Money money money money, money… Don’t let, don’t let, don’t let money fool you
For the Love of Money, The O’Jays, 1973
“[W]ith inflation running persistently below 2 percent, our policy will aim to achieve inflation outcomes that keep inflation expectations well anchored at our 2 percent longer-run goal.” Federal Reserve Vice-Chair Richard Clarida, January 8, 2021.
Over the past quarter century, disagreements about monetary policy have been narrow. A broad consensus developed that central banks should be independent, with a primary focus on price stability. Furthermore, nearly everyone agrees that policymakers should explain their actions and strategy, publishing summaries of their deliberations and standing for questions at frequent intervals.
For more than a generation, this inflation-targeting framework has delivered impressive results. From 1995 to 2007, U.S. inflation averaged 2.1% (as measured by the Federal Reserve’s preferred index). Since 2008, average inflation dropped to only 1.5%, but expectations have fluctuated in a narrow range: for example, the market-based five-year, five-year forward (CPI) inflation expectation rarely dipped below 1.5% and never exceeded 3%.
The pandemic brought with it many dramatic changes. Fiscal and monetary policy mobilized, responding swiftly to the economic plunge with a combination of extraordinary debt-financed expenditure and balance sheet expansion. U.S. government debt held by the public already is up by 20 percent of GDP since the end of 2019—or more than $4 trillion—and now exceeds $21 trillion. Over the same period, the Federal Reserve’s balance sheet expanded by 15 percent of GDP and now stands at a phenomenal $7.4 trillion.
As a matter of accounting and arithmetic, these actions have had a profound impact on the balance sheets of banks and households, spurring dramatic growth in traditional monetary aggregates. From the end of February to the end of May 2020, broad money (M2) grew from $15.5 trillion to $17.9 trillion—a 16% jump in just three months.
Won’t this record gain in M2 be highly inflationary? We doubt it, and in this post we explain why. At the same time, we highlight the chronic uncertainty that plagues inflation. In our view, the difficulty in forecasting inflation precisely makes it important that the Fed routinely communicate how it will react to inflation surprises—even when, as now, policymakers wish to promote extremely accommodative financial conditions. The better the private sector understands the Fed’s policy reaction function, the more effective policy will be in stabilizing both inflation and growth.
Before we get started, we should say a few words about the mechanism behind last year’s surge in the stock of broad money. Five factors are at play. First, demand for currency rose by 15%, more than double the pace of the previous decade. The 2020 increase was $275 billion. Second, as a precaution early in the pandemic, businesses drew down lines of credit by something in the range of $600 billion. When this happens, the lending bank credits the borrowing firm’s deposit account, which is a part of M2. Third, spurred by fiscal transfers and diminished spending opportunities, household savings skyrocketed, rising by more than $1 trillion. Fourth, the Federal Reserve’s bond purchase programs mechanically boosted both commercial bank reserves (an asset) and (at least initially) customer deposits (a liability) of those who sold securities to the Fed. Finally, with interest rates so close to zero, firms and households faced virtually no opportunity cost of keeping funds in a bank deposit.
Will the 2020 M2 spike lead to substantially higher inflation? As we discuss in an earlier post, the simplest version of monetarism states that controlling money growth is both necessary and sufficient to control inflation. So, if we see money growth rise, then inflation must be on the horizon. The following chart is Exhibit A in the case for this view. Using data for 90 countries on average annual inflation and money growth over nearly four decades, we can see that there are no examples of countries with either sustained rapid money growth and low average inflation or the converse. And, if we were to assume that the 2020 M2 growth rate in the United States were the new average—rather than a temporary spike—then this picture would lead us to anticipate U.S. inflation beyond anything we have seen since the end of World War I.
Average Annual Consumer Price inflation and Broad Money Growth, 1980 to 2017
Source: IMF World Economic Outlook, World Bank, and authors’ calculations.
However, this conclusion is profoundly misleading. First, no one seriously believes that U.S. monetary aggregates will continue to grow rapidly and unabated for years. Consistent with the relative stability of inflation expectations, there seems to be agreement that the 2020 jump is a one-off shock (see the chart here). Second, at low levels of inflation, the short-run link between money growth and inflation is loose, at best. Our recent post shows how in recent years, fluctuations in the two have been pretty much independent. Third, and related to the previous point, low nominal interest rates favor holdings of deposits included in M2.
That said, we thought it would be interesting to explore the implications of the 2020 monetary developments for forecasts of inflation over the next few years. To do it, we employ an incredibly naïve model in which inflation depends on its own history and the history of money growth. Technically, we ran a set of regressions of the four-quarter percentage change in the price index of personal consumption expenditures (PCE) on eight lags of itself and eight lags of the four-quarter percentage change in M2. We did this over a long sample from 1962 to 2019, as well as a shorter one starting in 1995. Then, assuming from 2021 onward that M2 growth returns to its pre-2020 average of 6 percent per year, we construct dynamic forecasts of inflation for the five years through 2025.
Before getting to the forecasts, we note a few things about this simple, single equation model. First, as we emphasize in our 2017 study of inflation for the U.S. Monetary Policy Forum, virtually all the explanatory power in forecasting inflation comes from inflation’s own history. Second, in the full sample, the long-run impact of an increase in money growth is large: a persistent one-percentage-point increase in money growth yields an 0.9 percentage-point-increase in inflation (asymptotically). However, in the shorter sample, a one-percentage-point increase in money growth perversely anticipates a very small decline in inflation (about 0.1 percentage point). Put differently, the 2020 M2 spike signals little about inflation unless it somehow restores the U.S. inflation process to its pre-1995 pattern.
As an aside, some people suggest that the recent surge in public debt itself will be inflationary. This would surely be true if there were fiscal dominance—in which fiscal policymakers control the supply of money—so it likely explains some of the high inflation observed in the first chart. While we do not expect the U.S. government to usurp monetary control from the Fed, we expand our simple M2 growth model of inflation to see if changes in the ratio of federal government debt to GDP drive U.S. inflation. Starting in 1971, when quarterly data become available, we find that lags in debt growth are uninformative about future inflation, so we exclude them from our further analysis.
Turning to the results, in the following two charts we plot the point estimate (the red line) and a 90% confidence interval (the black dashed lines) for forecasts of inflation over the 2021 to 2025 period using the full sample from 1962 and the recent sample since 1995. Starting with the full sample in the top panel, the results are largely driven by the 1960s and 1970s, leading to an inflation peak in early 2023 of slightly less than 6 percent. Even by the end of 2025, the forecast based on the long sample falls only to 3 percent. Importantly, however, uncertainty surrounding these point estimates is large: at the peak two years from now, the 90% band ranges from 2 to 9 percent!
Looking at the projections based on data starting in 1995, the central forecast is completely different. While inflation peaks in early 2023, the cap is 3 percent. Moreover, inflation sinks below the 2 percent objective in later years. Importantly, the 90% confidence band is still wide, ranging from 1.5 to 4.7 percent. Moreover, these confidence bands are quite similar in size to those in the Federal Open Market Committee’s (FOMC) newly expanded, quarterly Summary of Economic Projections (SEP; see Figure 4.C. and Table 2 in the December 2020 edition).
PCE inflation forecast using the full sample from 1962 to 2019 (Quarterly), 2021-25
Source: FRED and authors’ calculations using Eviews.
PCE inflation forecast using the recent sample from 1995 to 2019 (Quarterly), 2021-25
Source: FRED and authors’ calculations using Eviews.
We draw two inferences from this simple analysis, one about the level of future inflation and one about forecast uncertainty. On the level, in our view, the unprecedented spring 2020 burst in money growth does not herald a return even to high single-digit inflation, much less the double-digit price increases of the mid-1970s and early 1980s. Experience shows that inflation is driven by a highly persistent process, with short- and medium-term fluctuations only very loosely related to the time path of monetary aggregates. Of course, given the wide forecast error bands, it could still rise significantly. But if it does, we would expect to first see an increase in inflation expectations—beyond the pattern of recent decades—driven by anticipation of strong and sustained economic growth. The latter also ought to drive up longer-term real interest rates—something that clearly is not happening (see chart).
As for uncertainty about future inflation, the unmistakable message is that it is chronically high—a point that the SEP makes clear. To put some numbers on this, while the FOMC’s longer-run inflation goal remains 2%, their uncertainty is consistent with a 90% confidence band even one year out that is a full 3 percentage points wide. That is, in the view of Fed policymakers, inflation could be as high as 3.5% or as low as 0.5%. As is the case in our simplistic analysis above, these measures of uncertainty come from looking at the performance of models over the past several decades. From our vantage point today, still in the midst of the COVID pandemic, this past experience looks strikingly tranquil. As a result, current uncertainty about future inflation likely is even higher than the pre-2020 history suggests.
To conclude, we return to our 2019 analysis of the Federal Reserve’s monetary policy communications framework (see here). There, we emphasize the importance of communicating in advance how policy would react to changing conditions. As Vice Chair Clarida states in the quote at the outset of this post, the issue is whether inflation persistently deviates from the longer-run goal, not why. Furthermore, what is important for monetary policymakers is that everyone understand how they plan to react to deviations. While we appreciate and support the need for extraordinary accommodation in the current circumstance, it is no less important for the FOMC be clear about how they would react to inflation across what is a broad range of plausible outcomes.
What will they do if inflation projections for 2023, currently equal to 2%, were to fall to 1% or rise to 3%? What is important for safeguarding stability is that everyone correctly anticipate policy plans when inflation or inflation expectations diverge from the FOMC’s longer-term goal. So, while we are untroubled by the 2020 spike in M2, to make the Fed’s average inflation-targeting regime as effective as possible, it is essential that we know how future inflation developments will prompt policymakers to act.