"In response to the surging CPI in mainland China, some economists have proposed that the RMB be anchored to the prices of 50-100 goods. The goverment would fix a certain price for the basket (say 100 RMB), and adjust the money supply to try to keep that price fixed, no matter what happens to the economy, so that people can always buy such a basket with 100RMB. Is it feasible? And what are the advantages and shortcomings? Were there any precedence in the history of the world economy?" - Roy
This sounds like a version of price level targeting, which has been tried before- in Sweden, 1931-1937 (see this paper for a discussion). Normally, we would think of price level targeting as focusing on the entire basket of goods and services enjoyed by consumers by targeting the Consumer Price Index, but there's nothing to stop the Government targeting a select basket of goods and/or services.
Provided the central bank is given the automony to focus fully on its target (by adjusting interest rates or, equivalently, the money supply), such a policy target is achievable. If the price level increased above the target, the central bank would need to raise interest rates (decrease the money supply) to slow the economy sufficiently to get the price level to return to its target, and vice versa if the price level was below the target. This may not always be politically popular, but with sufficient autonomy, it can work.
The principle benefit of such a target is that it gives the Central Bank a numerical objective, and individuals can easily see whether or not the objective is being achieved. If indeed the target is achieved, then inflation expectations should remain under control, which should in turn make it easier to maintain stable price levels in future- all in a nice, virtuous circle.
This same benefit can in principle be conferred by any nominal anchor for monetary policy- for example an inflation target (as is common in many countries, starting with New Zealand in 1990), or an exchange rate target (China's official form of monetary policy target until recently), or any other numerical objective for that matter.
There may be some other benefits as well. Some people have argued that one problem with monetary policy is that Central Banks cannot credibly commit to future actions. In modern Macroeconomics models with sticky prices, the policy that the Central Bank should choose today depends on the policy that they will choose in the future. If they cannot commit to the policy that they will follow in the future, then they will be reduced to choosing a sub-optimal policy today.
A price level target can help to reduce the costs imposed by their inability to commit, because of the manner in which it ensures that the policy that the central bank will follow in the future is related to the policy that the central bank follows today.
I've written a couple of papers related to this, published in the Journal of Macroeconomics (see here and here). Personally, I'm sceptical that there are any gains to price level targeting by this mechanism, since the results are very sensitive to assumptions about how believable the central bank is, and how price setters set prices.