Willem Buiter argues in favour of negative nominal interest rates in his FT maverecon blog. He identifies the bearer security nature of currency (whose owner remains anonymous) as the fundamental cause of the zero lower bound on nominal interest rates and discusses three possible strategies to relax the lower bound.

First, to abolish currency. As a consequence, central bank seignorage might fall and criminals would need to find new stores of value that guarantee anonymity. Limited privacy could be preserved by ‘cash-on-a-chip cards’ and for practicality reasons, small denominations could be kept. The price of the remaining cash expressed in terms of electronic money would fluctuate, however.

Second, to tax currency. Since cash can be held anonymously, this poses difficult incentive problems. It could be done but would be complicated and costly.

Third, to unbundle two functions of money, namely the medium of exchange/means of payment function on the one hand and the numéraire/unit of account function on the other. Suppose that there are two dollars, one unit of account dollar or “dollar” and one medium of exchange dollar or “m-dollar” (Buiter talks of “rallods” rather than m-dollars). Central bank reserves constitute dollars and might pay positive or negative interest while cash constitutes m-dollars and does not pay interest (or at least not negative interest). Monetary policy is conducted as usual by setting interest rates on dollars. In addition, the stock of m-dollars is fixed by the central bank, letting the market determine the exchange rate between dollars and m-dollars; or the central bank fixes an exchange rate between the dollar and m-dollar and elastically supplies m-dollars at this rate. In either case, the exchange rate will typically differ from unity and vary over time, in contrast to the current situation. Zero interest on the m-dollar and non-zero (positive or negative) interest on the dollar are consistent with no-arbitrage as long as the appreciation or depreciation of m-dollars relative to dollars compensates for the interest rate differential. For instance, if the central bank sets a negative interest rate on dollars, then the price of m-dollars (which pay zero interest) expressed in terms of dollars must fall over time that is, m-dollars must depreciate relative to dollars.

According to Buiter the third strategy suffers from just one possible problem: If for some reason, the numéraire ‘followed the currency’ and people started to quote prices in m-dollars then nothing would have been gained. But he argues that the government has means to coordinate society on using a specific money as unit of account, for instance by requiring taxes to be paid in that money (that is, in dollars rather than m-dollars).

Buiter refers to contributions by Eisler (1932), Goodfriend (2000), Buiter and Panigirtzoglou (2001, 2003), Davies (2004), Buiter (2004, 2007) as well as Mankiw’s blog post (April 19) on a graduate student proposal to depreciate cash by means of a lottery.

In another blog post a few days later, Buiter offers further discussion and a rather sarcastic comment on option two:

Taxing currency will, I am afraid, remain rather intrusive and administratively cumbersome.  This may of course recommend it to some of our leaders.

He also notes that Charles Goodhart has been talking for years about the lottery proposal by Gregory Mankiw’s graduate student. And he points out that this proposal is not fool proof: Even when a lottery rendered bank notes with a specific last digit in their serial number “officially” worthless people might still continue to value them; confiscation threats etc. might therefore be needed in addition to the lottery in order to sustain the scheme.