gougou wrote: ↑Fri Feb 18, 2022 12:32 am
Here is a pretty recent example:
ASX:GRR is a small Australian iron ore miner with current market cap of A$943M. It was trading very close to its net working capital from 2017 to 2020 while also making a lot of money selling iron ores. Its current P/E is about 4x but it was mostly 1x to 2x in 2020. It was very undervalued and probably worth a lot more than its market cap if liquidated.
ASX:GRR closed at A$0.61 on Dec 8, 2021. ASX:GRR announced a special dividend of A$0.1 (16% of its market cap) on Dec 9, 2021. Share jumped on Dec 9 and closed at A$0.745. On the ex-dividend date Dec 14 share closed at A$0.665. So this 16% special dividend was basically free money plus 9% appreciation in share price from declaration to ex-dividend.
If GRR didn’t declare this special dividend, in all likelihood its share isn’t going to move much from Dec 8 to Dec 14. It obviously got a significant boost from the special dividend so this dividend definitely mattered.
The axiom under the usual 'frictionless' assumptions is that the stock price will drop by the dividend at the moment it starts to trade ex-dividend. It is not that the stock would drop by the dividend from the close on the day before the ex-dividend to the close on the day it goes ex-dividend (that contains a whole day's noise), and definitely not that the stock would drop by the dividend from the day prior to the *declaration* date till the moment after ex-. The axiom applies to a *known* dividend. A company declaring a higher or lower dividend than the market expected is a new piece of information about management's view of future profitability. It's no surprise if a company with very low valuation that *declares* a huge dividend generates a price pop: the market will likely interpret it as company's situation from management POV (people who want to keep their jobs, not make their company to disappear by paying out an unaffordable dividend) as better than the market thought. The problem with using that information practically is if you didn't know on Dec 8 what was going to be announced Dec 9. If I knew what was going to happen tomorrow, there'd be all kinds of ways for me to make megatons of money.
Back to the actual axiom, the stock price must drop by the *known, already declared* dividend at the moment it goes ex- or it would be arbitrageable, again under the somewhat idealized but typical assumptions in financial theory of continuous trading, ignoring transactions costs and taxes. If divs of companies are $.10 and stock prices consistently drop less than $0.10 at the moment of ex-, the free money would be attained by buying stocks just before the moment of ex- and selling them just after. If the price consistently drops more than div, short a micro-second before ex-, buy back a microsecond after (net profit despite having to make up the dividend to the person you borrowed the stock from). That is the correct concept of 'dividends don't matter'. It does not apply to unknown future dividends. Nobody on the thread who knows what they are talking about at all has claimed it does. There's a valid debate to be had if dividend *policy* affects risk adjusted return on a forward looking basis, but I'd reiterate a strong preference for original academic work showing a practical way to separate sheep from goats and only invest in part of the market according to dividends, not typically junky DIY Portfolio Visualizer analysis.
More prosaically the axiom that an already known current dividend results in equal price drop at the moment of ex- would not necessarily strictly hold once introducing taxes. Assuming price drop equal to dividend and 'buy the dividend' strategy of buy/sell pico-seconds before and after the moment of ex-, you get offsetting dividend income and short term capital loss. To the extent those two things have different tax treatment, for the investor at the market clearing price, the equilibrium price drop doesn't have to be exactly the dividend. Transactions costs are nowadays very small and ignoring them is not typically important anymore, though a problem with past data is that transactions costs were much higher than now if you go back far enough. And, the fact you can't typically actually trade micro-seconds before and after ex- should just introduce statistical noise, not a bias in either direction. Tax would be the main thing which might affect the axiom in real life.