Future Yield

F

At the moment, one of the most highly voted features in Friends of SGXcafe is “Use various known methods to compute the intrinsic value of stocks”.

As much as I want to do it, it is not easy to compute the real “intrinsic” value of stocks, assuming it even exists. All methods that I have found require subjective opinions of the future. And often, small deviations in assumed values influence the results significantly. One popular way to overcome this is to acknowledge that the computed value would be simply an estimation and introduces some form of “margin of safety” to prevent overestimation. Another major issue is determining when will the market ever reflect this “intrinsic” value in the stock price, assuming it will ever. In a year? 5 years? 10 years? Almost all methods do not even discuss this. This issue makes it difficult to validate any suggested method.

Future Yield

After reading up on various methods for computing intrinsic value and learning about their strengths and weakness, I was inspired to develop “Future Yield”. Instead of computing the elusive intrinsic value, I built a model to predict the amount of dividends one is likely to receive in the next 365 days. And if we divide that with the current price, it would give us the Future Yield.

For example, DBS currently has a predicted Future Yield of 3.0%. What this means is that the model predicts that you are likely to receive dividends of about SGD 0.5679 in the next 365 days, and if you buy DBS at its current price of SGD 18.93, your expected future yield would be 3% (0.5679 / 18.93 = 3%).

Future Yield is computed only for stocks that meet certain criteria. This is to improve the accuracy of the model. That is, it only makes a prediction when it is confident. The model is also tuned to be conservative to prevent overestimation. This also helps to improve accuracy by trading in coverage. (Learn more about accuracy and coverage below.)

The advantages of Future Yield are that it is highly accurate (>90% based on backtest results), it does not require explicit assumptions on the future for each stock individually, and it is a number useful either by itself or when compared with another stock.

You can use it by itself to determine if a given stock is giving you sufficient dividend yield at current price to assume the risks associated with the stock. Or you can use it to compare across several stocks with similar risk profile to determine which is more superior.

The weaknesses of Future Yield are that it can only make predictions for ~10% of all active stocks, and that it tends to underestimate the actual dividends received since it is conservative in order to maintain high accuracy. It does not consider the future capital gain or loss of stocks.

Do you have any thoughts on this new Future Yield? Let me know in the comments or by sending me a message!

Happy investing!

Note: While Future Yield is fully available to Friends of SGXcafe, it is only displayed to all users for top 10 most followed stocks (e.g. DBS).

The following sections are for users who are interested in the details.

Backtest Results

One advantage of Future Yield is that it is easy to validate. Simply compare the prediction made on day X and forward to day X + 365 and check if the collected dividends per share is greater than or equal to the predicted value.

Accuracy: By backtesting with 2014 and 2015 data, Future Yield was able to correctly predict ~91% of the time (i.e. Actual collected dividends is greater or equal to predicted dividends). If we are to become even more conservative and only assume half of what is predicted, the accuracy is ~98% (i.e. Actual collected dividends is greater or equal to half of predicted dividends).

Coverage: The predictions was able to correctly predict ~80% of the actual collected dividends (i.e. Total predicted dividends value is ~80% of the total collected dividends value). It is also able to make prediction on ~10% of active stocks (i.e. It makes predictions on >100 active stocks).

Methodology

Selection Criteria: Future Yield only makes prediction on stocks that have given out dividends in the last 3 years consecutively, AND has dividend strength >= 75. This is to select stocks that are consistently giving out dividends and have decent fundamentals.

Formulae: Future Yield = Min(Current Yield x Dividend Strength, Median of last 3 year dividend yield) / Latest Close.

12 Comments

  • Interesting, I found all future yield going down. Does that mean given constant current price, dividend will down? Or all current price too high?

  • All future yield are meant to be lesser than current yield to be conservative (hence higher accuracy). It does not mean that dividend will surely go down or current price too high.

  • Quite useless feature. I am puzzled why all the future yield should go down. if it is to be conservative then we can check the current yield and mark down the future yield our self and decide whether to buy or not.
    For example Micro mechanics (5DD) how did you comeup with the current yield and future yield. I am very sure both values are wrong.

  • “if it is to be conservative then we can check the current yield and mark down the future yield our self and decide whether to buy or not.” => Of course, you can use the current yield and mark it down yourself but how much to mark down is the challenge. The future yield given is a referenced marked down based on a model which attempts to balance between accuracy (90%) and coverage (80%) with a consistent method. It also saves you from spending time to mark it down yourself. Although arguably, it might be more interesting if users could set their own formulae to do so and even better backtest it.

    “For example Micro mechanics (5DD) how did you comeup with the current yield and future yield. I am very sure both values are wrong.” => Why do you think the current yield is wrong? In the last 365 days, there were two dividends payout on 4th November which totaled to SGD 0.04 and with the latest close of SGD 0.955 hence 4.19% for current yield seems reasonable. Definition of current yield in SGXcafe is total dividends of last 365 days divide by latest close.

    Although I assume you meant SGXcafe should include the 5th Feb SGD 0.02 dividends since it is only slightly beyond 365 days or the upcoming 13th Feb SGD0.03 dividends. I agree these few days is a little awkward for Micro mechanics where its current yield is lower than it should be. However, if we change the definition, we might have other issues like including too many dividends in the computation or not updating to the latest dividend collected. As usual, happy to hear suggestions.

  • Ah… One more thing, in cases when companies have those one-off large dividends distribution, future yield would also automatically adjust for that and ignore it whereas current yield would include it.

  • OK, I understood. I am still not sure what kind of formula you have used to come up with the future yield. I am not convinced the data is accurate. Micro mechanics increased their dividend from 2 cents to 3 cents so the future yield should be more instead of lower. 5th of Feb is passed, so why the 2 cents is not added to the current yield for Micro mechanics?
    Frankly saying I use the dividend yield from sgxcafe as reference only. I always cross check with other sources for the correct value.

  • I think it is better to use last year dividend to calculate the current yield instead of 365 days since the Ex Date is not always the same.

  • “I am still not sure what kind of formula you have used to come up with the future yield.” => Okie. Just updated. The formulae is Future Yield = Min(Current Yield x Dividend Strength, Median of last 3 year dividend yield) / Latest Close.

    “5th of Feb is passed, so why the 2 cents is not added to the current yield for Micro mechanics?” => 5th Feb is more than 365 days from today. Current yield only considers the last 365 days.

    “Frankly saying I use the dividend yield from sgxcafe as reference only. I always cross check with other sources for the correct value.” => I understand. It is always good to have data from multiple sources and do let me know if SGXcafe is wrong 🙂

    “I think it is better to use last year dividend to calculate the current yield instead of 365 days since the Ex Date is not always the same.” => This would have problem when it is nearer to the end of the year where last year dividends would not accurately reflect the latest situation.

  • So you could see, We have to check with other sources to confirm the current yield is accurate. Why do we need another data (future yield) on top of the current data which we are not sure correct or not.

    I still use Sgxcafe to filter and buy stocks based on Dividend yield. So accurate dividend yield is desirable.
    We have to work on correcting the current yield as much as possible because this is applicable for all the stocks instead of the top 10% percentage.

  • There are two different issues here.

    1) Dividends not being recorded accurately in SGXcafe. That is missing dividends, extra dividends or incorrect value of dividends. This is really bad if it is incorrect and I do have scripts in place checking what my data provider gives me against various sources attempting to keep them correct and up to date.

    2) Difference in definition. Currently in SGXcafe, current yield uses the last 365 days, it does have its weakness but it also have its advantages. What you suggested of using the last year data is useful at times like now for MICRO-MECHANICS but it also has its own weakness.

    Anyway, lets continue this conversation offline.

Evan Koh

Links

Website:
StocksCafe

Knowledge Base:
StocksCafe Academy

YouTube Tutorials:
StocksCafe Youtube Channel