Let's call the entity to control over 50% of the hashing power Carlos so we can call it something.
Carlos' ability to double-spend doesn't mean he is able to create money. His double spendings are temporary and you can read up here how it works.
You can play around with double spending but doing it when buying stuff from others is illegal, just like paying with a check you know will bounce is illegal. In both cases, the person you allegedly payed to will not receive to money and you actively caused it to happen. This is fraud.
Anyone with enough money to control over 50% of the hashing power in the network doesn't care about scamming the tiny bit of money they can get away with without being caught and convicted of fraud. Billionaires don't care about allegedly giving $2000 to a shop when buying a good PC but getting their money back after they left the shop. And this is about the most expensive something gets without the seller knowing enough details to report you to the police for sure (like when you buy a car or a house).
Regular Joe isn't even affected by this problem. The only times he's ever paid are at the end of the month by his employer, and for a few gifts or exchanges of moneys by his family and friends. The latter surely aren't billionaires able to control over 50% of the hashing power and wouldn't try to defraud him anyways, and the former has a legally binding contract with him, requiring them by law to pay him. If they don't do so, Joe can sue them and win, and the prospect of winning probably doesn't get smaller if they maliciously defrauded him, rather than straight-up not paying him.
That's of course not to say that a single entity controlling over 50% of the hashing power is harmless. If Carlos owns 10% of the total hashing power, he makes 10% of the total profit from mining. If he owns 20%, he makes 20%. But if he owns over 50%, he makes all of the profits because he can create a consecutive of arbitrary length of blocks he created, therefore eliminating anyone else making profit.
It goes like this: Suppose, you mine block
n and your hashing power is << 50% total hashing power. You now have an interest of showing it to others asap so either you or they can mine block
n+1, securing the money you made by mining block
n. If you don't show your block
n to others, chances are, someone else's block
n is the one others use to mine block
n+1, rendering your block
n worthless. If you own over 50% of the total hashing power, suppose someone else mines block
n before you do. You want to have all the mining profits for yourself, so you continue mining your own block
n. Once you got it, you mine block
n+1 on of of your own block
n. Because you control over 50% of the hashing power, your block sequence
(n, n+1, ..., m) will reach some block
m faster than the block sequence
(n, n+1, ..., m) of everyone else. Once that's the case, you publish all of your mined blocks until block
m. This gives everyone else an incentive to mine on top of your block
m is the largest block number of any published block chain.
It's already possible to gain such an advantage when controlling less than 50% of the hashing power but above 50% of it, it's easy to understand.
This in fact is a big problem, as you can see. For the most part for the miners, of course, but also for the people using the crypto currency. This is because it's against one of the ideas Bitcoin was founded upon: Taking control of payments away from the few power holders currently controlling them.
However, it is to be noted that an entity controlling over 50% of the hashing power would harm itself a lot when taking profits away from other miners as trust in the currency would diminish, taking away its own profit's, too. Even more so if it prevents or hinders payments.
A good counter measure I see for this is making the hashing function use a lot of the instructions of a modern CPU with a very large instruction set like the one the x86 processor family features. This makes designing specialized circuits very hard, meaning that that original idea of Bitcoin being mined on people's home PCs would come true, again. The resistance to specialized circuits being designed for it would have to be so big that the possible advantage one can get from using one of them is smaller than the profits, compared to mining using the desktop PC one buys for purposes other than mining Bitcoin.
However, while having lots of upsides, this idea also has a downside: People had to agree on a different hash function when the x86 processor family is being superseded by a different architecture.
Furthermore, the people using Bitcoin had to be persuaded that switching to a different hash function is a good idea. That's especially hard because miners already invested a lot of money in their specialized hardware. But if there is an entity threatening the entire project, people will be willing to take counter measures.