This paper is based on a long-term research program with Rachel Kranton on the implications of identity

28

inflation will be ever decreasing.  The natural rate then is the only sustainable level of

unemployment without accelerating or decelerating inflation.

Acceptance of Natural Rate Theory

As revealed by textbook presentations, most macroeconomists do not just view natural

rate theory as a useful null hypothesis.  They also view it as a description of reality.  Because of

this acceptance and its implications for macroeconomic policy, I shall discuss this neutrality at

considerably greater length than any of the other four.  But the basis for questioning this

neutrality will be remarkably similar to the other four; it will concern whether the assumed

behavior is based on a realistic view of preferences.  In this case the preferences concern the

wages and prices that employers and consumers respectively think should or should not be set.

Economists accept natural theory for theoretical and empirical reasons.  Theoretically,

they view its assumptions as realistic.  A standard criterion for an economic model is that

participants in the economy care only about real outcomes.  That is the fundamental assumption

of natural rate theory.  Also, unlike most other neutrality results, natural rate theory is relatively

insensitive to the to deviations from the perfect information competitive model.  As long as these

“frictions,” such as imperfect information or transaction costs, can be expressed solely in real

terms, the neutrality result of natural rate theory will be robust.  

Natural rate theory gained acceptance, not just for theoretical reasons, but for empirical

reasons as well.  The original Phillips curve showed a close fit between the rate of change of

nominal wages and inverse of the  unemployment rate for 97 years of British data, between 1861

and 1957.  However, in the United States in the late 1960's and early 1970's such a simple

---

36

See, for example, Gordon (1977, Table 3, p. 260, lines 6 and 7).

37

Given the importance of such findings, it is remarkable that their robustness to specifications of time

period, data, and exact specification of the Phillips Curve have never been subjected to tough tests—even though

everything else about the Phillips Curve, including the natural rate of unemployment itself is considered to be

estimated with great imprecision. Akerlof, Dickens and Perry (2000) show a range of estimates for both wage and

price equations with many different specifications.  These estimates, particularly when made for periods of low

inflation, show considerable variation in the sum of the coefficients on lagged inflation, dependent on the

specification.  Another bit of evidence that suggests such estimates will be sensitive to specification comes from the

high standard errors on the natural rate itself (Staiger, Stock and Watson (1997)); it would be surprising that the sum

of lagged coefficients could be estimated precisely if another component of the Phillips Curve, the natural rate could

be estimated only with very low precision.  Gordon’s own estimates show very different values for this sum of

coefficients.  Of course, there is a theoretical reason why estimates of such a sum should not be robust.  With

rational expectations, rather than a simple mechanical theory of formation of inflationary expectations, Sargent

(1971) shows that there is no theoretical reason that they should sum to one.   

29

inverse relation between changes in nominal wages and unemployment broke down as both price

and wage inflation rose, along with the unemployment rate.  Natural rate theory offered an

explanation for this occurrence: it explained the rise in inflation by the large oil supply shock

and also an increase in inflationary expectations, both of which shifted the Phillips Curve

outward; it explained the rise in unemployment by a decline in demand.  

Furthermore, new estimates of Phillips Curves seemed to show that the theory closely fit

the data.  If inflationary expectations are formed as a simple lag of past inflation, estimates of

Phillips Curves should find that the sum of coefficients on past inflation should sum to one. 

Many Phillips Curve estimates fail to reject that this sum is equal to one.

36, 37

 

The textbooks thus typically present natural rate theory as a “just-so” story.  It runs as

follows.  The previous Keynesian economists had posited a Phillips Curve without a dependence

on inflationary expectations.  Friedman (1968)  and Phelps (1968) perceived that such a theory

could not result from models where the participants in the economy are concerned only with real

variables.  They modified the relationship so that wage and price equations would be affected

one-for-one by inflationary expectations.  Such judicious use of economic theory explained the

---

38

These distitbutions have accumulations at zero, and they are also asymmetric: there are more wage

changes above zero than below zero.  This suggests that the accumulations at zero do not just occur because there is

a menu cost for changing wages.  

39

The following studies have all found significant signs of nominal wage rigidity: Bewley (1999), Card and

Hyslop (1997), Kahn (1997), Lebow, Saks and Wilson (1999), and Altonji and Devereux (1999) for the United

States, by Fortin (1996) for Canada, by Cassino (1995) and Chapple (1996) for New Zealand, by Dwyer and Leong

(2000) for Australia, by Castellanos et al (2003) for Mexico, by Kuroda and Yamamoto (2003a, 2003b, 2003c) and

Kimura and Ueda (2001) for Japan, by Fehr and Goette (2003) for Switzerland, by Bauer et al. (2003) and Knoppik

and Beissinger (2003) for Germany, by Nickell and Quintini (2001) for the United Kingdom, and by Agell and

Lundborg (2003) for Sweden.  

40

See, for example, O’Brien (1989) and Hanes (2000).

30

otherwise-mysterious finding of the simultaneous increases in inflation and unemployment of the

late 1960's/early 1970's. 

Facts about Wage Behavior Suggesting Nominal Considerations

But a considerable body of findings run contrary to natural rate theory.  They indicate

that nominal considerations do affect decisions about wages and prices.  I shall begin with a

discussion of wages, and then later turn to prices.  There are at least six observations that

nominal considerations affect wage setting. 

First, money wages are downwardly rigid.  Such wage behavior can be easily perceived

statistically since wage-change distributions are truncated at zero.

38

  Careful studies have

documented wage stickiness in the United States, Canada, New Zealand, Australia, Mexico,

Japan, Switzerland, Germany, and the United Kingdom.

39,40

  There seems to be no way to

account for such nominal wage rigidity with the basic assumptions underlying natural rate

theory: that participants in the economy only care about real prices and real wages.

Second, Truman Bewley (1999) has examined money wage rigidity, but from a very

different perspective.  His open-ended interviews sought to elicit the reasons why employers did

---