Arborbridge wrote:ULVR is pretty good for yield right now, but anyone who didn't buy in March 2020 missed the best recent high. ...
Something odd about your chart or your comment there, because the chart shows Unilever's yield as being higher in January 2021 than March 2020 (both dates approximate, due to eyeballing them off the chart).
I think it's probably something odd about the chart, as
https://dividenddata.co.uk/dividend-yield.py?epic=ULVR (click the "5y" button) shows the March 2020 peak as being higher than the January 2021 peak. And on clicking the "10y" button, it also shows the March 2020 peak as being higher than the peaks in 2013 and 2014, again unlike your chart.
A
possibility about what the oddity is: the March 2020 share price low (which of course will coincide with dividend yield peaks over short periods in which no new dividends happen) was quite a sharp downwards spike: the lowest intraday price was 3583.5p on 16 March 2020, the lowest closing price was 3726p, also on 16 March 2020, and the share price only went below 3840p at all on the four days 12, 13, 16 and 17 March 2020 and only closed below it on the first three of those days. So if the price data your chart is based on happens to miss those days, the yield spike could be cut off in your chart. For example, if you use weekly closing price data that happens to be taken on Wednesdays (quite a good choice because it minimises the number of weeks that need special treatment because of bank holidays), the actual closing share price low in March 2020 would be 3726p but the March 2020 low in your data would be 3977.5p on the 25th. That would make the actual yield peak a factor of 3977.5p/3726p = 1.0675 times higher than would be visible in the chart - which would make a difference between a peak of about 3.7% as shown in your chart and an actual peak yield of about 3.95%.
Or if the weekly closing price data were taken on the last trading day of the week (which would normally be Fridays, is probably the more natural choice, and is what the FT does), the March 2020 low in your data would be 3854.5p on the 13th, the factor would be 3854.5p/3726p = 1.0345, and the actual peak yield would be about 3.83%.
Not saying that's the actual explanation of the discrepancy between the peak heights in your chart and in dividenddata's - you're in a far better position to determine that than I am, as you know the actual data your chart is based on and I can only look at a few possibilities. But it does illustrate one of the difficulties of producing charts that completely accurately reflect reality: markets can produce some very short-term spikes that are liable to be missed simply because the spike happens entirely between the data points you use! And such spikes are particularly likely to occur around share price peaks and lows when markets are turbulent... (This can be dealt with by judicious use of Open/High/Low/Close share price data when preparing the chart, but that's liable to be quite a bit more work if you've only gathered closing prices so far...)
I should also say that the reality of share purchases for most HYPers is of course that they pick times to make them that fit in with other things they're doing, rather than watching the market price every day (or even at all times during market hours if one were to try to catch intraday peaks and lows!), and of course you never see that a price is a significant peak or low until after it's passed! So the peaks on a completely accurate yield chart are only the yields one might have happened to catch if one had happened to be very lucky -
not yields one might reasonably expect to catch!
Finally, another three possible explanations of or contributors to the difference between your yield chart and dividenddata's: firstly, exchange rates. Dividenddata's chart shows a March 2020 peak pretty close to a quarter of the way up from 3.75% to 5.0%, i.e. to 4.0625%. That's far too many significant figures for a figure 'eyeballed' off a chart, of course, but it is pretty clear that the peak shown is over 4%, and that's a bit higher than the plausible actual yield peaks indicated above because of share price spikes between data points. But if you look at dividenddata's yield calculation (at the bottom of its chart page), it basically assumes that the shareholder receives and accumulates their dividends in euros over the year
without converting them to sterling, and then translates them to sterling at the prevailing exchange rate at the end of the year. This is of course not what most UK-based HYPers are likely to do: they will probably receive their dividends in sterling, already converted from euros by Unilever.
Secondly, dates: examining dividenddata's calculation further, they're including a dividend in it if it was announced in the last year (e.g. the dividend announced yesterday is already in it. I suspect a more natural thing for a HYPer to do is to include it if it was paid in the last year - or one might also reasonably include it if it went ex-dividend in the last year.
Thirdly, 'dividend lag': if a HYPer bases their yield calculation on their actual shareholdings and dividends received, and they bought some of their shares in the preceding year, they may not have received a full year's worth of dividends from all their shares. This will depress the average yield of their holding for that year if it's calculated as (dividends received from holding during the preceding year)/(value of holding at the end of the year).
Gengulphus