Do monetary frictions matter?

In some cases, maybe. But not in understanding the effect of conventional monetary policy, at least to any significant degree.

First things first: To be clear about these issues, we need to be specific about the type of “money” and “frictions” we’re discussing. There are, after all, many assets that are sometimes labeled “money”. First, there’s “base money”, the paper currency and electronic reserves issued by the Fed. Then there are all the different kinds of “money” created by banks, both traditional and shadow: transactions accounts, saving accounts, money market funds, repo, and so on. And if that’s not enough, the Treasury creates “money” too: Treasury Bills are so liquid and bereft of nominal risk that they’re essentially as good as bank money. (Often they’re held in money market funds, which add an additional layer of convenience, but Treasury does all the heavy lifting in creating short-term liquidity.)

When all these assets are given the blanket title of “money”, you can’t have any sensible discussion. The properties distinguishing $1000 in cash from $1000 in a money market fund are very different from the properties that, in turn, distinguish $1000 in the money market fund from $1000 in an S&P index fund. So let’s confine our discussion to the narrowest possible definition of money: currency and reserves issued by the Fed.

Does this restrict us to some kind of meaningless special case? Not at all. In normal times, the Fed implements monetary policy by adjusting the federal funds rate (and its expected path). This is the spread between the interest rate on currency (zero) and loans in the federal funds market—in addition to T-bills, commercial paper, and many other assets that end up with essentially the same rate. In other words, it’s the spread between base money and a much broader set of money-like instruments. If monetary frictions matter in understanding the impact of a certain shift in the federal funds rate, their effect must boil down to the difference between base money and “money” more broadly. Base money must be useful in addressing monetary frictions in a way that “money” in general is not, and this usefulness must have nontrivial macroeconomic consequences.

Does it? I find that exceedingly hard to believe. Paper currency, which in normal times comprises the vast majority of base money, simply isn’t that important to the macroeconomy—at least not at the current margin. (Quick question: can you think of any cash transactions you would stop making if interest rates went up to 4%, which makes you lose $4 a year for every $100 in your pocket? I didn’t think so. Even harder question: are there any transactions you would stop making, period, because of this cost—rather than simply shifting to some non-cash form of payment? Again, I didn’t think so.)

That leaves us with reserves. Before the crisis made reserves costless, required reserves amounted to about $40 billion—that’s 10% of the roughly $400 billion in checking accounts subject to the requirement. That’s compared to $150 trillion in total financial assets—or, if we want to avoid double-counting, $50 trillion in financial assets held by households. The impact of a 1% increase in the federal funds rate is to increase the implicit cost of holding those reserves by 1%—that’s $400 million. But to a first approximation, the 1% increase also changes the expected rate of return on all financial assets by 1%. If we use the household total of $50 trillion, that’s a $500 billion effect. The direct effect of interest rates is over a thousand times as large as the secondary effect of making checking accounts more expensive. In other words, the effect where monetary frictions come into play—because a certain kind of money is made more expensive—is vanishingly small in comparison to the standard effect in New Keynesian models. And if the Fed sets the federal funds rate using interest on reserves,  the former effect disappears entirely.

If this isn’t enough to convince you, consider the following: the role of monetary frictions is key to understanding what matters in monetary policy. Scott Sumner and other market monetarists doggedly insist that the true stance of monetary policy is determined by the expected future path of nominal variables, not just whatever the interest rate or monetary base happens to be today. I completely agree! We use different languages—I think that it’s better to talk about expectations in terms of interest rate rules, while they like to talk about nominal GDP and quantities of money—but we share the fundamental understanding that expectations are more important than current levels.

If you think that frictions are essential to understanding the effects of monetary policy, on the other hand, you have to think that the present matters much more. Why? The extent to which frictions are a tax on economic activity is determined by the current cost of holding money rather than less liquid assets. (After all, if money pays as much as all other assets for the next few weeks, you can hold all your wealth as money, and frictions won’t be much of a problem in the near-term!)  Yes, it’s possible that the future trajectory of monetary policy will affect your capital investment decisions—if extreme frictions in 5 years will make it difficult to sell your products, you won’t want to build the factory today—but the current cost of money (and therefore the current burden of frictions) has a vastly disproportionate influence on your actions.

This result pops out of virtually any model where there is no nominal rigidity and the only impact of money comes through frictions. For instance, if you parse Proposition 2 in Chari, Christiano, and Kehoe 1991, you’ll see that the proof for optimality of the Friedman Rule comes entirely through static considerations—setting R=1 so that a certain first-order condition is satisfied at each point in time. Admittedly, if you’re deviating from this “optimum” and setting R > 1, the future trajectory of policy matters, but in strange ways that don’t do a very good job of matching intuition: tight monetary policy in the future depresses capital investment today, yet all else equal it actually increases current consumption. (Tight future policy makes investing in capital less worthwhile, so you choose to consume more today instead.) I don’t think this is what market monetarists have in mind, to say the least.

Bottom line: if you say that expectations rather than current levels matter in monetary policy, you can’t think that monetary frictions are very important. This isn’t an argument against monetary frictions, of course—the model should drive policy conclusions, not the other way around—but it is a useful check for mental consistency.

To a large extent, I think this issue is confusing because under conventional policy, monetary frictions must matter to some degree—otherwise, the Fed couldn’t convince anyone to earn a lesser rate on money than other assets, and it wouldn’t be able to manage interest rates using open-market operations. But “some degree” can be extremely small in practice, and the details of how the Fed manages rates aren’t necessarily relevant to the effects of those rates, in much the same way that the mechanics of your gas pedal are peripheral to the consequences of driving quickly. Meanwhile, the Fed can implement policy when there are no frictions at all. Even when the market is saturated with money, it can set rates by adjusting interest on reserves. (In fact, it’s doing that right now.)

All in all, I see very little practical role for monetary frictions in understanding the impact of conventional monetary policy. The slight changes in cost of holding paper currency or maintaining a checking account simply don’t have serious macroeconomic consequences. The notion that we need a comprehensive model of these frictions to understand monetary policy is the kind of idea that’s plausible in theory but dead wrong in practice—just like the Friedman rule.

(Unconventional policy like QE is a different story, but let’s save that for another post.)