Did globalization flatten the Phillips curve? U.S. consumer price inflation at the sectoral level

ABSTRACT

It is well-known that the slope of the U.S. Phillips curve has flattened in recent decades. In this article we examine the Phillips curve at the sectoral level to better understand why the slope has flattened. We decompose the core personal consumption expenditures (PCE) deflator into its sectoral parts and find, using a standard new Keynesian Phillips curve econometric specification, that the goods and services sectors contributed unequally to the change in slope. Our analysis is novel in its consideration of PCE inflation dynamics at a microsectoral level within the goods and services categories. The hypothesis that we think accounts for our findings is that globalization and the shift in U.S. output away from manufacturing production toward services reduced the impact of U.S. labor market conditions on the determination of goods prices and likely created a glut of workers seeking employment in the low-wage domestic service sector. The latter implies that workers in the service sector have a de minimis ability to bargain for higher wages, which ultimately manifests itself as a flat Phillips curve slope in the aggregate data. Our results suggest that the full-employment level of unemployment in the U.S. labor market is likely much lower than commonly estimated. Consequently, we argue that aggregate-demand-management policy in the postglobalization era has been persistently biased by an artificial perception of scarcity in the labor market. This bias has restrained wage growth for those working in the domestic service sector, ultimately resulting in widening earnings inequality.

KEYWORDS:

• Globalization
• inequality
• inflation
• monetary policy
• post Keynesian economics

JEL CLASSIFICATIONS:

• E12
• E31
• E52
• E60

Notes

¹The NAIRU is closely related to the so-called natural rate of unemployment, the rate at which labor supply and demand are in equilibrium (meaning that there is no excess supply of or demand for labor at prevailing wages), but is different in that equilibrium is defined by the NAIRU as a scenario where the economy-wide inflation rate is stable. Practically, estimates of the NAIRU are commonly used as (short-run) proxies for the natural rate, as aggregate labor supply and demand are always unobserved.

²At the time of this writing, the median consensus of the voting members of the Federal Open Market Committee (FOMC) is that the federal funds target rate, which currently rests at 0.5 percent, will be set within the range of 0.9–1.4 percent at the end of 2016. See the Economic Projections of Federal Reserve Board Members and Federal Reserve Bank Presidents, December 2015 (available at www.federalreserve.gov/monetarypolicy/files/fomcprojtabl20151216.pdf).

³As opposed to seeing inflation as a result of cost-push shocks such as union wage pressures, price increases by oligopolistic firms, and increases in the prices of commodities, which are all outside the purview of monetary policy.

⁴The persistently loose stance of monetary policy in the 1970s was allegedly due to policy being crafted on the basis of poor estimates of potential gross domestic product [GDP] (corresponding to estimates of the natural rate of unemployment that were too low), thereby leading to a chronically overheated economy. See Orphanides (2003).

⁵See Bernanke (2004). Interestingly, Bernanke in his speech also mentions but does not entertain various other structural causes that may have stabilized inflation, including the shift of production in the United States away from manufacturing toward services. This is the structural change that we put the most weight on, based on our reading of the sectoral inflation data (see the next section).

⁶Truman Bewley was among the first to elucidate this dynamic. He famously interviewed hundreds of managers and job recruiters to better understand why wages do not fall as much as they should during recessions—that is, why wages are “sticky” relative to how the prices of other commodities adjust in competitive markets. Bewley's main conclusion is not only that nominal wage cuts negatively impact worker morale, but that managers also believe that the reduced morale will translate into lower worker productivity, so the cuts are also resisted from the perspective of the employer. See Bewley (2002).

⁷The Fed made the switch to targeting PCE inflation rather than CPI inflation in the 2000 Humphrey–Hawkins report. The reasoning behind the switch is that the PCE chain-type index is constructed from a formula that reflects the changing composition of consumer spending in the economy and thereby avoids the upward bias associated with the fixed-weight nature of the CPI. See Monetary Policy Report to the Congress (2000).

⁸To the best of our knowledge, our analysis is original in its attempt to understand why the Phillips curve changed using microsectoral PCE inflation data.

⁹A full list of the sectors included in our constructed core PCE deflator is shown in Figure A1 in the Appendix.

¹⁰Each sector's weight is calculated as the amount of spending (measured in nominal dollar terms) occurring in the sector as a share of total spending for all 162 sectors. We obtain nominal spending data for our considered sectors from Table 2.4.5U in the National Income and Products Accounts.

¹¹We do not consider data prior to 1986, because we feel that the Fed's regime shift toward anchoring inflation expectations would likely bias our analysis. By 1986, inflation expectations were already well-anchored, dropping noticeably from the late 1970s, a drop that persisted into the early 1980s. Long-lasting disruptions from the Fed's regime shift were over by 1986.

¹²We are not the first to entertain this hypothesis. James Galbraith, in his 2008 book The Predator State (chap. 4), suggested that the rise of Chinese trade in manufactured goods may have structurally altered consumer price determination in the United States by placing downward pressure on global labor costs. Janet Yellen, when she was president and chief executive officer of the Federal Reserve Bank of San Francisco, made a similar argument, even going as far as to say that the flattening of the slope of the Phillips curve may be a result of globalization reducing the bargaining power of American workers who have become more fearful of job loss (Yellen, 2006).

¹³As is common in the existing literature, we assume that expectations of future inflation can be modeled as a series of lags of inflation (we use two).

¹⁴For comparison, we show in Figure A2 in the Appendix the results of the same regression except with the BEA's headline core PCE deflator rather than with the one we constructed using sectoral data.

¹⁵For an illustrative example of how p-values get distorted when the true value of an estimator is very close to zero, see the video “Dance p 3 Mar09” on YouTube (www.youtube.com/watch?v=ez4DgdurRPg).

¹⁶Again for comparison, we show the analog of Figure 10 for the BEA's headline core PCE deflator in Figure A3 in the Appendix.

¹⁷The Fed expects the unemployment rate to fall to between 4.6 percent and 4.8 percent by the end of 2016, and core PCE inflation, which currently is running at 1.3 percent, to rise to between 1.5 percent and 1.7 percent. See, again, the FOMC's December 2015 economic projections (www.federalreserve.gov/monetarypolicy/files/fomcprojtabl20151216.pdf).

¹⁸Figure A1 in the Appendix, which, as mentioned, lists the 162 non-food-or-energy sectors included in our analysis, also denotes in the second column whether each sector falls into the category of goods (denoted by G) or services (denoted by S).

¹⁹Although this result does not allow us to reject at the 5 percent level the null hypothesis that the slope of the Phillips curve is zero. This finding is different from that of the New York Fed's analysis, which finds that while the estimated slope of the Phillips curve has declined when considering aggregate core services inflation data (from the CPI), the slope is statistically different from zero at the 5 percent level—we replicate and show the relevant confidence intervals using the core services CPI in Figure A4 in the Appendix.

²⁰In Figure A5 and Figure A6 in the Appendix, we show more details regarding these sectoral regressions, including further breaking down the services supersector into services excluding housing and housing services. In these more detailed tables, we show which specific sectors were isolated as having a Phillips curve slope whose 95 percent confidence interval is entirely negative as well as the estimated slopes and their corresponding sector weights.

²¹These weighted supersector slopes are calculated as the sum of the weighted slopes within each supersector, where each individual weight denotes the average amount of spending (measured in nominal dollar terms) occurring in each sector as a share of total spending for all 162 sectors over the time period under consideration.

²²Again, see Figure A5 and Figure A6 in the Appendix for a list of the specific sectors driving the results in Figure 13 and Figure 14.

²³Of our 162 examined sectors, we classify eight as housing. They are: Group Housing, Tenant-Occupied Stationary Homes, Tenant-Occupied Mobile Homes, Tenant Landlord Durables, Housing at Schools, Owner-Occupied Stationary Homes, Owner-Occupied Mobile Homes, and the Rental Value of Farm Dwellings.

²⁴In Figure A7 in the Appendix, we reinforce this point by comparing the estimated Phillips curve slope from a fifteen-year rolling regression using our constructed core goods PCE deflator in Equation (1) against the number of manufacturing workers employed in the United States. This chart suggests that the shift of labor-intensive production out of the United States served to break the classical relationship between declining labor market slack and higher domestic goods inflation.

²⁵As mentioned, the CBO produces an estimate of the NAIRU that helps guide policymaking. In addition, the Fed frequently publishes its view of what the “longer-run unemployment rate” is in its Summary of Economic Projections, which is released following selected FOMC policy meetings each year.

²⁶We should stress here that the 5.5 percent level is our sense of the historical norm. More recently, the Fed has actually been systematically revising down its estimate of the “longer-run unemployment rate.” At the time of this writing, the consensus view of the FOMC is that the longer-run unemployment rate is within the range of 4.8–5.0 percent. See the Economic Projections of Federal Reserve Board Members and Federal Reserve Bank Presidents, December 2015 (www.federalreserve.gov/monetarypolicy/files/fomcprojtabl20150318.pdf).

²⁷Tangentially, an abundance of workers seeking employment in the low-wage service sector may not be the only force holding down wage growth. The service sector is also qualitatively different than the manufacturing sector in that collective bargaining—and thereby the wage gains that come with it—is less common in the former. This is because workers in the service sector are more spread out both geographically and across smaller company franchises, as opposed to manufacturing workers who were traditionally concentrated in large factories in specific geographical hubs (think Detroit for automobile manufacturing or Pittsburgh for steel production). These qualitative differences are usually discussed in terms of economies of scale when it comes to unionizing: if, for example, a union can dispatch 15 organizers to a big factory, they may be able to walk away with 1,000–2,000 new members, whereas if organizers are sent to a fast food restaurant, they may be looking at only 20–30 new members.

²⁸In the current cycle, it has become common to examine broader measures of unemployment—such as the U6 unemployment rate, which includes not only those unemployed and searching for work but also those who are marginally attached to the labor force (i.e., those not in the labor force but who want and are available for work) and those working part-time for economic reasons (meaning they would prefer to work full-time if the opportunity were available)— in order to gauge the amount of slack in the labor market. Indeed, the Fed's staff members routinely aggregate a plethora of alternative labor market indicators (19 of them) into a so-called labor market conditions index (LMCI) using a dynamic factor model. Janet Yellen and other FOMC members have referred in various public testimonies to the Fed's LMCI as a useful measure to gauge slack. See Chung et al. (2014).

²⁹This point is made analytically by Dean Baker and Jared Bernstein in their book Getting Back to Full Employment. Baker and Bernstein pay specific attention to the late 1990s when the Fed did not tighten monetary policy as the headline unemployment rate approached conventional estimates of the NAIRU. As it happened, the unemployment rate proceeded to fall two percentage points below the estimated NAIRU (to around 4 percent) without generating destabilizing upward pressure on consumer prices. This period of late 1990s tightness in the labor market was the only period during the Great Moderation when wage growth at the bottom of the income distribution advanced in line with growth at the top (see Figure A8 in the Appendix).

³⁰For an in-depth discussion about how the Fed's full employment mandate could/should be reinterpreted to better promote shared prosperity, see Palley (2015).

Additional information

Notes on contributors

Joe Seydl

Joe Seydl and Malcolm Spittler are economists employed in the private sector.

Malcolm Spittler

Joe Seydl and Malcolm Spittler are economists employed in the private sector.