A Simple Test of the Summers-Blanchard Inflation Hypothesis

Economics proceeds one funeral at a time

Old ways of thinking are almost impossible to kill. They usually only go away when the heads which hold the old ideas end up in the grave. As the old saying goes, economics proceeds one funeral at a time. These mental rigidities are most obvious in the implicit theories of the inflation process in the heads of the doyens. Larry Summers, Olivier Blanchard and Steven Rattner, usually all inflation doves and card-carrying Democrats, have wagered that the $1.9 trillion spending bill just passed by Congress will trigger inflation. In turn, the revival of inflation will force the Fed to hike, which would undo much of the macroeconomic stimulus of the fiscal expansion. Markets in turn are buying into the story to some extent. The prospect of the removal of monetary accommodation has already triggered a selloff in bond markets:

The markets are not completely wrong — these economists are. It is possible that the Fed could again revert to the harebrained idea of hiking in anticipation of inflation, and this policy risk may warrant more conservative bond pricing. But the notion that inflation is now around the corner is completely mistaken.

The economists, beginning with Summers, reckon that the scale of the fiscal stimulus is out of proportion to the scale of the output gap. And therefore, the US economy will overheat as spending by the recipients of the Federal largesse generates demand in excess of capacity. More money in people’s pockets means more demand for goods and services, whose supply is presumably fixed in the short term, which would therefore generate inflation.

This is the Standard Model of inflation. The relationship between domestic slack, usually captured by the output gap, and inflation is given by the Phillips curve. It used to work reliably in the postwar period. In the 1970s, the old Phillips curve had to be abandoned as inflation expectations were destabilized. What replaced it was the expectations-augmented Phillips curve. That seemed to work until hard currency-issuing central banks, led by the Fed, succeeded in taming inflation and anchoring inflation expectations on target. This process was complete by the mid-1990s, ushering in what was called ‘The Great Moderation’ — a period of stable inflation and low volatility in output growth. The same period saw globalization and trade liberalization.

What then happened was that the Phillips curve relationship between domestic slack and inflation completely broke down. Three global developments drove this process. First, the addition of hundreds of millions of workers to the global labor pool, above all in China, dramatically reduced the bargaining power of labor globally, with the result that AE and, in particular, US wage growth stalled even as output per hour kept rising. This put a lid on wage pressures as a driver of inflation.

The second development was the oligopolization of industrial sectors organized by Wall Street in response to the loss of dynamism that became evident in the late-1970s. In order to deter entry, the oligopolists had to build massive moats in the form of excess capacity. This parlor trick restored Wall Street profits, and upper class investment incomes, without restoring underlying dynamism. It also led to a secular decline in capacity utilization across industrial sectors. The dramatic increase in excess capacity meant that even large demand shocks could be absorbed by the supply side without generating inflationary pressures.

The third development was the ‘second unbundling of global production’ [Baldwin]. This was, to a great extent, a regional unbundling — dictated by flying distance from the ‘headquarter economy’. American firms expanded operations into Canada and Mexico, Germans firms across Europe and especially into Eastern Europe, and Japanese and Taiwanese firms across the Pacific Rim. But it was also a global process related to the rise of the Chinese behemoth. In short order, China became practically everyone’s largest trading partner as global industrial firms located low and medium value-added processes to mainland China in order to exploit cheap but disciplined and hardworking Chinese industrial workers. This unbundling of global production meant that what came to drive inflation was the capacity of global value chains to keep up with domestic demand, not the slack in the domestic industrial sector.

All these global factors were implicated in the breakdown of the Phillips curve and the secular decline in inflation in the center countries. A number of economists began to study the new inflation process. The old hypothesis that domestic slack drove inflation has been increasingly abandoned in favor of the hypothesis that global slack drives inflation. These two hypotheses have been tested by a number of authors including Borio, Ciccarelli, Mojon, Ragot and others. The main challenge is to construct a good proxy of global slack. Most authors use the weighted average of output gaps — weighted either by GDP or trade intensity. The problem is that neither of these weights capture the structure of global value chains.

A satisfactory measure of slack in the global production system would use weights derived from the datasets now available on global value chains. In other words, the new model of the inflation process would pay attention to the degree of slack in the global value chains of firms supplying the domestic economy. This is a big research project that we shall leave for future work. For now, we shall look at a particularly simple proxy of global slack — the common component of inflation in the advanced economies. Specifically, we define the global factor as the unweighted mean of AE inflation.

The motivation for this proxy is straightforward. Advanced economies are essentially supplied by global industrial firms whose value chains are spread out over the world. Moreover, in the advanced economies, inflation expectations have been anchored at a low level since the mid-1990s. This means that AE inflation rates are not confounded by expectations. Furthermore, trade across AE borders flows freely so that they are more exposed to global production than the more autarchic developing nations. The common component of their inflation rates thus contains information on the degree of slack in the global production system as a whole. We therefore expect mean AE inflation to track global slack closely.

Already in 2010, Ciccarelli and Mojon had shown that the global factor accounts for 70 percent of the variation in AE inflation rates. They show that a particularly simple error correction model accounts for almost all the diachronic and synchronic variation: domestic inflation in the advanced economies is a linear function of global inflation plus an idiosyncratic error term centered at zero.

Following these observations, our model of the inflation process says that domestic inflation in the advanced economies is a linear function of global slack, proxied by mean AE inflation. The standard model says that domestic inflation in the advanced economies is a linear function of the domestic output gap. In what follows, we test these hypotheses together. We obtain quarterly CPI and nominal GDP data for 20 advanced economies (the five Anglo-Saxon countries, 14 from the continent and Japan) from the IMF’s International Financial Statistics dataset. We construct output gaps using the standard Hodrick–Prescott filter (with the parameter set at 1600, as is standard for quarterly data) applied to real GDP. We have data from 1994Q4-2020Q2. We drop the last two observations because of the extreme shock in 2020. Our final dataset consists of 2000 observations for 20 countries and 100 quarters. We regress country-specific inflation on the country’s output gap (Domestic Slack), mean AE inflation (Global Factor), and lagged country-specific inflation (Persistence).

Figure 1 displays the global factor along with US qtr-on-qtr inflation. The series look highly correlated and they are. The correlation between the two series is r = 0.77, P < 0.0001.

Figure 2 displays the AE output gap together with the US output gap. These two series are also highly correlated (r = 0.75, P < 0.0001).

The next figure presents our main result. It shows the t-statistics for the main regression specification. The slope coefficient for domestic slack is statistically indistinguishable from zero in 17 out of 20 advanced economies. Domestic slack is significant but bears the wrong sign for Finland (t = -2.62) and Portugal (t = -2.11). Only for Sweden is it significant and bears the right sign (t = 2.27). Meanwhile, the global factor is significant and bears the right sign in all 20 country-specific regressions.

The following table displays the same t-statistics in detail. There is simply no comparison between the global factor model and the standard model of the inflation process. The mean t-statistics are 10.1 for the global factor, -0.9 for the AR(1)-persistence factor, and -0.3 for domestic slack. The evidence couldn’t be more one-sided if it tried.

Displaying the results of 20 different regressions is tricky. But we are largely interested in the US. The next figure displays scatterplots for US inflation against domestic slack and the global factor. US inflation simply does not respond to the domestic output gap (t = -0.6). Instead, it is governed by the global factor (t = 11.5).

So all this inflation talk by supposedly hard-nosed economists is completely incongruent with empirical reality. It is based on an implicit theory of the inflation process that has not worked for twenty years. Where there should be a new theory of inflation in their heads there is nothing more than a mental rigidity. Nor is this mental rigidity restricted to a small handful of old economists. It is shared by perhaps a majority of macroeconomists whose mental frames of reference were fixed in the 1970s. The world has changed around them but their habits of mind have proven harder to kick. And these intellectual rigidities have real world implications. They are the reason why the Fed kept hiking in anticipation of inflation that never arrived. As I mentioned earlier, the real risk is a model risk or policy risk — the Fed could revert back to the bad old days of hiking in anticipation of inflation based on a broken model of the inflation process. That’s the risk faced by market actors and everyday people suffering the rule of these technocrats.

Thankfully, the Democrats have ignored the technocrats for now. But the risk remains that incompetent technocrats will again kill the party before it gets started.