The problem is that Harvest, like Tribute and other 'turn cards over on a deck, get stuff' cards, is quite unpredictable - in a deck with good variation, you should almost always get at least $2, usually $3, and sometimes 44.
It's a lot more predictable than tribute : you know it will be money. And from my experience, it's usually $4, sometimes $3.
This depends heavily on the board:
1. No trashing. Okay, you start with 7 coppers. Say you have N cards. When you hit a copper, how likely are you to hit a second copper? Well assuming I did the permutations correctly, you need an N of about 26 cards to have better than 50% odds assuming no other duplicates. Reality is worse because you will be getting duplicates of other cards. Gaining 16 cards is looking at around T9 on a quick board to actually start paying out at $4.
2. Low draw control. If you can discard cards, life is fine, but Harvest has the trouble that if your intend to draw your deck and if it ends up being in the bottom 4 cards itself, it gets progressively worse. Say I want to hit double province and I expect to hit $3.5 per Harvest. This means I need to hit either 4 and some treasure or 5 ... but if my deck is only 16 cards (say I opened Remake/Ratcatcher) then I will completely whiff on one Harvest each turn. Getting enough Harvests to ensure big VP buys becomes somewhat self-defeating as they draw down your expectation value for each Harvest play. A thinner deck where you can more reliably draw several Harvests is also a deck where you are much more likely to get a complete $0 whiff on one or more Harvests.
3. +action requirements. At best, Harvest costs 2 actions, 1 buy and 2 draw to yield a province. Gold costs 0 actions, 1 buy, and 3 draw. There are a lot of boards where you will buy villages, and not want to burn all that on a Harvest. With only one village pile, but with terminal draw, light trashing, and +buy it is pretty steep to skip out on the draw. After all Smithy x2/Silver x4 is 2 actions, 1 buy, and 0 net draw, yet you need a good reason to go for it instead ditching some silver and using a gold.
On the positive side:
1. It is generally easier to gain $5 actions over $6 golds. Yeah a few cards (Mine, Tax collector, etc.) can go direct to gold, but you can snag Harvest from University, Smasoning gold, and Remodeling Silver. When you can drop it into an effective deck it can save a turn or two in ramp up (e.g. Kc/Artificer/Margrave can drop in Harvest mid-turn and quickly add $12 to the buy total). University can be absolutely insane here where you go from maybe $5 worth of buying power to something like $20 in mid-turn.
2. There are now a good number of VP setups where it is preferable to gold: Vineyards, Arena, Bandit fort, Basilica, Colonnade, Defile shrine, Museum, Orchard, Tower, Triumphal arch, and maybe wall.
3. With deck top sorting cards, it can be useful to clear dross off the top. E.g. Apothecary leaves you some green and two random cards - discard and forget it. Top deck a province with a Count or Mandarin - dump it for cash. Even Pearl diver can make tiny marginal improvements on its ability to skip dross.
All told Harvest is just not THAT much cheaper than Gold. In the early game you will NOT be hitting better than Gold all that often and will have plenty of terminal silver hands. Worse when you do hit big hands, it will almost always be because you dumped a card you wanted to play (like a non-terminal trasher). Late game, you can save a $1 now instead of gold and later you can spend an action ... to net maybe $1 more than Gold. Sometimes +actions are that cheap and Harvest is not bad. Other times Gold or even silver is vastly more action and space efficient.