Pieria #2—>The Costs and Benefits of Repealing the Zero Lower Bound...and Then Lowering the Long-Run Inflation Target

blog.supplysideliberal.com tumblr_inline_mvx539vIjs1r57lmx.png

Link to the Column on Pieria

Here is the full text of my 1st Pieria exclusive “Going Off the Paper Standard,” now brought home to supplysideliberal.com. It was first published on Pieria on October 28, 2013. 

This post complements my recent column “Larry Summers just confirmed that he is still a heavyweight on economic policy," which could have been called "Larry Summers and the zero lower bound.” In brief, Larry Summers gave a powerful speech at an IMF conference, emphasizing the costs of the zero lower bound–which might include the kind of “secular stagnation” (Larry’s words) that Japan has suffered in the last two decades. I then argue that we should simply eliminate the zero lower bound.

But I did not explain in “Larry Summers just confirmed that he is still a heavyweight on economic policy," why we shouldn’t just steer away from the zero lower bound by engineering higher inflation (assuming we can). This Pieria post on the costs and benefits of inflation in the absence of the zero lower bound makes that case. (Also see the Powerpoint file for my November 1, 2013 presentation at the Federal Reserve Board, and my Twitter discussion with Daniel Altman on the costs and benefits of inflation in the absence of the zero lower bound.)

In ”Larry Summers just confirmed that he is still a heavyweight on economic policy,“ I address the politics of eliminating the zero lower bound by saying

Politics will stay the same until a critical mass of people do what it takes to make them different. Summers proved at the IMF conference that he is still an economic policy heavyweight—someone who could contribute a lot toward reaching that critical mass in the war against the zero lower bound, if he is willing to join the fight.

I don’t know Larry’s views on repealing the zero lower bound in the way that I advocate, but Larry Summers’ IMF talk has led to discussion in other quarters about eliminating the zero lower bound. (Update March 15, 2018: Larry Summers is now an advocate of eliminating the zero lower bound and occasionally refers favorably to the proposal I have made for how to do it. I know this mainly by personal communications with Larry and others whom Larry has talked to. In print, you can see it here.) Matthew Yglesias renewed his advocacy of abolishing paper currency in "The Biggest Problem in Economic Policy Today” a few hours after my column appeared. Brad DeLong picked up on my column here, and Paul Krugman picked up on Brad DeLong’s post in his “Secular Stagnation, Coalmines, Bubbles, and Larry Summers.” (And others are picking up on Paul’s post.) In his post, Paul said the most positive thing I have seen him say so far about negative nominal interest rates as a real-world policy:

If the market wants a strongly negative real interest rate, we’ll have persistent problems until we find a way to deliver such a rate.

One way to get there would be to reconstruct our whole monetary system – say, eliminate paper money and pay negative interest rates on deposits. 

Finally, Dylan Matthews of Wonkblog interviewed me last Thursday about repealing the zero lower bound to add negative interest rates to the policy toolkit. That interview might appear even as early as today.

If you want to mirror the content of this post on another site, that is possible for a limited time if you read the legal notice at this link and include both a link to the original Pieria exclusive and the following copyright notice:

© October 28, 2013: Miles Kimball, as first published on Pieria. Used by permission according to a temporary nonexclusive license expiring June 30, 2015. All rights reserved.

Historically, there have been many different monetary systems. Tom Sargent surprised me with the range of monetary systems that have existed in the United States since 1776 when he presented his paper “Fiscal Discriminations in Three Wars” at the University of Michigan this Fall. Nevertheless, we have become used to our current monetary system, and have gained useful experience with it, so any proposal to change it should be carefully justified.My efforts in that regard are laid out in “How and Why to Eliminate the Zero Lower Bound: A Reader’s Guide.”  By “our current monetary system” I mean the monetary system of most advanced economies and most emerging economies in 2013. The proposed new monetary system I call an “electronic money system” because the electronic dollar, euro, yen, pound, or the like would be the unit of account. 

The brief appeals for eliminating the zero lower bound at the bottom of “How and Why to Eliminate the Zero Lower Bound: A Reader’s Guide.” plus my column “America’s Big Monetary Policy Mistake: How Negative Interest Rates Could Have Stopped the Great Recession in Its Tracks” are my attempts to show how the polemics for repealing the zero lower bound could be approached in the political arena.  The links collected under the heading “Operational Details for Eliminating the Zero Lower Bound” in “How and Why to Eliminate the Zero Lower Bound: A Reader’s Guide.” address the nuts and bolts of eliminating the zero lower bound. 

For an overview of the operational details, I recommend starting with my Pieria post “Going Off the Paper Standard.” What is missing in my justification for changing our monetary system is a point by point tally of the costs and benefits of repealing the zero lower bound. Hence this post: costs first, then benefits. My title reflects an important complementarity that exists between repealing and lowering the long-run inflation target.

The Costs of Repealing the Zero Lower Bound

Repealing the zero lower bound as I have proposed is not without some costs. The most obvious cost is the extra computation needed to deal with what is, in effect, an exchange rate between paper currency and electronic money that periodically inches away from par during serious recessions, and then gradually returns to par after the recession is over. But for consumers, this computational cost is of a similar type to the computational cost of dealing with sales taxes that are added on to the price of purchases, and for business people, it is much easier than many other computations they need to make. And for both consumers and business people, any computational cost from an exchange rate between paper currency and electronic money is likely to apply to a smaller and smaller share of goods as technological change makes the use of electronic money look progressively more convenient compared to paper currency.  

The most important costs of repealing the zero lower bound are costs of the negative interest rates themselves. Given the level of inflation, going from zero to negative interest rates has all of the usual costs and benefits of lower short-term nominal interest rates and lower short-term real interest rates, including important distributional effects. In addition, nominal illusion makes even the concept of negative interest rates unfamiliar and confusing to some people. Beyond any direct psychological distress confusion about negative interest rates causes, that confusion could cause harm by opening up new strategies for financial hucksters and bubble-mongers.  

Additional costs could arise if political or legal constraints prevent the full policy prescription from being followed. Most important among these are the extra dangers to financial stability if equity requirements for banks and financial firms are not raised substantially beyond anything in current law. Also important are the distortions to the intent of financial contracts if amounts of money specified in old contracts that are ambiguous are interpreted as amounts in paper currency rather than according the electronic unit of account.  

The Benefits of Repealing the Zero Lower Bound

Direct Costs of the Zero Lower Bound.The benefits of repealing the zero lower bound come from avoiding the costs of keeping the zero lower bound. The obvious cost of the zero lower bound is in preventing a central bank from lowering its short-term interest rate when negative interest rates would be helpful for macroeconomic stabilization. That includes not only the cost of having less stimulus than would otherwise be optimal, but also the cost of using other ways to stimulate the economy, such as those arising from

  1. the deficits traditional fiscal stimulus generates,
  2. the unusual spreads that large-scale purchases of long-term government debtgenerates,
  3. the danger of reigniting a bubble in home prices by large-scale purchases of mortgage-backed securities, and
  4. any reduction in the responsiveness of monetary policy to future needs that forward guidance engenders. 

Indirect Costs of the Zero Lower Bound Through the Long-Run Inflation Target. In addition to such direct costs of the zero lower bound and responses to a currently binding zero lower bound, there are costs from efforts to avoid running into the zero lower bound in the future. In particular, if there is any fear of these direct costs of the zero lower bound, central banks are likely to choose long-run inflation targets that are higher than they would otherwise choose in order to take into account the danger from these direct costs. The zero lower bound should not be taken as a given. But if it is, many find the logic behind tilting the inflation target higher to steer away from the zero lower bound compelling. Ben Bernanke gave the conventional view for an inflation target at 2% rather than zero in his March 20, 2013 press conference, saying:

… if you have zero inflation, you’re very close to the deflation zone and nominal interest rates will be so low that it would be very difficult to respond fully to recessions. And so historical experiences suggested that 2 percent is an appropriate balance …

And Brad DeLong counts himself, Olivier Blanchard, Larry Ball, and Paul Krugman as serious advocates of an even higher 4% inflation target due to their worries about the zero lower bound.

The contrary view is that a nominal anchor such as price level targeting or NGDP targeting can make running into the zero lower bound so uncommon that the optimal inflation rate would be quite low even with the zero lower bound. Olivier Coibion, Yuriy Gorodnichenko and Johannes Wieland found this in a formal model for price level targeting. Scott Sumner argues the corresponding view nominal GDP targeting.  (Scott Sumner also argues that nominal GDP targeting will do the trick even when the economy actually up against the zero lower bound.) But claims that with a better monetary rule the zero lower bound would be easy to avoid even with a low inflation target remain speculative. I advocate both repealing the zero lower bound and following a version of nominal GDP targeting that leans in the direction of price-level targeting.

The Costs of Inflation

As argued above, gauging the benefits of repealing the zero lower bound requires assessing the costs and benefits of inflation in the long run. My goal here is to point out how those costs and benefits would be affected by the repeal of the zero lower bound. What remains is to examine how the repeal of the zero lower bound affects the other costs and benefits of inflation. The answer is not immediately obvious because my proposal involves at some points a higher rate of inflation relative to paper currency than to the electronic money that serves as a unit of account. So it is important to pay attention to whether each cost or benefit of inflation is about inflation relative to the unit of account, or inflation relative to paper currency. 

Messing Up Price Signals. Many of the costs of inflation have to do with messing up price signals in one way or another. Sticky prices and sticky wages mess up prices signals to some extent even in the absence of trend inflation. But unless price changes are fully synchronized across firms, trend inflation tends to lead to different prices leap-frogging each other in a complex dance that distorts signals about which goods have the lowest social costs. This is a potential issue for

  • varieties of final goods
  • varieties of intermediate goods
  • varieties of labor inputs
  • leisure over time
  • each good over time

Every one of these costs has to do with the setting of sticky prices–including sticky wages (the prices of labor and of leisure). So for these costs, it is inflation relative to the unit in which sticky prices or wages are set. My contention is that (with the measures discussed in “Going Off the Paper Standard.” ) retailers can successfully be encouraged to set almost all prices in terms of the electronic unit of account, with a single store-wide conversion factor for converting the electronic price of a bundle to the amount of paper currency that would be charged for those who prefer to pay in paper currency. The choice of conversion factor itself is likely to be determined in large measure by retailers’ costs from credit and debit card fees, desire to price discriminate and some desire to keep paper and electronic prices equal. If there is only one conversion factor for the entire line of goods at a given retailer, I would expect it to be relatively flexible once it departed from par. (Not only should the menu costs for a single conversion factor be low, the information relevant for deciding on the conversion factor is relatively straightforward.) If so, once the conversion factor is away from par, the stickiness would be in terms the electronic unit of account. To the extent sticky prices and wages are set in terms of the electronic unit of account, zero inflation relative to the electronic unit of account minimizes the distortion of price and wage signals from sticky prices and sticky wages.

As for the initial stickiness of the conversion factor at par, I have argued on several occasions that the initial stickiness of the conversion factor at par could ease acceptance of the exchange rate between electronic money and paper currency, since there would be a few months near the inception of the electronic money system in which households could obtain paper money from the bank at a discount, but have it accepted at par by retailers.

What If Some Prices Are Set in Both Electronic and Paper Terms?  Even if most goods are priced in terms of the electronic unit of account, it may be that there is also a paper currency price for relatively inexpensive goods that are frequently (though not always) purchased with paper currency. These are goods for which “convenient prices” in Ed Knotek’s sense matter a lot, so a formal analysis would be complicated. But they are goods that have a relatively small budget share and are unlikely have a big effect in inducing people to go to the wrong store (“wrong” from a social-welfare-maximizing point of view). And, once a customer has chosen whether to use paper currency or electronic money to pay, the within-store price ratios are unaffected by the existence of both an electronic and paper price for these items. (I suspect that most stores can adequately discourage most people from purchasing some items with paper money and some with electronic money. The choice of paying by electronic money or paper money becomes interesting when there are these two sets of prices.) The bottom line is that these effects are likely to be complex in many ways(including depending on both inflation relative to the electronic unit of account and inflation relative to paper currency), but small.

Resources Used Up By Menu Costs. Menu costs are incurred by changing sticky prices. So they are also affected by inflation-induced price-leapfrogging. For the direct use of resources to pay menu costs, what matters is the unit in which prices are set. If prices are set in terms of an electronic unit of account, menu costs will be minimized at zero inflation relative to the electronic unit of account.

Causing Confusion.The costs under the heading “messing up price signals” all persist in models in which all the agents are infinitely intelligent and optimize fully subject to (sometime ad hoc) constraints. There are serious additional costs of inflation that arise when cognition is finite. I consider these the main costs of inflation. Let me list some of the likely types of confusion and some of their consequences.

  • making people blame something they call “inflation” (though it is not the general rise in prices and wages that macroeconomist refer to when they say “inflation”) for the fact that their real wage is not higher than it is
  • causing unintended distortions in the tax code, and in particular a higher effective rate of capital taxation than elected representatives may have intended
  • leading people to mistake nominal rates of return for real rates of return when deciding how much they need to save for retirement
  • muddling intertemporal comparisons more generally.

Greg Mankiw gives this parable about the cost of muddling intertemporal comparisons in his best-selling Brief Principles of Macroeconomics textbook (p. 260):

Imagine that we took a poll and asked people the following question: ‘This year the yard is 36 inches. How long do you think it should be next year?’ Assuming we could get people to take us seriously, they would tell us that the yard should stay the same length—36 inches. Anything else would complicate life needlessly.

All of these costs are from inflation in the unit of account, where in this case it is the most literal sense of “unit of account” that matters. As long as people are thinking in terms of electronic dollars, euros, yen, pounds, etc., confusion costs will be minimized by zero inflation in the electronic unit of account.

Note that having zero inflation in the electronic unit of account would, in turn, encourage people to think in those terms. In addition, public education, accounting rules, and the tax system can be used to explicitly encourage households to think in terms of the electronic unit of account.

Unpredictability of Inflation. Zero inflation is quite focal for many decisions. So zero inflation is likely to minimize the costs of unpredictable inflation. Being focal has to do with the yardsticks people have in their minds. Thus, this is about inflation relative to the unit of account.

Causing People to Use Too Little Paper Currency. Socially, there is very little directcost to providing paper currency. To the extent that paper currency provides convenience and helps avoid transactions costs associated with credit and debit card transactions, private costs to using paper currency will lead people to use too little paper currency. As long as paper currency is at par with electronic money, the key private costs to using paper currency are

  1. the chances of theft,
  2. the gap between the checking account interest rate and the paper currency interest rate, and
  3. the “shoe-leather costs” of making more trips to the ATM in order to keep the first two costs down. 

The key point here is this: although an electronic money system sometimes has a negative paper currency interest rate, that would occur when checking account interest rates are very low or negative. That is, the spread  between the checking account interest rate and the paper currency interest rate can be kept small–except when paper currency is already back to par and checking account rates are at distinctly positive levels. (Note also that if the inflation target is lowered, nominal interest rates won’t be as far above the zero paper currency interest rate when paper currency is kept at par.) Thus, any substantial costs from people using too little paper currency would arise from

  • a choice of the central bank to leave some spread between the paper currency interest rate and the target interest rate so that retail banking as we know it could continue to make non-negative economic profits, and
  • a choice of the central bank to keep paper currency from going above par to obtain the benefits of paper currency being at par much of the time.

If neither of these considerations were a concern, the central bank could keep the paper currency interest rate equal to the target rate at all times–or even above it by the extent of the theft rate–to avoid all costs coming from people using too little paper currency.

Let me try to drive home the point with these additional remarks:

  1. Costs coming from too little use of paper currency are primarily about the spreadbetween the paper currency interest rate and other interest rates, not about the level of these rates. They have nothing to do with the rate of inflation per se either relative to electronic money or relative to paper currency, except when paper currency is being kept at par. When paper currency is being kept at par, lower inflation will lead to lower nominal interest rates on everything but paper currency, and so will lead to less underuse of paper currency. 
  2. If there are any serious problems from too small a spread between paper currency and other interest rates, the central bank’s ability in an electronic money system to choose the paper currency interest rate can help avoid these costs.
  3. The point of making it possible to have negative paper currency interest rates (by time-varying paper currency deposit fees) is not to disadvantage paper currency. Rather, it is to make it so that there is nowhere to hide from the negative interest rates (either in paper currency or in the bank) without taking on risks in a way that reduces risk premia, buying goods or services, or generating capital outflows and thereby stimulating net exports. (If negative interest rates, both in the bank and in paper currencies prevailed worldwide, then the only place to hide would be by taking risks in a way that reduces risk premia and thereby leads to additional physical investment purchases,  or by directly buying goods and services.)

Will Having Paper Currency Away from Par Discourage People from Using It? The one remaining issue about how much paper currency is used is the effects of being away from par. To the extent people get a discount on paper currency at the bank in a way that exactly makes up for the extra paper currency needed to make purchases at the store, the effects on use of paper currency should wash out, except to the extent the extra computation cost for paper currency discourages its use. But that effect should be overwhelmed by the fact that retailers can choose the conversion factor to make those who purchase with credit or debit cards pay for the extra transactions fees. Thus, under an electronic money system, the true resource cost of credit and debit card transactions is likely to be somewhat better transmitted to the customers who make the decision of whether to use credit or debit cards or paper currency. Getting what is in effect a “cash discount” some of the time should encourage people to use paper currency more, in a way that gets closer to the socially optimal level of use of paper currency. 

The Possible Benefits of Inflation

All of the benefits of inflation come from a second-best narrowing of some other distortion. Here are the three logical possibilities that I am aware of. The first depends directly on the statutory unit of account used for tax calculations. The other two depend on the unit in which prices and wages are set, which is also likely to be the electronic unit of account.

Raising the Effective Rate of Capital Taxation (If that is Good Rather than Bad). In my view, rates of capital taxation are too high. If I am right, one additional cost of inflation is that it raises capital taxation even further, when capital taxation is already too high. But if one thought that elected representatives had set rates of capital taxation too low, one might be in favor of higher inflation to raise the effective rate of capital taxation. Here it is important to recognize that the models that are the most favorable to capital taxation involve relatively sudden capital taxation, either once at the beginning of fiscal time, and never again (with the problem of needing a clear legal definition of the beginning of fiscal time) or during particular bad contingencies. Thus, that kind of capital taxation cannot be achieved by steady inflation, and is correspondingly more costly and dangerous.

Making It Easier for Firms to Lower the Real Wages of Particular Employees. There is a considerable amount of evidence that it is difficult for firms to lower nominal wages because of negative effects on the morale of both the employee whose wage is cut and all the other employees to whom that one complains. When inflation is positive, real wages can be lowered by leaving nominal wages the same, or increasing nominal wages by only a small amount. The lower inflation is, the more difficult it is to lower real wages without lowering nominal wages. The trouble with not being able to lower real wages is that a firm might then want to reduce how much it uses employees whose marginal product has declined, but whose real wage has not. It could lay those employees off or cut their hours. Both options are socially inefficient compared to reducing those employees’ wages and continuing to employ them fully.

This cost of blocking otherwise appropriate cuts in the real wage is a potentially important benefit of inflation. However, I think it is relatively easy to deal with this issue in ways other than inflation. In particular, having a substantial portion of pay in an annual bonus makes it much easier to reduce annual nominal wages. This is the way things work for a large share of firms in Japan. In a very low or zero-inflation environment, it is likely that firms would gravitate toward this solution on their own. But it is also straightforward to encourage this kind of solution by public policy, as Martin Weitzman details in his underappreciated classic The Share Economy, which is the bible for minimizing whatever costs are caused by nominal wage rigidity (including the costs of messing up price signals discussed above).   

Leading Firms to Lower Their Markups of Price Over Marginal Cost and Wage Setters to Lower Their Markups of the Wage Over the Opportunity Cost of Time. Because of discounting of the future, a sticky price or wage will be adapted somewhat more closely to the immediate future than to the more distant future. First consider firms. Since inflation tends to give an increasing track to the price a firm would want to have absent any costs or limitations on price changing, the immediate future of the firm tends to suggest a lower price than the more distant future. Thus, heavier discounting should interact with positive inflation to encourage the firm to have a lower price. If price is above marginal cost to begin with, a lower markup tends to increase efficiency. But firms should discount at the shadow interest rate they should use to evaluate investment projects. The interaction of this interest rate with the amount of inflation that occurs in other prices while a particular price is fixed creates only a small effect.

For workers involved in setting wages above the opportunity cost of time, Liam Graham and Dennis Snower have argued in their well-written paper “Hyperbolic Discounting and Positive Optimal Inflation that workers who disagree with their future selves about the way to discount one month relative to the next might plausibly have very high effective discount rates. They go on to argue that then inflation could have a significant benefit in leading these present-biased workers to accept lower wages, which would be closer to their opportunity cost of time.  In my view, their story, while intriguing, seems fragile. It depends on work hours being determined by contractual wages at a quite fine-grained level. It also requires believing workers could see this as a significant issue and still not press for more frequent wage adjustments. Finally, it depends on workers being at once hyper-rational and internally conflicted, when experiments by Daniel Benjamin, Sebastian Brown and Jesse Shapiro (reported in their paper “Who is ‘Behavioral’? Cognitive Ability and Anomalous Preferences” ) suggest that present-bias is associated with low cognition.  It is probably safe to say that Liam Graham and Dennis Snower have given a best case scenario for their effect, in finding that it raises the optimal level of inflation in a model with no zero lower bound issue from zero to 2.1%.

The Bottom Line for the Long-Run Inflation Target

Whatever the optimal target for long-run inflation is when there is a zero lower bound, the optimal target for long-run inflation is likely to lower in the absence of a zero lower bound. The overall benefit of repealing the zero lower bound is

  • the benefit of repealing the zero lower bound would have if the long-run inflation target held fixed, PLUS
  • the benefit of lowering the long-run inflation target from its previous value to whatever value is optimal in the absence of the zero lower bound. 
  • I remain unimpressed by the purported benefits of inflation other than steering away from the zero lower bound. I will be surprised if a nation that repeals its zero lower bound does not also gradually lower its long-run inflation target to zero.


Repealing the zero lower bound has some costs, but those costs should be weighed against the benefits: not only ending recessions, but also ending inflation. The key analytical point is that by and large the costs of inflation are costs of inflation relative to the unit of account.Thus,

  • if electronic money provides the unit of account (including the unit of account for price and wage setting),
  • and inflation is close to zero in terms of the electronic unit of account,
  • then one can have inflation relative to paper currency without serious costs,
  • as long as the central bank keeps the spread between the paper currency interest rate and the checking account interest rate small.