Comp/bonuses went up for some. It went to zero for many others. In 1985, banks had armies of line workers processing trades, managed by officers making decent money. The average comp at a bank might be the average of 1 officer and 100-500 low level line workers. It would be better than other office jobs, but not great.

In 2010, the officer specs out business requirements after talking to traders and regulators, sends the high level work to IT and the CRUD apps to Bangalore. 475 of the 500 line workers have been replaced by a data center in Nutley, NJ, 5 remain in the US handling cash wires and other time-zone sensitive work, and 20 work via email out of an office park in Pune.

For obvious reasons, the average comp of 1 officer + 3 IT guys + 5 US line workers is going to be far higher than the average comp of 500 grunts + 1 officer.

Additionally, for many reasons (IT being one of the biggest ones), it's also vastly easier today to strike out on your own than ever before. In 1985, if you wanted to quit and start a hedge fund, it was tough. Today, a hedge fund can be operated by 3-4 guys + servers. Banks are forced to pay their traders enough money to prevent them from quitting and doing exactly this.

(Remind you of another high comp industry, with established players giving out big comp packages to retain talent?)

Edit (in response to your edit): RE: hiding the big trades in noise. What's wrong with that? Why should they have to develop a special HFT algorithm in order to simply move shares around without getting their lunch eaten? This is exactly the kind of value extraction I'm talking about. HFT shops have now created a huge expense for these retail shops that didn't exist before..

When you make a large trade, there will be a price impact. This is unavoidable - if you are selling millions of shares, supply went up, and price must drop.

In 1985, institutional investors (not retail investors) tried to make sure that other people (their counterparties) suffered the price impact. They would sell shares at full price, retail investors (i.e., my Mom) would buy from them, and then shares would go down once the market figured out supply dramatically increased.

In 2010, some HFT shops will sell to my Mom at a favorable price, and then buy from the institutional investor after prices go down. This benefits the HFT shop and retail investors at the expense of the institutional investor. I'm not saying this is good or bad, I'm saying it benefits "dumb money" (little guys who can't afford to rent an algorithm) at the expense of "smart money" (Goldman, Vanguard, Calpers Pension Fund).

[edit2: just a disclaimer - specific numbers/cities should not be taken as precise. The data center might be in Jersey City or Brooklyn, Pune might be Cebu or even Baltimore, and there might have been 250 or 1000 line workers rather than 500.]

Corporations can now protect themselves from all sorts of risks which weren't possible before using swaps and future instruments.

Innovations in debt financing drove the huge economic growth of the last two decades. Mobile phone technology would never have taken off if cellphone companies couldn't have used debt financing to build huge cell tower infrastructures.

CMOs for all the problems they caused, did allow for a huge increase in home ownership. For every default there are dozens of paid-up home owners who could never have afforded to buy a house otherwise.

Electronic Government bond auctions have driven down the cost of government borrowing, leading to lower taxes.

Forex costs have dropped massively. Look at your laptop. It's likely that during the manufacture dozens of countries were involved and likely hundreds of currency transactions. You paid less for your laptop because of a decade of innovation in the FX markets.