trade the ToM effect if it’s not significant?
You can, but you must account for transaction costs and risk. A non-significant effect may not persist out-of-sample, so position sizing and risk management are crucial.
Why do some studies find a significant ToM effect and others don’t?
Differences in time period, market, window definition, and statistical methods lead to different conclusions. Later studies often find weaker effects as markets become more efficient.
Does the ToM effect exist in crypto or other assets?
Some studies have examined it in Bitcoin and other cryptocurrencies, but results are mixed. Crypto trades 24/7, so the definition of “days” is different, and the effect is even less stable.
How many years of data do I need to detect a 0.05% daily ToM effect?
Given typical daily volatility of 1%, you’d need thousands of observations to reach significance—far more than the ~360 months in 30 years. That’s why the test often falls short.
Related
Formula used
in t | Holding period (days) | days | 5 | 1 in d | Days in month | days | 30 | 28 out avg_daily | Average daily return (ToM) | % | 0.1 | 2 out ann_return | Annualized return (ToM) | % | (pow(1 + avg_daily/100, 252) - 1) * 100 | 1
Each line is one input (in, sel, list) or one result (out: label, unit, formula, decimals). How the calculators work
Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.