Many, especially academic economists (particularly on the left - Krugman, DeLong - but also some on the right) have derided the 'financial innovation' performed by Wall Street as useless and say that we'd be better off with less of it.
This is a willfully simple view.
Instead of financial innovation, which people don't see every day, let's talk about cars.
Your most basic car is like the Model T. It has a steering wheel, it has four wheels and an engine, and it runs on gasoline. It doesn't go particularly fast. It's functional, but nothing special.
As the car evolved, it got lots of features. Some of them, like seatbelts, airbags, and anti-lock brakes, are explicitly safety features. Occasionally they can be a source of harm (as with the airbag decapitation debate), but for the most part, they save lives.
Some of them are convenience-oriented. Things like power steering, remote keyless entry, etc. have definite value. Occasionally they can be useful in safety situations, but for the most part, they just make our lives easier, and don't increase risk substantially.
Then you have the personalization-oriented tweaks. The purpose of these is so that you can "do more" in a car. People who want to cart lots of people can get a large SUV. People who want to save money on gas can buy a Prius. People who want to go fast can buy a sports car. People who want to haul things can buy a pickup truck. People who want to feel the wind can buy a convertible. People who really want maneuverability get motorcycles. Etc.
Here's the funny thing about the personalization: We'd all be a lot safer without it. Sports cars encourage speeding that can kill people. SUVs are prone to rollover, and when they crash into other cars, those other cars are demolished. Convertibles are way more dangerous than regular cars, and motorcycles kill people at something like 28 times the rate of regular cars. Everyone would be a lot safer if we took the Model T, added the safety features and convenience features, chose the body type that killed people the least (perhaps a sedan, or maybe a station wagon/minivan) and made all of our cars like that.
Now please go look in your garage, assuming you have one. Unless you are driving a Volvo S80, a Ford Five Hundred, or one of the other small handful of cars that top the charts in safety ratings and are not massive, you are violating the "safety first" code - and driving kills a lot more people than banking!
So why do we accept this variability in cars?
Generally speaking, the heterogeneity of cars provides substantial benefits. Maybe SUVs kill people at a higher rate than sedans, but for soccer moms, they improve the quality of life immeasurably. If we knew everyone would drive well, we want every soccer mom to have access to a SUV because for her, it is significant.
Similarly, if we knew everyone would drive well, sports cars should be more common! They're maneuverable, they reduce stress, and it makes people happy.
The same logic applies to financial innovation.
There are some features of the financial system which are explicitly safety features. Giving customers access to government bonds, for example, lets them not rely on their own bank's solvency, which takes the FDIC off the hook for their deposits. Similarly, linking bank accounts brokerage accounts to access other non-bank securities diversifies the consumer's counterparty exposure through convenient access to other capital markets, also sparing the FDIC. Mandated interbank check acceptance is a way of gaining confidence in the banking system and not limiting bank use to one or two large brands.
Others are convenience innovations. The ATM doesn't guarantee the financial stability of the US, but it is certainly a major improvement to banking services in terms of quality of life. Linking credit cards to accounts isn't necessary, but as someone who has both a linked credit card and an unlinked credit card, the link makes it much easier to manage my monthly payments. These are clearly convenience benefits for me. Similarly, the system itself has convenience clauses. Each bank could probably negotiate its own interbank interest rates, but LIBOR makes it so they don't have to (and yes, I know this is a holy-mother-what-a-simplification given the current system, but it is conceptually consistent, I believe)
And finally, others are in fact personalization issues. Through a bank, airlines and refineries can hedge their fuel costs in opposite directions to reduce their individual risk. International companies can hedge away their currency risk. Banks and insurance companies can hedge away their interest rate risk, banks can diversify and hedge default risk, etc. These complex transactions increase the overall risk of the system, but like an SUV, they also facilitate a lot of beneficial activity that simply couldn't happen. I understand that we're overlevered, but on the margin, credit is still helpful, so a complex transaction that reduces risk for a bank may facilitate more credit and thus economic expansion.
The problem is not that they create risk, it's that these transactions make risk harder to measure accurately, and if you are not conservative enough in your estimate, you can end up in trouble because you overestimated how much additional risk you could take on elsewhere based on the risk you mitigated in that transaction.
Thus, killing "financial innovation" and deriding it as evil and stupid misses the point. The problem isn't too many innovative bankers, it's not enough bankers capable of assessing risk well. When you have relatively inelastic demand for banking, uncertainty as to the quality of each individual banker, and too few high quality bankers to service that large demand, you have a recipe for high salaries and positional externalities, as well as the potential for periodic systemic overleveraging by bad bankers believed to be good, which is exactly what happened. You don't fix this by outlawing the SUV, you fix this by ensuring that people are better drivers and SUVs are safer in crashes.
Reducing the interconnectedness of banks (which is far, far more important than size - too big to fail is a dumb concept, as i've outlined a million times before - check prior posts) is pivotal to ensure that these SUVs are, in fact, safer in crashes. This can be done with things like creating derivatives exchanges to minimize counterparty risk. Banning prop trading by banks has a couple negative features - it's hard to know how much this will reduce access to risk reduction measures, if the bank takes the other side of trades, but if most transactions have counterparties anyway, then you functionally just trade away severe periodic crises for more frequent but less severe crises (by reducing the diversity of bank exposure).
Killing financial innovation entirely, though, as many on both sides of the aisle seem to want, and restricting banker pay, and capping the size of banks, and all of these drastic, drastic measures, will affect the "real" economy far more than people seem to be acknowledging anywhere because by increasing the risk banks need to take on each individual trade, you reduce the credit they can make available. The banks are so unpopular now (and in some cases, rightly so - high frequency trading, flash trading, M&A bankers with terrible incentives, etc) that they can't articulate this, and nobody else seems to be willing to do so.