The Econmist - USA (2021-11-06)

(Antfer) #1

The Economist November 6th 2021 GraphicdetailInflation 81


flaws.  Changes  in  food  and  energy  prices
are  not  necessarily  unusually  large  or
short­lived.  And  trimmed  means’  weight­
ing schemes are plagued by abrupt cliffs. In
the Cleveland Fed’s version, which lops off
the  top  and  bottom  8%  of  the  index,  an
item in the 93rd percentile when sorted by
price changes is removed entirely, whereas
one in the 92nd gets its full weight.
With this in mind, we have devised an
alternative  inflation  index.  Like  trimmed
means,  it  adjusts  items’  weights  based  on
their recent price changes. But its weights
are shaped like a smooth hill rather than a
box.  Components  with  inflation  near  the
median get the most emphasis, and those
with the biggest price changes get the least.
Our hill looks a bit like Uluru in Austra­
lia:  a  broad  central  plateau,  flanked  by  a
steep  slope  on  the  left  side  and  a  gentler
one on the right. (The Dallas Fed’s trimmed
mean  is  also  asymmetric,  counteracting
bias caused by price­change distributions’
lopsidedness.) Most items with negative or
low  inflation  get  a  hefty  weight;  those
whose  prices  are  rising  fastest  count  for
25% as much as those in the middle do. 
When  using  the  past  year  of  data  to
predict pceinflation during the following
year, this method is more accurate than ei­
ther using core inflation or expecting infla­
tion  to  remain  constant.  Since  1959,  its
one­year forecasts have also outperformed
those of the Dallas Fed’s trimmed mean.
Some of this apparent advantage stems
from the design of our study: the Dallas Fed
sought to maximise accuracy for different
time  periods  and  forecast  horizons  than
ours.  However,  its  trimmed  mean’s  errors
in  the  1970s  illustrate  the  risks  of  such  a
deep  trim.  Amid  two  oil  shocks,  some  of
the fastest­rising prices—those of energy­
related  items—just  kept  rising.  The  less
weight an index placed on such goods, the
worse it predicted inflation one year out.
The current episode of concentrated in­
flation  differs  from  the  1970s  in  many
ways.  Surging  prices  for  goods  like  home
appliances are unlikely to feed through to
other costs, as oil prices do. And in general,
treating  outliers  the  same  as  more  repre­
sentative items has been a mistake. But in
cases where such price changes foreshad­
ow  broader  supply  constraints,  ignoring
them entirely can be an even bigger error.
As  a  sense­check  of  our  “Uluru”  meth­
od, we also built a model that forecasts pce
inflation  using  only  excess  concentration
and the one­ and ten­year trailing inflation
rates. Both this approach and the Uluru in­
dex  yield  a  4.1%  prediction  for  inflation
during  the  next  12  months.  That  is  below
the  current  level  of  4.4%,  but  above  the
Fed’s target of 2% and the 2.3% value of the
Dallas  Fed’s  trimmed  mean  for  thepast 12
months. If this forecast comes true,inter­
est­rate rises will almost surely follow.n


→ Four ways to weight an inflation index

Headline inflation*
Weightsitemsattheirshareofconsumerspending

Core inflation
Excludesfoodandenergyprices

Trimmed-meaninflation
Excludesproductswithbigpriceswings

“Uluru”inflation
Smoothlyreweightsitems

*Weightedaverageofeachitem’syear-on-yearchangeinpersonalconsumptionexpendituresindex
†Differenceinroot-mean-squareerrorpredictingheadlinePCEinflationduringthesubsequent 12 months

→“Uluru”inflationislessvolatilethanheadline,anda betterpredictor

Headlinev “Uluru”inflationindices,%changeona yearearlier

Housing
Food

Prescriptions

Prescriptions

Petrol

Food

Petrol

Food Petrol

Petrol

Housing

Housing

Housing

Food

0

0

0.5x

Equal to
headline

Equal to
headline

1.5x

0.5x

1.5x

2.0x

2.5x

← Lower Percentile of price change Higher →

← Lower Percentile of price change Higher →

← Lower Percentile of price change Higher →

← Lower Percentile of price change Higher →

Weight ratio, May 2021

Weight ratio, May 2021

↓Headlineclosertonextyear’sinflation†

↑“Uluru”closertonextyear’sinflation†

Headline index

“Uluru” index

-2.5

0

2.5

5.0

7.5

10.0

12.5

1960s 70s 80s 90s 2000s 10s 20s

Covid-1
pandemic

Global financial
crisis

Oil
shock

Oil
shock

Fed-induced
recession

Sources: Bureau of Economic
Analysis; The Economist
Free download pdf