Firstly, "always" is a bit of an exaggeration. A recent earnings release from a certain large tech company comes to mind.
Secondly, why can't analysts be that accurate? It's not like there's a lack of data. Moreover, I ain't an expert, but I imagine earnings exhibit a lot of autocorrelation, which should make things easier.
Ask Jack Welch. If memory serves me he did pretty well laying out analysts "expectations".
I looked it up:
“During the heart of the Jack Welch era,” writes Martin, “GE met or beat analysts’ forecasts in forty-six of forty-eight quarters between December 31, 1989, and September 30, 2001—a 96 percent hit rate. Even more impressively, in forty-one of those forty-six quarters, GE hit the analyst forecast to the exact penny—89 percent perfection. And in the remaining seven imperfect quarters, the tolerance was startlingly narrow: four times GE beat the projection by 2 cents, once it beat it by 1 cent, once it missed by 1 cent, and once by 2 cents. Looking at these twelve years of unnatural precision, Jensen asks rhetorically: ‘What is the chance that could happen if earnings were not being “managed’?”’ Martin replies: infinitesimal.
How is this possible. Analyst says Q3 EPS will be 48 cents. Company reports 49 cents.
How can these analysts be so accurate?
That would be like playing roulette and always getting the number or one off every time.
Statistically something seems odd.
The companies tell the analysts what it's going to be minus a small % so that the company will look like they beat estimates. The analysts, knowing that the companies want to beat estimates, publish estimates some small % higher than what they're told. Then the company does their accounting such that they at least meet those estimates. It's a self-fulfilling prophecy.