Sumner often says that increases in the quantity of base money are inflationary as long as they are expected to be permanent. According to Sumner, it doesn't matter if the money is created by the purchase of Treasury bills that have a zero nominal interest rate. While he doesn't dispute that this is just a swap of two safe, zero interest financial assets today, he argues that eventually the interest rate on Treasury bills is going to rise above zero. Assuming that the Fed doesn't raise the interest rate paid on base money with the T-bill rate, this is going to result in an excess supply of base money. Eventually, this will result in a higher price level.
Sumner usually explains this in the context of some extreme example. Suppose the quantity of base money is doubled, and the Fed has committed to keep it at that new level. It might seem that with short and safe nominal interest rates being very low, a doubling of base money along with a matching decrease in the amount of T-bills in the hands of private investors would have little direct and immediate effect on expenditures.
But consider 5 years from now. Unless interest rates on short and safe assets remain extremely low, then those holding all of that base money, probably banks, will use it to purchase what are now safe assets with higher interest rates. There will be a boom in lending and spending, and much higher prices for goods and services. As a rough rule of thumb, the price level will be approximately twice its current value.
This implies that there will be a 100% inflation rate over the next five years. While we could imagine that it all occurs between year 4 and year 5, this is very unrealistic. Households and firms would be motivated to purchase goods in year four rather than wait until year five when the goods will be twice as expensive. Since loans can be paid back with money worth half as much, borrowing at low interest rates to purchase these goods would be very attractive. Further, lending, including by holding securities with very low interest rates, rather than buying goods now, would be very costly. This is exactly what generates the Fisher effect, which would cause nominal interest rates between year four and five to skyrocket--in theory to something slightly higher than 100%.
But all of those arguments suggest that the inflation would not all be between year four and five. Prices would skyrocket between years three and four too. Perhaps all of the increase would occur between year three and four, with prices remaining more or less stable at the much higher level, between year four and five.
However, this reasoning applies between year two and three. Rather than wait for the high prices in year four, people would buy in year three. But if prices rise between year two and year three, the same arguments apply between year one and two.
And so, if base money is doubled, and it is going to remain doubled, even if it is all created by the purchase of zero interest T-bills, the result will be an immediate rapid inflation. Further, the interest rates on the T-bills will immediately skyrocket. The entire issue of zero interest T-bills and base money being identical financial assets would almost immediately disappear.
Of course, the quantity of base money has increased--more than doubled. And so, why no immediate inflation? Sumner's logic suggests there are only two possibilities. One would be that the near zero interest rate environment is expected to last forever. The other, entirely realistic possibility, is that the increase in the quantity of money is expected to be temporary. When short and safe interest rates begin to rise in five years, four years, or two years, the Fed will contract the quantity of base money as the demand to hold it falls. Or, perhaps the Fed will pay interest on that base money, particularly that part that takes the form of reserve balances, raising the amount it pays so that it is competitive with other short and safe assets.
I always find this argument irritating, because in my view, this is exactly what the Federal Reserve should do. The demand to hold base money rose in 2008 and remains quite high, and the Fed increased the quantity of base money to accommodate that added demand. When the demand to hold base money falls again at some future date, the Fed most certainly should decrease the quantity of base money accordingly.
Sumner claims that any fiat money issuing central bank that wants to inflate can. He argues that a permanent increase in the quantity of base money is always inflationary, regardless of the current level interest rates on short and safe assets.
While I think his arguments are persuasive, I also find them irrelevant. I don't want a central bank to try to inflate. I don't think the Fed should ever "permanently" increase the quantity of base money. Any increase or decrease in the quantity of base money should follow changes in the demand to hold base money, and so should be tentative, reversed if the demand to hold base money later moves the opposite way.
Sumner also has no interest in a permanent increase in base money or generating inflation. He favors a target growth path for nominal GDP, with the targets rising at a 5 percent rate. Suppose the Fed doesn't double base money in an effort to create massive inflation, but rather commits to get nominal GDP to its target growth path as soon as it can. Further, rather than a 5 percent growth path, suppose the growth rate is 3 percent, equal to the trend growth rate of potential output. The trend inflation rate will be zero.
Suppose the initial target for nominal GDP is $10,000 billion and nominal GDP is on target. During the first year, it falls 3% to $9,700 billion. At the same time, there is an increase in the demand for T-bills, and their yield falls to zero. The Fed does expand the quantity of base money to try to offset the decrease in velocity. But its open market purchase of T-bills just swaps one zero interest financial asset for another, and there is no motivation for added expenditure by the private sector. Instead, nominal GDP just remains stuck at $9,700 billion.
The Fed, however, doesn't change its target for nominal GDP. While nominal GDP is $9,700 billion, the target has increased to $10,300 billion after one year. Then it is $10,609 billion, and on and on. The Fed is falling further and further behind its target, but its open market operations in T-bills would seem to have no effect, because T-bill rates remain approximately zero, and swapping one zero interest rate financial asset for another would seem to have no effect on spending.
As above, in year 5, finally the demand for T-bills falls off, and so their yields rise. Now, the Fed's open market purchases of T-bills have some effect. By year five, the Fed's target for nominal GDP is approximately $11,600 billion. Compared to the level after year one, that requires an increase of nearly 20% to return nominal GDP to target, which the Fed can finally accomplish.
Consider three scenarios. In the first scenario, the productive capacity of the economy fell 3 percent in the first year and then has remained stagnant all five years. It so happened that nominal GDP exactly tracked the productive capacity of the economy, and the price level remained constant. Suppose the price index is 100. The approximate 20% increase in nominal GDP that the Fed will finally accomplish in year 5 would cause a 20% increase in the price level.
It is possible, as explained above, that this entire increase occurs between year four and five. And so there will be a 20% inflation rate over that period. However, such a scenario is very unrealistic. In year four, there will be a motivation to purchase goods before their prices rise. Borrowing at very low interest rates to fund those purchases will be very attractive. Further, lending by holding low interest rate securities rather than purchasing goods in year four will be very costly. The demands for goods would rise substantially in between year three and four, and nominal interest rates should rise quite a bit, to something close to 20%. As above, with the permanent change in the quantity of base money, this same argument applies in between year three and four, between year two and three, and between year one and two. In reality, the near 6% inflation rate that would be necessary to bring nominal GDP back to its target of $10,300 billion from its below target $9,700 billion level should immediate result in T-bill rates rising well above zero, making ordinary monetary policy effective.
The second scenario involves the same path for nominal GDP, it falls to $9,700 billion and stays there, but this time, potential output continues to grow 3 percent. Fortunately, all prices and wages are perfectly flexible, so that price level falls from 100 to 94 over the first year. Productive capacity rose from $10,000 billion to $10,300 billion, but spending on output fell from $10,000 billion to $9,700 billion. With a price level of 94, real expenditure rises 3 percent and real output grows with productive capacity.
As before, the Fed is trying to keep nominal GDP on target, but the open market operations supposedly have no impact. Remember, the T-bill rate is near zero and so open market operations swap just one zero interest rate asset for another. So, nominal GDP remains $9,700 billion. And each year, the price level falls roughly 3 percent more. By year 5, the price level is 84. And finally, the T-bill rates rise above enough above zero so that open market operations are effective. (I am not sure that deflation forever is impossible in this scenario. The real T-bill rate is 3 percent and for the nominal rate to rise above zero, the real T-bill rate most rise above 3%.)
Once conventional monetary policy becomes effective, the Fed raises nominal GDP back to target, which will cause the price level to rise back to 100. As before, this involves an inflation rate of approximately 20 percent between year four and five. This is unrealistic, because there would be a strong motivation to buy in year four before the prices rose. And further, nominal interest rates would rise to something like 20 percent. But, if that happens between year 3 and 4, the same argument applies between year 2 and 3 and between one and 2. When the price level approaches 94, the needed 6 percent inflation to get the price level back to 100 in year two would almost certainly cause the T-bill rate to rise well above zero and so make monetary policy effective almost immediately.
Market monetarists are usually most interested in a third scenario. In this scenario, nominal GDP falls to $9700 billion and stays at that level, while the productive capacity of the economy continues to grow. But actual production falls from $10,000 billion in the first year to $9,700 billion, and then remains at $9,700 billion. The price level, being sticky, remains at 100. As before, in year five, the T-bill rate finally rises significantly above zero, and the Fed can generate increased expenditure on output, so production rises approximately 20%, from $9700 billion to approximately $11,600 billion. All of the excess capacity, that has been growing over the years, is finally utilized.
Since the price level remains 100, there is no inflation between year four and five. Avoiding the massive inflation of prices that occurs between year four and five in the previous two scenarios is not an issue. However, the large increase in real income in year five is likely to impact demand for output in year four. For example, consumption smoothing suggests that those who expect a large increase in real income in year five would shift some of the consumption made possible back to year four. This could be done by reduced saving or even increased borrowing.
On the other hand, to the degree that the reduced income was due to unemployed workers earning little or nothing, then those who are already fully employed may expect little or no increase in their income and so do nothing different. The unemployed would have little ability to reduce saving or borrow against their future income. Still, at least some workers and entrepreneurs may be saving out of their currently reduced incomes and reduce that saving due to expectation of higher income.
Also, in a world where people lose jobs frequently, high unemployment implies that anyone who loses a job will have a more difficult time than usual finding a new one. Expectations that unemployment will fall will relieve such worries and so allow for reduced saving, and perhaps some debt financed purchased of consumer durables.
Equally important is the motivation of firms to purchase capital goods to expand productive capacity. Of course, if productive capacity had been maintained all along, with firms producing well below capacity yet still purchasing additional capital goods, this effect would not occur. But that is hardly realistic. One of the reasons why spending on output remains depressed is that firms have little motivation to purchase additional capital goods when their current sales are low. The massive increase in demand in year five is going to give firms an incentive to purchase needed capital goods in year four. (I am not at all sure that this can be managed without an inflation of capital goods' prices, but that just motivates the purchase of the capital goods in year four before their prices rise.)
The increase in demand for capital goods along with purchases of consumer durables results in increased credit demands. Further, firms that currently hold "cash," made up, for example, of the Treasury-bills that had been in such high demand, will sell them to buy capital goods. And some those those who have been saving by purchasing those bonds may sell off some of them to purchase consumer goods now due to the expected increases in income. This decrease in the supply of loans and increase in demand will tend to raise both real and nominal interest rates.
Again, the increase in demand in year 4, will result in increased output and income. And the increase in the demand for credit and reduced supply will raise interest rates, including T-bill rates, making monetary policy effective in year 4. Again the argument steps back to the initial year. Does an expected 6 percent increase in real output, from the initial reduction to $9,700 billion back to $10,300 billion cause the interest rate on T-bills, both nominal and real, to rise enough to make monetary policy effective immediately?
Realistically, some combination of scenario two and three would be likely if the problem were entirely a decrease in spending on output. A reduction in real output and a mild deflation would be the initial effect. The expected recovery of output would tend to generate higher demand in the present as well as a higher real interest rate. And the recovery of the price level, implies an inflation that motivates purchases at current low prices, and an inflation premium on the interest rate on T-bills.
Thinking about the last few years, it is likely that some element of the first scenario would apply as well. The slowdown in productivity growth in the context of nominal GDP targeting would result in a higher price level and inflation. This would tend raise the nominal interest rate on Treasury bills.
On the other hand, all of these processes require that many people have confidence that monetary policy will eventually regain its effectiveness. Of course, if the Fed hasn't even said that it is trying to target the growth path of nominal GDP, and will let spending on output languish at a low level in perpetuity, then making such a commitment now is one obvious solution to the problem.
Still, I believe that rather than limit open market operations to securities with a zero yield, the obvious next step is to expand base money to a point where longer term to maturity securities and even riskier securities are purchased. A target for a growth path for nominal GDP is necessary and might be sufficient. But if it is not sufficient, then large open market operations, including purchases of securities with yields above zero, would certainly be sufficient to break through any perverse expectations and bring nominal GDP to target.