Are Past Dividends a Predictor of Future Dividends?

“Past dividends aren’t an indicator of future dividends”.

Spend enough time on FI related social media and it won’t be long before someone says this. But is this true?

How much do past dividends correlate with future dividends? And what causes someone who hasn’t won previous dividends to suddenly become a dividend earning machine?

Past Dividends vs Future Dividends

To work out how much past dividends relate to future dividends is quite tricky, for a few reasons. My first thought was to compare total dividends earnt in the 2018-19 season with dividends in this current season (many thanks to @FIDataStephen and @FootyIndexLDN on twitter, as well as the lovely folks here at IndexGain for helping me with some of this data).

But given the dividend increase earlier this season this makes it difficult to do as players have earnt more this season than last season.

So next I thought I would compare rankings, to see if how high a player finished one season correlated with the next. But this is also difficult to do, as with media opening up to more to squad players, this saw the roughly 200 dividend winning players of 2018-2019 double to 400 players this season. So, therefore, finishing 100th last year is not the same as finishing 100th this year. You wouldn’t be comparing apples with apples.

The next best thing was to take a large mainly random sample of players and work out where they rank relative to each other for total dividends earnt each year and normalise these scores to 100 to compare to data sets on the same scale (the nearer to 0 the more divs they have earnt, with 100 being 0p dividends earnt). This isn’t ideal but offers a good indication. I also didn’t include IPD’s as this would be a nightmare to calculate over a season for each.

Doing this and plotting it on the graph produced a correlation of 0.72 and looked like this:

So What Does This All Mean?

Well let’s start with the most important finding. 0.72 is a pretty high positive correlation. 1.0 would be a perfect correlation.

This means high dividends won last season are associated with high dividends won so far this season. And likewise, lower dividends won last year are associated with low dividends the next. To give a correlation of 0.72 some comparison, it’s more than the correlation between a student’s IQ and how well they do at school. So a significant one.

So why is the correlation not a perfect 1.0? There are probably three reasons (of which a player can benefit from in combination or just on its own):

  1. A change to how FI do things – this includes changes to the PB matrix, opening media and position changes. People get very hung up on matrix changes (often citing Trent Alexander-Arnold as one, and he has improved this year (going from being one of the top 20% div earners in terms of ranking to being within the top 5 – this is probably due to the next reason below. The likes of De Bruyne and Kroos are better examples here.
  2. Player development/decline – Players tend to improve in the formative years as a professional footballer and decline as they get nearer retirement. This is reflected within their ability to win dividends, and as such means last years performance wouldn’t perfectly map on to this one. The likes of Sancho, Alexander-Arnold, Rashford fit here.
  3. Change in player circumstance – most notably, moving from a non-PB league to a PB league has the biggest impact. This partly explains three of the biggest climbers from last year on this graph (Ighalo, Haaland, Bruno Fernandes).

Could The Correlation Be Random?

If you compare enough sets of data, you will eventually find some random relationships. In research terms, this is known as p-hacking and generally provides some poor findings. The most famous example of this is how the number of films that Nicholas Cage has appeared in actually correlates positively with and number of deaths by drowning (0.66). Clearly, one doesn’t actually relate to the other, but if you go looking for enough stuff you will find something.

That is unlikely to be the case here. As the two variables (last year dividends and this season’s dividends) are measuring the same sort of thing (how well a player does on both the PB matrix and the media algorithm).

And of course, correlation doesn’t mean causation. Ice cream sales are positively correlated with the temperature, but clearly buying more ice-cream does not make the temperature rise. So this does not show that last years dividends cause next years dividends, but it is reasonable to say looking at this data set that one would expect last years dividends be a decent predictor of next year’s yield.

What Extra Interesting Stuff Stands Out from This Data

I think there are four interesting areas on this graph:

 

  1. You will see how the data by point 1 has a few extra circles on it. This is because a lot of players are all on this data point. Almost 50% of players who won 0p in dividends went on to win 0p in dividends next year. About 75% of players who won 0p in dividends last year won 5p or less this year. So if you buy a player who hasn’t won dividends last year, don’t be expecting a sudden change next year as the odds are against you.
  2. If a player is in the top 30% of dividend winners one season, there is a good chance they will be in the top 30% next year. These are mainly your elite players and/or ones very well suited to FI.
  3. There is a smaller pocket of players who were ok on Football Index last year (ranked either side of average) who made the step up to being in the top 10%. This means there is good capital appreciation and dividends to be had in this group.
  4. These are the players who won 0p in dividends last season (and as such score 100 on the x axis) who are in the top 10% this season (with a score of 0-10 on the y axis). There are less than five players on the whole index who would be in this category. If you manage to pick these at the start of the season a) lots of people would probably have laughed at you at the time, b) you would have experienced one of the best capital appreciation opportunities on the platform and c) you were either very very lucky and/or skilled.

In Conclusion

Past dividends are definitely a good indicator of future dividends. Don’t believe anyone who says otherwise. But there are always a few exceptions to the rule. The big profit obviously comes from spotting the ones who haven’t won many previously and go on to do so. But this is very difficult to do. As such, it is a riskier trading strategy. But get it right and the rewards are huge.
Share This Post
Sigmund Freund
Written by Sigmund Freund
Psychologist by day, FI Trader by night. Mainly tweet about my current strategy, thinking biases, decision making + how to assess value
Have your say!
0 0
IndexGain
Join our 8000+ Football Index trader community on Slack for free. Click here to sign up!
Holler Box