Download Understanding Inflation: The Impact of Energy Prices and Supply Shocks and more Exams Monetary economics in PDF only on Docsity!
The Supply-Shock Explanation of the
Great Stagflation Revisited*
Alan S. Blinder Jeremy B. Rudd
Princeton University Federal Reserve Board
February 2010 revision
ABSTRACT
U.S. inflation data exhibit two notable spikes into the double-digit range in 1973- and again in 1978-1980. The well-known “supply-shock” explanation attributes both spikes to large food and energy shocks plus, in the case of 1973-1974, the removal of price controls. Yet critics of this explanation have (a) attributed the surges in inflation tomonetary policy and (b) pointed to the far smaller impacts of more recent oil shocks as evidence against the supply-shock explanation. This paper reexamines the impacts of the supply shocks of the 1970s in the light of the new data, new events, new theories, and new econometric studies that have accumulated over the past quarter century. We find that the classic supply-shock explanation holds up very well; in particular, neither datarevisions nor updated econometric estimates substantially change the evaluations of the 1972-1983 period that were made 25 years (or more) ago. We also rebut several variants of the claim that monetary policy, rather than supply shocks, was really to blame for the inflation spikes. Finally, we examine several changes in the economy that may explain why the impacts of oil shocks are so much smaller now than they were in the 1970s.
September 2008. Blinder gratefully acknowledges research support from Princeton’s^ * Paper presented at the NBER conference on The Great Inflation, Woodstock, VT, Center for Economic Policy Studies. We also thank Olivier Blanchard, other conference participants, and two referees for useful suggestions. The opinions expressed here are our own, however, and do not necessarily reflect the views of any of the institutions with which we are affiliated.
“Everything should be made as simple as possible, but not simpler.” Albert Einstein
1. Preamble
Between, say, the first OPEC shock and the early 1980s, economists developed what has been called “the supply-shock explanation” of what this conference calls “the Great Inflation,” that is, the period of high inflation seen in the United States (and elsewhere) between 1973 and 1982.^1 At the conceptual level, the supply-shock explanation can be succinctly summarized by four main propositions:
- At any given moment, there is an underlying (or “core”) inflation rate toward which the actual (or “headline”) inflation rate tends to converge. This rate is determined by the fundamentals of aggregate demand and aggregate supply growth.
growth rate of aggregate demand. On the supply side, the fundamental driving factor in2. Many factors, including but not limited to monetary and fiscal policy, influence the the long run is the growth rate of productivity, but occasional abrupt restrictions in aggregate supply (“supply shocks”) can dominate over short periods.
change of prices for all items other than food and energy.3. For empirical purposes, the core rate of inflation can be proxied by the rate of
- The headline inflation rate can deviate markedly from the core rate over short periods. Rapid increases (or decreases) in food or energy prices, which are largely exogenous, can push inflation above (or below) the core rate for a while. There may beother special one-shot factors as well, such as the 1971-1974 Nixon wage-price controls.
This model, if you want to call it such, was applied by a number of scholars to explain the history of the Great Inflation with six additional propositions:^2
- The dramatic rise in inflation between 1972 and 1974 can be attributed to three major supply shocks—rising food prices, rising energy prices, and the end of the Nixon wage-price controls program—each of which can be conceptualized as requiring rapid adjustments of somestory.) relative prices. (Thus nominal rigidities play a central role in the
- The equally dramatic decline in inflation between 1974 and 1976 can be traced to (^12) For a short but comprehensive summary in an earlier NBER volume, see Blinder (1982, pp. 262-264). specific six points listed here follow Blinder (1982).Among the many who could be listed, see Gordon (1975), Phelps (1978), and Blinder (1979, 1982). The
coherent explanation of the inflation of the 1970s must explain both the ups and the downs. In addition, however, Figure 1 displays a clear upward drift in core inflation, from under 2 percent in 1964, to around 4 percent by 1970, and then to about 6 percent by 1976—before it falls back to 4 percent or so after 1983. This upward drift, which is presumably explainable by the fundamental factors listed in point 2 above, constitutes an interesting and important macroeconomic episode in itself—and one that has certainly not gone unnoticed! 6 But it is not the subject of this paper. Had the upward drift in inflation from 2 percent to 6 percent (and then back down to 4 percent) been all that happened, no one would have dreamed of calling this episode the “Great Inflation.” Hence we focus squarely on the two big “inflation hills” that are so evident in the figure. Second, the Great Inflation was really the Great Stagflation. Any coherent explanation must also explain the contemporaneous deep recessions. In particular, the economy did not merely experience real output declines over these two periods. Unemployment also rose sharply, implying that what was going on in each case was more than just a neoclassical drop in output in response, say, to the rise in the relative price of energy.
(^6) Among these fundamental factors, we would count Vietnam War spending in the late 1960s, over- expansionary monetary policy, and the post-1973 productivity slowdown. According to CBO estimates,the unemployment rate was below the NAIRU in every year from 1964 through 1974.
Figure 1 Consumer price inflation, 1964-
A. Current-methods CPI
Headline
Core
B. PCE price index
Headline
Core
Note: Four-quarter log differences.
0
2
4
6
8
10
12
1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984
0
2
4
6
8
10
12
1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984
Section 6 draws some conclusions. But we can end the suspense right now by stating that, at least in our judgment, the “old fashioned” supply-shock explanation holds up quite well.
2. What is the supply-shock explanation of the Great Stagflation?
First we must define what we mean by a “supply shock.” We begin, as is now conventional (but was not in 1973), by dividing the various influences on output and prices into two categories: factors that influence aggregate supply (“supply shocks”) and factors that influence aggregate demand (“demand shocks”). Their respective hallmarks can be described in either of two ways.
- Supply shocks affect the ability of firms to produce the gross domestic product, which means that they directly affect either the prices or quantities of factor inputs or the production technology. The resulting changes in output can be thought of as basically neoclassical in nature. On the other hand, demand shocks affect spending by the households, businesses, and governments that purchase the GDP. Naturally, any demand shock will have short-run Keynesian effects ( e.g. , result in changes in real output) if the economy has Keynesian properties— which it does.^7
- Supply shocks are events that, on impact, move the price level and real output in opposite directions ( e.g. , an adverse shock causes prices to go up and output to go down). Demand shocks are events that, on impact, move the price level and real
(^7) For this purpose, we define “Keynesian properties” as the presence of nominal rigidities plus some inertia in wage and price setting (whether from expectations or not) that makes this behavior at least somewhatbackward looking. We exclude purely forward-looking models with rational expectations. As is well- known, models in this latter class carry starkly different—and generally counterfactual—implications.
output in the same direction ( e.g. , an expansionary demand shock pushes up both prices and output). The second definition is exemplified by the standard aggregate supply and demand diagram (shown in Figure 2) in which an upward-sloping aggregate supply curve shifts inward along a fixed aggregate demand curve, thereby simultaneously raising the price level and reducing output—a stagflationary outcome. The non-vertical aggregate supply curves AS 0 and AS 1 , of course, embody some sort of nominal wage-price stickiness.
Figure 2Supply shocks in the AS/AD framework
P (^) AD AS 1 (^0) AS 0
P 1 P 0
Y 1 Y (^0) Y Either of the two definitions will suffice for our purposes. But it is important to note that some shocks have both supply-side and demand-side elements. A shock to the price of imported oil is, of course, the most prominent example. We will show later that neoclassical supply-side considerations alone cannot come close to explaining the magnitudes of the two recessions that occurred during the Great Stagflation. Rather, to explain these episodes, the two big oil shocks must be viewed as having affected both
Figure 3 Real Oil and Energy Prices, 1965-
A. Real oil price (2000 dollars)
B. Real consumer price of energy (PCE-based, 2000=100)
Note: Series deflated by headline PCE price index; see Appendix for data definitions. Last observation is Sept. 2008.
0
10
20
30
40
50
60
70
80
90
100
110
1965 1970 1975 1980 1985 1990 1995 2000 2005
70
80
90
100
110
120
130
140
150
160
170
180
1965 1970 1975 1980 1985 1990 1995 2000 2005
Figure 4 gives these qualitative points a quantitative dimension. To study pass- through empirically, we estimated a relatively standard backward-looking price-price Phillips curve model of U.S. inflation on monthly data from January 1961 to December 1984. The basic specification takes the form
t 0 A ( L ) t 1 B ( L ) xt 1 G ( L ) t 1 + εt ,
where πt is the inflation rate; x is the detrended unemployment rate, used here as a measure of slack; ζ is a supply-shock term; and ε is a stochastic error.^10 The supply- shock variable in our baseline specification (we tried several variants) is a weighted average change in relative food and energy prices, using smoothed PCE shares as weights.^11 We take a six-month moving average of this weighted relative price change variable, and use its first lag in the model (additional lags did not enter). For this exercise, we treated the relative price of energy as exogenous, with panel A of Figure 4 plotting the precise path of energy we assumed: Relative energy prices rise by 35 percent (30 log points) over a period of 12 months and then return to where they started over the next 12 months.^12 Panel B of the figure shows the simulated inflation results, which are just as expected. Headline inflation rises quickly and sharply by about (^10) The number of monthly inflation lags used in the model was determined with the Akaike criterion, with twelve lags used as the default. Note, however, that we did not impose the “accelerationist” restriction A (1) = 1. The estimated model also includes additional terms to capture the impact of the Nixon wage- price controls—see Section 3.1.3 for a discussion. (Appendix 2 provides more details on this and otherempirical specifications employed in the paper; the calculations that underpin Figures 4 through 6 are described more fully in Appendix 3.) (^11) Ideally, a CPI-based model would use CPI relative importance weights rather than PCE shares. Unfortunately, there are significant breaks in the relative importance weight series over time—notably in1978, when the CPI moved from measuring prices faced by wage earners to prices faced by all urban consumers, and again after owner-occupied housing costs moved to a rental equivalence basis. In anyevent, whether weighting is used turns out to make little difference to the results. (^12) The actual OPEC II peak was spread out over a longer period: Oil and (especially) energy prices did not return to their pre-shock levels until the collapse in oil prices in 1986. Note that this and the other twosimulations assume a 30 log-point increase in relative energy (not oil) prices. Following OPEC I, real energy prices rose about 20 log points; the corresponding increase after OPEC II was 35 points, and thefive-year net increase after the end of 2002 was around 45 points—so a 30 point increase is (in round terms) close to the average increase in log real energy prices over these three episodes.
how we entered this type of energy-price shock into our econometric model: The energy price is assumed to jump by 30 log points (35 percent) over two quarters and then to remain there forever. The speed of this simulated shock is not too different from what actually happened in 1973-74: While the rise in oil prices took place over a period of about four months, the bulk of the pass-through to retail energy prices occurred over an eight-month period.
Panel B of Figure 5 shows the simulated impact of the shock on headline and core inflation. Headline inflation leaps quickly and dramatically (by about 6 percentage points), but then recedes just as quickly. After six months, the direct contribution of energy prices to headline inflation is zero. Core inflation moves up much more slowly and by much less. But the effects on core and headline inflation are essentially identical as soon as energy prices have finished moving up to their new higher level—and they both die out very slowly. So, in terms of the basic supply-shock story, a permanent increase in the level of energy prices should cause a quick burst of inflation which mostly, but not quite (because of pass-through to the core), disappears of its own accord.
Figure 5Effect of a permanent jump in energy prices A. Level of real energy price B. Path of headline and core inflation (monthly change at AR)
Headline
0.01.0 Core
0.9 -4 0 4 8 12 16 20 24 28 32 36 40 44 48
-4 0 4 8 12 16 20 24 28 32 36
Once again, headline inflation quickly converges to core, but now core inflation remains persistently higher than it was before the shock. As is evident in Figure 1, a similar pattern can be seen in actual U.S. inflation during and after OPEC I. 14 Writing in the 1980s or 1990s, our typology might have stopped there. But this decade has taught us that we should perhaps consider a third type of supply shock; namely, a long-lasting rise in the rate of energy price inflation, as exemplified by the stunning run-up in the real prices of oil and energy from 2002 until mid-2008 (see Figure 3). We entered this third type of shock into our model as a permanent rise from a zero rate of relative energy price increase to a rate of 6 percent per year, which cumulates to a 35 percent increase in the level of real energy prices over five years. (Panel A of Figure 6 shows the first three years of the assumed real energy price path.) This hypothetical history is qualitatively similar to what actually occurred between 2002 and mid-2008, although the actual increase in real oil prices was, of course, followed by a spectacular decline. Panel B of Figure 6 shows the model simulation results. Headline inflation starts rising right away and continues to rise very gradually. Core inflation does the same, though with a short lag and to a smaller degree. But notice that inflation keeps on rising as long as the higher energy inflation persists. Headline inflation now does not converge to core until real energy prices stop rising. Nor does the impact on core inflation fade away until that happens.^15
(^14) It would be even more evident were it not for the effects that price controls had on the core. We discuss these in Section 3.1.3. (^15) The estimated effect on core inflation from this third simulation is almost certainly higher than current reality. As we discuss in Section 5, the pass-through of energy price shocks to core inflation appears to bemuch smaller now than in the 1970s and early 1980s, but the model used to generate these simulations is estimated through 1984.
Using national accounts data to compute the effects of higher prices of imported petroleum and products on real GDP, we find that the 1973-74 oil shock implies a cumulative reduction in real GDP of 1.1 percent through the first quarter of 1975. Similarly, the OPEC II shock implies a real GDP reduction of 1.7 percent through the second quarter of 1980. (Details of these calculations are provided in Appendix 3.) But the actual decline in real GDP in the United States (relative to trend) was much larger in each case. For example, real GDP fell a little more than 3 percent between its 1973:4 peak and its 1975:1 trough, a five-quarter period during which normal (pre-1973) trend growth would have called for an increase of around 4.5 percent. Thus, in round numbers, we lost nearly 8 percent of GDP relative to trend.^16 The period of the two oil shocks also saw large increases in the prices of other imported materials (in addition to oil). It is straightforward to extend the Bruno-Sachs framework to incorporate multiple imported inputs and to compute the real GDP effects of their price increases. Even with this extension, however, the impacts of the supply shocks are far smaller than the observed GDP declines. For 1973-75, the supply-side reduction in real output from both higher oil and nonoil material prices cumulates to 1.6 percent, while the corresponding estimate for the OPEC II period is 1.9 percent. (See Appendix 3 for details.) In addition, the pure neoclassical view does not provide any particular reason to think that unemployment should rise following an oil shock. In that framework, real wages and the rate of profit fall by enough to keep labor and capital fully employed. Put differently, a purely neoclassical oil shock reduces both actual and potential output
(^16) We obtain almost identical estimates of the cumulative GDP shortfall by using CBO’s ( ex post ) measure of potential output.
equally, leading to no GDP gap (if the gap is measured correctly). In fact, however, the U.S. unemployment rate soared from 4.8 percent in the second half of 1973 to almost 9 percent in the second quarter of 1975. Both of these calculations suggest that something else was going on—probably something Keynesian on the demand side.^17 2.3 The “oil tax” That something is often called “the oil tax.” The idea is simple: If imported energy— which mainly means imported oil—becomes more expensive, the real incomes of Americans decline, just as if they were being taxed by a foreign entity. The “tax” hits harder the less elastic is the demand for energy, and we know that the short-run price elasticity is low. Using OPEC I as an example, the nominal import bill for petroleum rose by $21.4 billion through the end of 1974, which represented about 1.5 percent of 1973’s GDP. If the marginal propensity to consume (MPC) was 0.9, this “tax” would have reduced non-oil consumption by almost 1.4 percent of GDP. If standard multiplier- accelerator effects created a peak multiplier of 1.5, the maximal hit to GDP would have been about 2 percent, or almost twice as large as the neoclassical supply-side effect.^18 Adding the two together would bring the total reduction in GDP to a touch above 3 percent, which is still far less than actually occurred. These calculations encompass only imported oil. But there was also an internal redistribution within the United States, as purchasing power was transferred from energy
(^1718) Or, possibly, something “new-Keynesian” on the supply side (cf. Rotemberg and Woodford, 1996). and Enzler (1974)—that used an early version of the MPS model to attribute approximately a 3 percentBy comparison, Blinder (1979, pp. 84-85) cited two econometric studies—by Perry (1975) and Pierce decline in real GDP to OPEC I. We are aware of the continuing controversy over the size of the multiplier(see, for example, Hall, 2009). Naturally, using a smaller multiplier would make these effects smaller as well.
Point (ii) above raises an important issue that we will return to several times in this paper: The impact of a supply shock on real output and inflation depends critically on how the monetary authorities react. Monetary accommodation to mitigate the incipient recession will produce larger effects on inflation and smaller effects on output and employment. Monetary tightening to mitigate the increase in inflation will produce just the opposite. This is one, though not the only, reason why responses to oil shocks vary both across countries and across time. 2.4 “Second-round” effects Another important issue, related of course to monetary policy, is how much “second- round” inflation is induced by the “first-round” price-level effects of supply shocks—as, for example, higher energy costs creep into the prices of other goods and services and into wages. Regarding the price channel, Nordhaus (2007, p. 223) recently used an input-output model to estimate that the long-run pass-through of energy costs into other consumer prices (which include airfares, apartment rents, and so on) is 80 percent as large as the direct effect of energy prices on the index. However, this estimate overstates the short- to- medium run effects of an energy-price shock. For example, airfares will react quickly to higher fuel costs, but the higher cost of the energy used to manufacture airplanes will probably not show up in airfares for years.^20 That said, there is still significant scope for sizable second-round price effects. Indeed, as might be expected, energy-intensive (^20) Moreover, it matters whether an estimate of this sort is based on crude or finished energy. Using the 1992 input-output accounts, we estimate that finished energy costs accounted for 3.4 percent of non-energyPCE, while crude energy costs only accounted for 1.5 percent (these estimates include an imputation for the energy costs incurred in transporting and distributing consumption goods). For core PCE, the estimatesare a little smaller (3 percent and 1.3 percent, respectively). That said, this still appears to be a reasonably large indirect effect given that the direct effect of finished energy on PCE prices (measured as the nominalshare of energy goods and services in total consumption) was 5½ percent in that year, and also given that crude energy price changes tend to be much larger than changes in finished energy prices.
consumption goods and services posted relatively larger price increases following the first two oil shocks.^21 As evidence, the first two columns of Table 1 report rank correlations between energy intensity and three-year price changes for various groupings of individual PCE components following OPEC I and OPEC II.^22 These correlations are similar whether one looks at total, core, non-energy, or non-transportation components of PCE. (As is evident from the rightmost column of the table, however, similar correlations cannot be found during the most recent run-up in oil prices—a point to which we will return in Section 5.) Table 1. Rank Correlations between Energy Intensity and Price Change Correlation with change in price from 1972 to 1975 1978 to 1981 2002 to 2007
- All PCE components 0.308^ 0.3430.283**** 0.150a 0.326 0.193*
- Nonenergy PCE 0.256**^ −0.0480.
- Nonenergy ex. transportation 0.270 **^ −0.
- Core components 0.267**^ −0.
- Core ex. transportation 0.275**^ −0. Note: *//a^ denotes significant at 1/5/10 percent level, respectively. One simple way to study the pass-through question is to examine the impacts of supply shocks on measures of core inflation, which by definition remove the mechanical impacts of energy and food prices on headline inflation. We did this earlier in Figures 4 through 6, which showed the results of passing three different types of stylized supply
(^21) In a more specialized context, Weinhagen (2006) finds evidence of significant pass-through of crude petroleum prices into the PPIs for plastics and organic chemicals over the period 1974-2003. (^22) The rank correlation measure that we use is Kendall’s “tau-b,” which is more robust to the presence of ties across rankings. Energy intensities are estimated using data on total crude energy requirements fromthe 1972 and 1977 input-output tables. Price changes are December-over-December increases computed over the relevant periods.
