Hasn't this been know for, well, ever? I mean as people retire it's been historically common to start moving into dividend paying stocks to provide income for them. I think most people understand that moving to dividend paying stocks means giving up some upside on the stock movement in return for the income and perceived stability of these stocks.
And dividend paying stocks tend to be big blue chip firms traditionally so the flight to them when markets are rough is again, almost a reflexive move that's been happening forever.
> This leads to predictable marketwide price increases on days of large aggregate dividend payouts, including stocks not paying dividends.
Well there goes that trading signal:(
That's the biggest takeaway for me - and something I'd backtest separately to confirm.
In a finance class I took many many years ago, we were taught about the Modigliani Miller "Dividend Irrelevance Theory"...which basically says that you, as an investor, shouldn't care whether you get a dividend or not. But if we've learnt anything from the past year or so, in low interest rate environments, yield is everything. Money now is better than money later.
I think more important than yield in low interest rate environments is that dividends are generally pretty reliable, as opposed to price appreciation, which is not. A long history of increasing dividends shows a strong underlying business model and history of execution. Furthermore, for mature companies, there is only so much money you can productively employ; retaining money that you cannot use effectively reduces shareholder returns. Conversely, you shouldn't be paying dividends if you are entering an Amazon winner-take-all market because you need it to grow. So you really should care about dividends.
Though maybe the authors know what they're doing and were under pressure to publish (publish or perish in effect?) so they just picked one of the more BS low-hanging fruit theorems in economics to poke holes in for the millionth time.
I don't think its totally irrelevant. It may not be applicable in a today's real world, as you point out. But the point of a model is to try to explain what happens when you change the input parameters.