The Run-Up Before Ex-Dividend Date (2024)

Investment thesis

The price of a dividend-paying stock tends to drift up before the ex-dividend date. In this article, I present 4 possible ways to take advantage of this price anomaly, with different holding periods based on different levels of belief in market efficiency.

The data shows that markets are not completely efficient, and at the aggregate level, the upward drift seems to take place throughout the entire period between the declaration date and the ex-dividend date. Thus, the method that performs the best with highest aggregate returns and highest risk-adjusted returns is to buy immediately after dividend declaration and sell right before the ex-dividend date.

Introduction

In my previous article on dividend stripping, I mentioned a few other price anomalies surrounding dividends. One of them was that the price of a dividend-paying stock tends to drift up before the ex-dividend date. I hypothesized that the upward drift anomaly was due to other people trading other variants of dividend stripping strategies, which was to buy before the ex-dividend date. This article will be an investigation of this price anomaly, whether it is significant across a range of stocks, and which variants of trading methods might be used to take advantage of it.

The period before ex-dividend date

How do we know that a dividend is coming? Well, we find that that a dividend is coming when the dividend is declared. This declaration (or announcement) of a dividend usually takes place at the same time when a company reports its earnings (could be annual earnings or quarterly earnings). This means that there is a limitation in our analysis where we cannot separate the effects of earnings and dividends, but I will talk more about this limitation in the conclusion.

The dividend declaration provides us with the details of the dividend, such as the ex-dividend date (and the record date), the pay-date, and the dividend amount. Since this article is about the time period before the ex-dividend date, the relevant time period for us is between the declaration date and the ex-dividend date. Before the dividend is declared, we do not know exactly when the dividend is coming, and after the ex-dividend date, the price is influenced by other factors (such as the ex-dividend date price drop, and price recovery which is another price anomaly).

This time frame varies widely among stocks. The average MSFT dividend has approximately 2 months between declaration date and ex-dividend date. On the other hand, the average PG dividend has only 1 week between declaration date and ex-dividend date. A special case can be seen in WMT where all the quarterly dividends for the year are all announced at the same time.

(Created by author using data from Seeking Alpha)

For yet other stocks, the time frame between declaration date and ex-dividend date varies widely among dividends of the same stock. One such example is CAT where some dividends become ex-dividend a week after being declared, and others become ex-dividend up to a month after being declared:

(Created by author using data from Seeking Alpha)

Another issue with dividend declarations is that they sometimes take place after the market has closed. Thus, for our analysis, we shall assume that we only get to trade the stock the day after the declaration date.

The Four Noble Truths

(In my last article, I referred to the trading method as “the Tao” referencing a concept from Taoism. Here I have 4 trading methods, so I referred to them as the four noble truths, referencing the Four Noble Truths of Buddhism. Get it?)

Through my research, I have found 4 plausible ways of attempting to profit from this upward price drift before the ex-dividend date, each implying different levels of belief in market efficiency:

1. On the day after the declaration date, buy at the open and sell at the close. This method believes that whatever the information conveyed in the dividend announcement, investors would finish their actions on the first day.
2. Buy at the close of the day after the declaration date, hold for one week, then sell at the close of the trading day. This method believes it takes slightly longer for investors to react to the informational content of the dividend announcement.
3. One week before the ex-date, buy at the close of the day, sell at the close of the day before the stock becomes ex-dividend. This method believes that investors will wait until near the ex-date before buying, just so they can collect the cash dividend.
4. On the day after the declaration date, buy at the close, hold for the whole period, sell at the close of the day before the stock becomes ex-dividend. This method believes that the entire time period from declaration date to ex-date has an upward drift due to the beliefs of the above 3 individual methods.

The below timeline summarizes the four methods:

(Created by author)

Do note however, that it is possible for “Ex-Dividend Date – 1 week” to be before “Declaration Date + 1 week”, if declaration date and ex-dividend date were close enough. Also, colors are chosen arbitrarily, to make it easier for reader to compare the different trading methods throughout this article.

Let’s use an example to illustrate. For the stock KO, the most recent dividend was declared on 25 Apr 2019, with an ex-dividend date of 13 Jun 2019. (Chart is stretched to provide more space for my writings.)

(Created by author using chart from Investing.com, data from Yahoo! Finance)

Here’s another example using JNJ, where all 4 methods would have lost money. The most recent JNJ dividend was declared on 15 Jul 2019, with an ex-dividend date of 26 Aug 2019.

(Created by author using chart from Investing.com, data from Yahoo! Finance)

If you have read my previous article on dividend stripping, you would already know what I’m going to ask. If a long-only strategy loses money, could it be due to a drop in the overall market? And if it is due to a drop in the overall market, can we hedge against it? Let’s find out by comparing the SPY movement during the same period as the JNJ analysis. (SPY on the left, JNJ on the right.)

(Created by author using chart from Investing.com, data from Yahoo! Finance)

Aggregate result of stocks in the Dow Jones Industrial Average

Before I present aggregated results, there are two points to note. One is that I used data for dividend declaration dates and ex-dividend dates from approximately 2013 onward. Prior to 2013, dividend details seem incomplete, with many of the dividends missing the declaration date data. This is true with dividend data from both Seeking Alpha and NASDAQ. (I have no idea how complete the dividend data from Bloomberg is, I will check it out some time.)

The second point to note is that I removed Walmart (WMT) from the analysis. This is due to the earlier-mentioned fact that WMT announces 4 dividends at once, and this renders most of the analysis useless. For example, if I were to buy WMT after the dividends were announced, do I buy 4 times of the trading size? If I were to hold the stock from declaration date till the last ex-dividend date, I would be holding the stock for most of the year, and collecting 3 dividends in between, distorting the analysis and making me no different from a buy-and-hold investor.

Below are the aggregate results for 29 stocks in the Dow Jones Industrial Average, with the results being a sum of individual trade returns, from approximately 2013 to August 2019.

(created by author using data from Yahoo! Finance, NASDAQ, and Seeking Alpha)

It is again important to note that the above results DO NOT represent annualized return, CAGR, or even total return. The magnitude of the numbers is irrelevant, and only whether they are positive/negative indicates the persistence of a price anomaly. The sum/average for the 4 individual trading methods can also be compared to see how much of an edge each method has.

All the aggregate sums/averages are positive, regardless of hedging. This indicates that there exists a price anomaly, and it is overall consistent with the idea of price rising between declaration date and ex-dividend date. Hedging also expectedly reduces both the sum/average, as well as the standard deviation of returns, like the last exercise of dividend stripping.

It is interesting to note that the returns seem highest for method 4 (holding from declaration date to ex-dividend date), returns seem lowest for method 1 (1 day holding period after declaration date), and returns seem in the middle for methods 2 and 3 (1 week holding period). This could imply that market efficiency is low, and it takes time for other investors to react to the informational content of a dividend announcement. The average and standard deviation of individual strategies can be compared by calculating some sort of “risk-adjusted return ratio” (though not exactly the Sharpe ratio we know), with that ratio showing the same pattern as returns (highest for method 4, lowest for method 1, in-the-middle for methods 2 and 3).

Conclusion

The price anomaly of an upward drift between declaration date and ex-dividend date exists at the aggregate level. This article provided 4 different ways of capitalizing on this price anomaly, but the best (best is subjective here) way is to buy the stock a day after the declaration date and hold it until the day before the ex-dividend date. This method captures the upward drift of the whole period instead of just a day’s or a week’s worth of the drift.

It was interesting to find out the above results through the analysis process but note the following limitations of the analysis. You never know exactly when a company will declare dividends (or announce earnings), the best you can do is to monitor the companies in your stock list for any announcements and refresh your sources daily. Some stocks like PG and MMM only declare dividends one week before it goes ex-dividend, the short time frame limits the number of permutations of trading methods that can be applied. Also mentioned in the article, the number of historical dividends with known declaration dates is quite small, most dividends before 2013 have incomplete dividend details, this small sample size could be one of the limitations of the analysis.

Also mentioned earlier, dividend declarations usually occur at the same time as earnings announcements, so the price movements could be due to earnings result instead of dividend effect. However, it is impossible to separate the 2 effects in my study. It is possible to study the effect of post-earnings-announcement drift by focusing on earnings announcements that do not come with dividend declarations. However, earnings announcement dates are not as easily available as dividend declaration dates, with this problem being more severe for non-dividend-paying stocks.

Lee KCM Research