Recent news articles (, ) and
blog posts (Economists View,
Big Picture) have discussed Bernanke’s March 2 speech on globalization and inflation.
Certainly, there appears to be some evidence of a shift in the relationship often termed the Phillips curve. In the figures below, the change in inflation rates are plotted against the output gap (expressed in percentage terms), in two samples — 1968-95, and 1996-2006:
Figure 1: The change in the year-on-year CPI inflation rate on the output gap modeled as the deviation of log real GDP from HP-detrended GDP, lagged a year, over the 1968-95 period. Source: St. Louis Fed FRED II, and author’s calculations.
Figure 2: The change in the year-on-year CPI inflation rate on the output gap modeled as the deviation of log real GDP from HP-detrended GDP, lagged a year, over the 1996-2006 period. Source: St. Louis Fed FRED II, and author’s calculations.
Obviously, there looks like there is a change in this bivariate relationship. However, I think it would be useful to think formally about how globalization can affect inflation,
and the sensitivity of inflation to the output gap, as well as other shocks. In some respects, Bernanke’s speech revisits a topic discussed by SF Fed President Janet Yellen in her talk on “Monetary policy in a global environment”
which I discussed in this post last May.
Bernanke discusses the implications of both financial and real-side globalization; I’ll focus on the real side first:
Other than through its influence on financial markets, globalization may also have affected the operation
of monetary policy by changing the relative importance of the various factors that determine the domestic inflation rate.
As national markets become increasingly integrated and open, sellers of goods, services, and labor may face more
competition and have less market power than in the past. In particular, prices and wages may depend on economic conditions
abroad as well as on conditions in the local market. These linkages suggest that, at least in the short run, globalization
and trade may affect the course of domestic inflation.
International factors might affect domestic inflation through several related channels.
First, the expansion of trade may cause domestic inflation to depend to a greater extent on the prices of imported goods–not
only because imported goods enter the consumer basket or (in the case of imported intermediate goods) affect the costs of domestic production,
but because competition with imports affects the pricing power of domestic producers.
Second, competitive pressures engendered by globalization could affect the inflation process by increasing productivity
growth, thereby reducing costs, or by reducing markups. Third, to the extent that some prices are set in internationally
integrated markets, pressures on resource utilization in foreign economies could be relevant to domestic inflation.
What, then, is the evidence for the view that globalization is affecting the inflation process in the United States?
Of the various channels that have been suggested, probably the most intuitive is the idea that greater openness to trade
has increased the influence of import prices on domestic inflation. In the major industrial economies over the past decade
or so, import prices–particularly the prices of imported manufactured goods–have generally risen at a slower rate than
other consumer prices, slowing overall inflation. The slower growth in import prices reflects to some extent rapid
productivity gains in the production of manufactures, an important component of trade. Increased exports by low-cost
emerging-market economies have also helped keep down the prices of imports received by the United States and other
industrialized countries. Indeed, the share of U.S. non-oil imports coming from the emerging Asian economies has increased
from 27 percent to 34 percent over the past decade or so.
Overall, research indicates that trade with developing economies in particular has slowed the rate of growth of import
prices faced by industrialized countries, with estimates of the reduction ranging widely from 1/2 to 2 percentage points.
One study, for example, estimated that trade with China alone has reduced annual import price inflation in the
United States by about 1 percentage point over the period 1993-2002 (Kamin, Marazzi, and Schindler, 2006).
However, imported goods make up only part of what people consume, and so the effect on overall inflation is less than the
deceleration in the prices of imports alone. Typical estimates of the short-term effect on the overall inflation rate of
less-rapid increases in the prices of imports stemming from trade with China are in the neighborhood of 0.1 percent or
less per year–a discernable but certainly not a large effect.
Janet Yellen is somewhat more comprehensive, and detailed, in her discussion of how globalization affects the conduct of monetary
In discussing how globalization potentially affects the inflationary process, it is common to focus on a number of
distinct channels, and I will follow that approach here. However, I want to emphasize that, at least in some cases,
these channels represent partial effects that may have repercussions on other variables—such as the exchange rate — in a
fully specified model. Movements in these other variables may materially affect one’s views on the impacts of globalization.
However, I will defer that consideration until I turn to assessing the interpretation of the empirical results in the
The first channel is the most obvious one—the direct effect of the reductions in the prices of imported goods and services
that may be caused by globalization, and which are included in the indices of consumer prices that central banks commonly
Import prices also could have indirect impacts on inflation. One such indirect linkage might operate through the labor
market if nominal wage demands are influenced by the prices of imported consumer goods. The argument here is that a
decline in the price of imports raises the real reward to work, namely, the purchasing power of a given nominal wage.
Such real wage increases may raise labor supply. Alternatively stated, lower import prices could reduce workers’ demands
for nominal wage increases.
Another indirect channel reflects the possibility that lower import prices may restrain the prices charged by domestic
producers of competing products. Increased global competition, as the “new view” emphasizes, may have made the demand curve
facing American producers more elastic, resulting in larger feedbacks from lower import prices into core inflation.
The now standard practice of including import prices in the price-price or wage-price Phillips curve provides a way to
capture both direct and indirect linkages from import prices to domestic inflation.
In addition, this constraint on pricing ability could affect other parameters in Phillips curves.
This effect might operate in a couple of ways. First, when lower domestic unemployment leads to higher wage demands,
firms may not be able to pass through the higher costs, but must absorb them in their markups. As a result, a Phillips
curve that expresses inflation as a function of slack, lagged inflation, and other variables (the so-called price-price
Phillips curve) would become flatter—with a smaller response of inflation to measures of slack—as the “new view” emphasizes.
This result would hold even if the response of wage growth to slack were unchanged.
However, it is also possible that globalization could reduce the sensitivity of domestic wages to changes in domestic labor
market slack—in other words, it also could make the wage-price Phillips curve flatter. Suppose, for example,
that globalization has enhanced the opportunities for firms to substitute imports for domestic output.
This could occur in part because firms operating plants in several countries may be able to shift production from plants
in the U.S. to those in lower-cost countries. As such opportunities for substitution increase,
firms might become less willing to grant wage increases that would impair their cost competitiveness,
even in the face of tight domestic labor markets. Such substitution effectively increases the degree of competition
between domestic and foreign workers. In the limit—when such substitution in effect creates a single global labor market —
it could be that global, not domestic, labor market slack explains changes in U.S. wages and inflation.
A distinct but related possibility is that globalization may be undermining the bargaining power of U.S. workers,
making them more fearful of job loss, thus lowering wage demands and holding inflation down.
This might show up as a downward shift in the Phillips curve, similar to the impact of more rapid productivity growth in
the second half of the 1990s. However, globalization is but one of several structural shifts that may have deepened worker
insecurity, especially among less-skilled workers. These shifts include increased use of domestic outsourcing and
skill-biased technological changes that have decreased the demand for less-skilled workers and constrained their wages
in most sectors of the U.S. economy. Alternatively, globalization, coupled with technological change, may simultaneously
have raised the bargaining power of many skilled workers with opposite effects on the Phillips curve.
A final linkage from globalization to inflation worth noting pertains to productivity. Some have argued that increased
global competition has raised firms’ incentives to innovate and their ability to achieve productivity improvements in part
via foreign outsourcing of intermediate goods, IT services, and back-office functions. Productivity growth (or its change),
as we saw during the boom of the 1990s, may affect the dynamics of inflation. In essence, faster productivity growth matters to inflation, at least for a time,
because it holds down cost pressures. Stated differently, more rapid productivity improvements make it easier for firms to satisfy workers’ aspirations for
real wage gains. Faster productivity growth thus tends to lower inflation unless or until workers real wage aspirations rise to match the productivity gains.
I think it’s useful to distinguish between the level of inflation being driven by international factors (such as the rate of
import price inflation) and the strength of the linkages between domestic and international factors that might be affected
by globalization. To make this distinction more concrete, let’s discuss systematically these issues in the context of the a formal model, starting
with the Phillips Curve equation.
π t = π et + f(yt-1-y*t-1)+φZt
- π is inflation at time t, πet is expected inflation,
- (yt-1-y*t-1) is the output gap,
- Zt is the percentage point change in input prices, which could incorporate import prices (themselves a function of exchange rates).
f is the slope of the Phillips curve in π-(y-y*) space. Assuming adaptive inflation
such that the accelerationist model holds yields:
(2) π t = π t-1 + f(yt-1-y*t-1)+φZt
By the way, Equation (2) motivates the graphs in Figures 1 and 2. Pulling the lagged inflation term to the left-hand side yields the change in the inflation rate. What this makes clear is that the effect of the Z term is omitted from those figures. In the subsequent discussion, I will treat Z as the inflation rate of import prices (including oil).
Yellen’s first channel is summarized by Z, as well as φ. This direct effect is also what Bernanke stressed in his speech; he also focused
on the strength of the relationship of Z on π. Imports may also reduce the pricing power of domestic firms. She also notes that the slope of the
Phillips curve, f could also be reduced. Finally, higher productivity would show up (at least in this specification) in more rapid growth of
potential GDP, Y*.
All of these points would seem to mean a moderation of inflationary effects, that in principle should be measurable. However
estimating these relationships are difficult, given that the Phillips curve is not a structural relationship. Simple OLS
would not be appropriate to the extent that the output gap last period is related to the output gap this period, and this period’s output gap is affected by the inflation rate this period (i.e., serial correlation with a predetermined right hand side variable makes OLS provide biased estimates of the coefficients). These are well known problems, and econometricians have attempted to adddress them by use of instrumental variables approaches. As always, the estimates are only as plausible as the instruments.
Another way to see this point is to realize the Phillips curve is only one curve determining the equilibrium inflation and output gap. The other one is a composite of the central bank’s
reaction function and the IS curve. Once one realizes this, one can see how difficult it would be to disentangle the effects of
real globalization from changes in the Fed’s reaction function. To see this, consider this derivation based on Hall and Papell Macroeconomics texbook (2006), Chapter 16.
Let the Taylor rule be given by:
(3) r t = πt + β (yt-y*t) + δ(πt-π*t) + R*t
Where r t is the nominal interest rate. Now suppose the IS curve is given by:
(4) Rt – R*t = – σ (yt-y*t)
Where Rt is the real interest rate, R* is the equilibrium real rate, and -σ is the interest semi-elasticity
of the IS curve. Subtract π t and R*t from both sides of the Taylor rule; that makes the left hand side of the Taylor rule equal the left hand side of the IS curve.
(5) (yt-y*t) = – [δ/(β+σ)](πt-π*t)
Hall and Papell term this last equation the “Macroeconomic policy curve”, indicating that it incorporates both aggregate demand and both monetary and fiscal policy.
Substituting the Macro policy curve into the Phillips curve leads to:
(6) π t= πt-1 – [(fδ)/(β+σ)](πt-1-π*t-1)+φZ t
The last equation shows that the reduced form expression for current inflation depends upon lagged inflation and the lagged
inflation gap, as well as the current import shock.
This result arises because of the purely adaptive expectations process assumed for inflation. Interestingly, the autoregressive properties
of inflation depend upon the slope of the Phillips curve, the parameters of the Taylor rule, and the slope of the IS curve. For (δf)/(β+σ) < 1,
(and constant target inflation), inflation exhibits positive serial correlation. When (δf)/(β+σ) > 1, inflation exhibits negative serial correlation.
In a stochastic world, if expectations of inflation are unbiased — and prices are perfectly flexible —
then one would find inflation fully determined by the current policy stance (up to a function of expectations errors). Of course, the joint assumptions of rational expectations and fully flexible prices are untenable. However,
it could be — indeed it is likely that there is — a forward looking component to inflationary expectations. (This perspective is consistent with the early
New Keynesian models, and New Keyenesian Phillips curve models.) In that case, no simple solutions can be obtained. However, it would be straightforward to identify the important variables —
inflation in the current period would be driven by lagged inflation and lagged inflation gaps, current import price shocks, as well as current and expected values of these variables.
Then certainly real globalization — operating through φ and f — would have the effects consistent with intuition. But of course, this more realistic version of the model makes
it that much harder to disentangle and identify the evolution of the key parameters over time. Thus, while I find it entirely
plausible that f has become smaller so the response of inflation to slack is reduced (perhaps because of globalization), I’m not sure how certain we can be that
the link has weakened substantially. Also, since the Phillips curve is not a structural relationship, one could plausibly argue
that the reduced f arises in response to low inflation. Such a point was made by George Akerlof, Bill Dickens and George Perry in
2000 (Brookings Papers on Economic Activity) as discussed in his recent AEA Presidential Address.
That being said, one point is unambiguously clear. While posited downward pressure from declining import prices can push down domestic inflation, there is no guarantee that in
general import prices will act in such a benign fashion. In addition to the points made by Bernanke about China and other emerging
markets pushing upward commodity prices, more rapid appreciation of the Chinese Yuan, holding all else constant, would mean upward pressure on import prices. While Bernanke
accords a small impact to this channel, recent work by Marquez and Schindler (2006) leaves the question open — in their study, long run pass through into import prices is about half (obviously pass through into consumer prices would be much smaller).
One last (related) point. It is probably wrong to interpret Z as completely exogenous, although we have thus far assumed it is. Rather, Z depends upon the conduct of monetary policy; in that sense Z and π are jointly determined at least in part.