Zero matters

The invention of the number zero transformed mathematics and laid the foundations for modern science. Zero is the additive identity (any number plus zero equals itself). It separates the positive and negative numbers. For a celebration of zero, see here.

Zero matters in economics, too. 

Zero growth separates cyclical expansions and contractions. We need zeroes to measure the trillions of dollars of GDP, and even more zeroes to measure hyperinflations (during the record Hungarian inflation of 1945-46, the quantity of currency in circulation grew to a number with 27 zeroes). Most importantly, zero (or slightly less) marks the lower bound on nominal interest rates and a downward barrier for wage changes.

This commentary highlights the key role of zero in separating inflations from deflations. That is, the difference between positive, sustained increases in the general price level, and negative, sustained declines.

In some ways, inflations and deflations are symmetrical, but in other important ways, they are not.

When unexpected and persistent, both deflation and inflation can cause serious disruptions. An unanticipated change in the rate of price increase, whether it is up or down, undermines the plans of households, firms, and even governments. Regardless of where you start, when prices rise more quickly than expected, it erodes household savings, while the reverse increases the real burden of repaying debt.

In the United States, in the 1960s and 1970s, it took many years for savers to realize that inflation was rising. As a result, real returns to lending were negative. For that reason, as well as many others related to unanticipated inflation, overall economic performance suffered.

In the Great Depression, the severe and unexpected deflation increased the real burden of repaying the debt and lowered the value of collateral. The result was pervasive defaults and a reduced supply of credit. It was this experience that prompted Irving Fisher to coin the phrase debt deflation to describe an unstable spiral of falling prices and increased bankruptcies that depress the economy and intensify downward pressure on prices.

A recent BIS study found that the relationship between economic growth and deflations – with the important exception of Japan, most sustained episodes are before World War II – is far from tight. For example, in the United States in the late 19th century, persistent deflation was associated with solid economic growth, suggestive of a positive supply shock. In the BIS study, the link between deflation and economic growth is clearly negative only in the Depression period, when aggregate demand collapsed.

So, when they are unanticipated, inflation and deflation appear to be bad in quite similar ways. The one thing we will say is that this symmetry breaks down when the numbers are big. A ten percent Great-Depression-style deflation is much worse than a 10-percent inflation. Such a large deflation is so costly that, even when it is a remote probability, the tail risk will always remain a policy concern.

How about when inflations and deflations are anticipated? Are they the same?

When price changes are expected, households and firms are able to adjust to limit the negative effects. A firm that correctly expects annual inflation to average 2% over the next 20 years can reasonably judge whether a particular new long-term project warrants borrowing at a nominal rate of, say, 6%. Similarly, if the firm correctly anticipates annual deflation of 2% over the next 20 years, it can reach a similar conclusion about whether to borrow at a nominal rate of 2%. In either case, the expected real interest rate is 4%.

The point is that, so long as it is anticipated, planning can ensure that neither moderate inflation nor deflation wreak havoc. Indeed, if all prices and wages are flexible, movements in the general price level would not distort relative prices and wages even in the short run. As a result, resource allocation would be efficient.

But there are two qualifications; and they are big ones. First is the zero lower bound for nominal interest rates; and second, the fact that it is much more difficult for employers to cut wages than it is to increase them.

As we recently observed, central banks cannot lower nominal interest rates significantly and sustainably below zero. This makes inflation of 2% and deflation of 2% – anticipated or not – different. With inflation of 2%, to battle a recession, the central bank can lower the expected real interest rate (the nominal interest rate minus the expected rate of inflation) to -2% (that is “minus” 2%). By contrast, with deflation of 2%, the floor under the real rate is +2%. Since the real, rather than the nominal, interest rate is what influences employment and growth, monetary policymakers have more capacity to stabilize economic activity in a world of modest inflation than in a world of modest deflation.

That seems like a pretty big difference already, and one which suggests inflation targets of +2% may be too low. (See our recent post.) But, added to this is downward nominal wage rigidity. You might think wages are just another price, so they would go up and down with equal probability. If that were true, then, when everyone expects deflation, nominal wages could simply fall to avoid an unintended rise of the real (price-adjusted) wage.

Experience suggests otherwise. When average wage inflation is low, there are fewer wage cuts and more “zeroes” than you would expect if wages were completely flexible. This is what the histogram below displays: the frequency of wage changes deviates notably from a bell curve. There is an abundance of zeroes (and some skew of the distribution to the right). Importantly, the clustering at zero is more pronounced when the average wage gain was around 1½% in 2011 than when wages were rising at a rate over 4% per year in 2006.

United States: Distribution of 12-month log wage changes in 2006 and 2011

Source: reproduced from Mary C. Daly and Bart Hobijn, “Downward Wage Rigidities Bend the Phillips Curve,” FRB San Francisco Working Papers, January 2014. For an updated distribution, see the FRB San Francisco wage rigidity meter.

Source: reproduced from Mary C. Daly and Bart Hobijn, “Downward Wage Rigidities Bend the Phillips Curve,” FRB San Francisco Working Papers, January 2014. For an updated distribution, see the FRB San Francisco wage rigidity meter.

The next chart confirms the close relationship between the share of zero wage changes and the average wage change in the United States. Since 1997 (when continuous data became available), that correlation is a remarkable -0.86. Over the past 15-plus years, changes in average wage gains account for nearly three-fourths of the variation in the share of zeroes. And, for each percentage point reduction in wage inflation, the share of zero wage changes rises by about two percentage points.

United States: Percentage of Zero Wage Changes and the Average Rate of Wage Inflation

Sources: Nominal Wage Rigidity Data Release (FRBSF) and FRED.

Sources: Nominal Wage Rigidity Data Release (FRBSF) and FRED.

Downward wage rigidity makes mild deflations much more troublesome than mild inflations. Of course, there are offsetting factors. For example, worker productivity typically rises with experience, so employers need not cut nominal wages for these employees to lower unit labor costs a bit each year. And it is easier for firms to pay new employees less than to cut the pay of existing ones.

But our concerns about downward wage rigidity are reinforced by the experience of euro-area countries facing huge pressures to cut wages to compete with Germany. Even with record unemployment, institutional factors inhibit such wage cutting. In Spain, for example, since mid-2007, the unemployment rate has soared by 15 percentage points. Meanwhile, hourly wages have risen by nearly 14%.

Finally, there is the case of Japan, the only major economy to experience a sustained deflation since the Great Depression. Measured by the GDP deflator, prices in Japan declined by 18% from the peak in 1994 to the trough in 2013. Over the same 19 years, scheduled wages (excluding overtime) of regular workers rose by 2.5%. Importantly, however, Japan’s labor institutions build in an unusual degree of downward wage flexibility through the mechanism of semi-annual bonus payments, which fell by roughly one third after 1994 (see chart). Countries lacking such arrangements likely would have suffered much larger employment and output losses from a comparable deflation than Japan did.

Japan: Scheduled Wages, Bonuses and the GDP Deflator, 1990-2014 (1994=100)

Sources: Japan Basic Wage Structure Survey and Historical Statistics of Japan Table 19-49.

Sources: Japan Basic Wage Structure Survey and Historical Statistics of Japan Table 19-49.

What to conclude? Fears of deflation can surely be overdone: the economic impact of annual price changes of -0.1% would be difficult to distinguish from changes of +0.1%. It would take nearly 700 years for the price level to halve or to double at this pace! Yet, policymakers are warranted in taking the view that even mild deflations of 1% or 2% annually are meaningfully different from comparable inflations. Conventional monetary policy tools – adjusting nominal interest rates – are ill-suited to restoring price stability in the face of modest single-digit deflation. And the presence of downward wage rigidities makes things even worse.

The bottom line: zero matters.