In an IMF blog post, Maurice Obstfeld, Gian Maria Milesi-Ferretti, and Rabah Arezki offer a solution to the "puzzle" of the weak positive macroeconomic response to low oil prices. Specifically, they posit a sharp rise in real interest rates due to falling inflation expectations is the culprit:
Even though oil is a less important production input than it was three decades ago, that reasoning should work in reverse when oil prices fall, leading to lower production costs, more hiring, and reduced inflation. But this channel causes a problem when central banks cannot lower interest rates. Because the policy interest rate cannot fall further, the decline in inflation (actual and expected) owing to lower production costs raises the real rate of interest, compressing demand and very possibly stifling any increase in output and employment. Indeed, those aggregates may both actually fall. Something like this may be going on at the present time in some economies. Chart 3 is suggestive of a depressing effect of low expected oil prices on expected inflation: it shows the strong recent direct relationship between U.S. oil futures prices and a market-based measure of long-term inflation expectations.
Initially, I was a bit enamored with this idea. As I thought on it more, however, I came to see it as a cautionary tale of chart crime. But digging underneath the surface a bit uncovered some interesting questions about monetary policy and credibility. Specifically, how worried should we be that inflation expectations will soon become unanchored?
Obstfeld et al. rely on a version of this widely publicized chart to support their contention:
The first and most obvious problem is that this chart really proves nothing. For example, I could just as easily presented this chart:
Now I can tell a story that the rising dollar (note the inverted scale) is driving down inflation expectations and thus driving up the real interest rate. Oil, on the other hand, is having exactly the effect we might expect - just look at sales of light trucks and SUVs, not to mention vehicle miles traveled: Hence, there is no paradox of oil. Lower oil prices are triggering the expected positive impacts. It's about the dollar weighing on inflation expectations that is creating the offsetting impact. Obstfeld et al. apparently do not try to distinguish their story from this one.
(Warning: wonkishness ahead.)
This problem, however, just scratches the surface. Look at either of the first two charts above and two red flags should leap off the screen. The first is the different scales, often used to overemphasize the strength of a correlation. The second is the short time span, often used to disguise the lack of any real long term relationship (I hope I remember these two points the next time I am inclined to post such a chart).
Consider a time span that encompassed the entirety of the 5-year, 5-year forward inflation expectations:
The correlation is less obvious to say the least (note too that changing the scale also suggests less correlation). Why does the correlation appear and disappear? Could any supposed correlation across selected time periods be spurious?
That gets to another issue. When you show me this chart:
and claim there is a meaningful relationship, I see two nonstationary variables you are claiming to be cointegrated. The trouble with that is that while oil is a nonstationary process - it is not mean reverting, nor is there reason to believe it should be a mean reverting series. Inflation expectations, however, should be a mean reverting series.
Or more specifically, it should be mean reverting if the central bank is credibly committed to their inflation target. If the central bank is credible, then we anticipate that policymakers will respond with policy that offsets inflation shocks to maintain their inflation target. Hence, inflation expectations should revert to that target and we would expect the series to be stationary.
If inflation expectations are a nonstationary series, then shocks build in the series and inflation expectations would drift persistently away from the central bank's inflation target. Inflation expectations would be unanchored. In other words, if inflation expectations are nonstationary, then we have a problem. More on that in a bit.
It appears that oil prices are nonstationary:
Dickey-Fuller Unit Root Test, Series DCOILBRENTEU
Regression Run From 1987:05:22 to 2016:03:21
Observations 7523
With intercept
With 1 lags chosen from 9 by AIC
Sig Level Crit Value
1%(**) -3.43430
5%(*) -2.86246
10% -2.56729
T-Statistic -1.47895
while, luckily, inflation expectations are stationary:
Dickey-Fuller Unit Root Test, Series T5YIFR
Regression Run From 2003:01:14 to 2016:03:24
Observations 3444
With intercept
With 7 lags chosen from 7 by AIC
Sig Level Crit Value
1%(**) -3.43524
5%(*) -2.86290
10% -2.56752
T-Statistic -4.63374**
Which leads me to conclude that the recent correlation between oil prices and 5-year, 5-year forward inflation expectations is not indicative of an underlying relationship and hence policymakers should be wary of accepting the Obstfeld at al. hypothesis. Of course, this should not be a surprise as the theoretical underpinnings for such a relationship are weak. A level shock to the price of oil should not change inflation expectations five years from now.
(The same is true for the dollar as well. And while both the dollar index and oil prices are nonstationary, they don't appear cointegrated, suggesting that instances of high correlation are more spurious than anything else.)
Digging a little deeper, note the University of Michigan Survey of Consumers has a longer series of 5-year inflation expectations which shows less variability than 5-year, 5-year forward inflation expectations:
The UMich inflation series also appears to be stationary:
Dickey-Fuller Unit Root Test, Series UMICH5YEAREX
Regression Run From 1990:08 to 2016:03
Observations 309
With intercept
With 3 lags chosen from 4 by AIC
Sig Level Crit Value
1%(**) -3.45322
5%(*) -2.87105
10% -2.57180
T-Statistic -2.92456*
Consequently, I think we have evidence to support the claim that the Federal Reserve is a credible policymaker in the most important arena, that of maintaining stable inflation expectations.
I suspect the high variability of the 5-year, 5-year forward measure is attributable to financial market structural issues (depth of the market for TIPS, for example) rather than rapidly shifting inflation expectations. Hence, we would expect that should those structural issues lessen in importance, the measure will revert to its mean (assuming the Fed remains a credible policymaker). Nor should we read too much about inflation expectations in this measure. Federal Reserve Chair Janet Yellen has reached the same conclusion, which is why she is wary of claims that shifts in the 5-year, 5-year forward measure reflect inflation expectations - and why she refers to these measures as inflation "compensation" not "expectations." From the March 2016 press conference:
In addition, the Phillips curve theory suggests that inflation expectations are also an important driver of actual wage- and price-setting decisions and inflation behavior, and I believe there’s also solid empirical evidence for that. And it’s one of the reasons that I highlighted in my statement, and we continue to highlight in the FOMC statement, that we are tracking indicators of the inflation expectations that matter to wage and price setting.
Now, unfortunately, we don’t have perfect measures of these things. We have survey measures. We know that household measures, even when households are asked about longer- term inflation—at longer-term inflation, they tend to move in response to salient changes in prices that they see every day. In particular, when gas prices go down, which is very noticeable to most households, you tend to see a view—you tend to see responses about long-term inflation marked down. So that’s kind of an overresponse to something that’s transitory. So it’s difficult to get a clear read from those survey measures.
Inflation compensation as measured in financial markets also embodies a variety of risk premia and liquidity premia. And so, it’s also—we monitor those closely and discuss them in the statement in paragraph one, but, again, there’s not a straight read on what’s happening to the expectations that influence wage and price setting. But this model continues to at least influence my own thinking, and it certainly is a factor that I and at least some of my colleagues are incorporating in these projections.
Note too that she also questions the importance of the recent slight downward drift in survey-based measures. I would place more weight on those measures (I think others on the FOMC, such as Governor Lael Brainard, are similarly inclined). In any event, I think the Fed is moving in a credible way to either measure by moving more cautiously than anticipated in December. Hence, we should expect inflation expectation measures to remain stationary.
Of course, if expectations devolve into nonstationary processes (it is a long-period property of the data, hence we cannot definitely declare the answer in any finite time period), we should be very worried that policymakers have lost control of inflationary expectations. And at the present time, they would be unanchored to the downside, not the upside as often feared.
Bottom Line: Be wary of claims that oil prices are influencing inflation expectations; the recent correlation is likely spurious. Inflation expectations look to be following a mean reverting process, indicating that the Federal Reserve's has credibly committed to their inflation target. We should expect policymakers will maintain such credibility if they continue to react to inflation shocks with offsetting policy.
