New Methods for Forecasting Inflation, Applied to the US*
Author | John Muellbauer,Janine Aron |
DOI | http://doi.org/10.1111/j.1468-0084.2012.00728.x |
Date | 01 October 2013 |
Published date | 01 October 2013 |
637
©John Wiley & Sons Ltd and the Department of Economics, University of Oxford 2012.
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 75, 5 (2013) 0305-9049
doi: 10.1111/j.1468-0084.2012.00728.x
New Methods for Forecasting Inflation, Applied to the
US*
Janine Aron† and John Muellbauer‡
†Department of Economics, Manor Road Building, Oxford OX1 3UQ, UK; Institute for New
Economic Thinking, Oxford Martin School, University of Oxford, Oxford, UK
(e-mail: janine.aron@economics.ox.ac.uk)
‡Nuffield College, Oxford OX1 1NF, UK; Institute for New Economic Thinking, Oxford Martin
School, University of Oxford, Oxford, UK (e-mail: john.muellbauer@.ox.ac.uk)
Abstract
Models for the 12-month-ahead US rate of inflation, measured by the chain-weighted
consumer expenditure deflator, are estimated for 1974–98 and subsequent pseudo out-
of-sample forecasting performance is examined. Alternative forecasting approaches
for different information sets are compared with benchmark univariate autoregressive
models, and substantial out-performance is demonstrated including against Stock and
Watson’s unobserved components-stochastic volatility model. Three key ingredients to
the out-performance are: including equilibrium correction component terms in relative
prices; introducing nonlinearities to proxy state-dependence in the inflation process and
replacing the information criterion, commonly used in VARs to select lag length, with a
‘parsimonious longer lags’ parameterization. Forecast pooling or averaging also improves
forecast performance.
I. Introduction
Stabilizing inflation is a key objective of monetary policy in the USA, and a large subset
of OECD and a few emerging market countries now target inflation as a primary objective
of policy. Since monetary policy is based on the likely path of inflation, it is important
that central banks have a reliable forecasting framework to avoid costly policy errors.
Forecasting inflation is notoriously difficult, however. Clements and Hendry (1998, 2002)
have long argued that structural breaks are the chief cause of forecast failure. Indeed, there
have been large structural shifts in the world economy, for instance in trade and financial
globalization, and in individual economies, such as a decline in trade union power and
ÅThe authors acknowledge funding support from the Economic and Social Research Council, UK (grant RES-
000-22-2066). This research was supported in part by the Open Society Foundation and the Oxford Martin School.
Comments by Chris Carroll, Kevin Lansing, Roland Meeks, John Williams and other seminar participants at the San
Francisco Federal Reserve are gratefully acknowledged. We are grateful to Mark Watson for advice and to Markus
Eberhardt for helping us implement Gauss software.
JEL classification numbers: E31, E37, E52, C22, C51, C52, C53.
638 Bulletin
the increasing economic weight of services. Monetary policy itself has generally shifted
to a greater focus on inflation. In addition, energy and food price shocks can be both size-
able and largely unpredictable, with the speed of price changes tending to rise with larger
shocks.
It is not surprising, therefore, that most forecasting models used by central banks put a
large weight on recent inflation. This approach tracks inflation quite well, except at turning
points, because the models miss underlying or long-term influences. In practice, central
banks do not rely on a single econometric model to guide policy making. Most central
banks augment the main model inflation forecasts by examining trends in individual price
components and specific sectoral information, for example on the expiry of gas supply
contracts, which might give clues to future changes. Outside the US Federal Reserve, the
New Keynesian Phillips Curve (NKPC) has been the dominant paradigm for modelling
inflation among macro-economists and central bankers.1
The current, mainstream version of the NKPC, proposed by Calvo (1983), claims
micro-foundations for sticky prices in the adjustment process. It assumes every firm has
the same probability of changing its prices: thus, one that has just adapted its prices, has
the same probability of adjusting prices again as a firm that has not changed for a long
time. Furthermore, this probability is constant over time – independent of the state of the
economy.This model implies that the rate of inflation is largely determined by the expected
rate of inflation (with a coefficient close to 1), and by the output gap (or the price level
relative to unit labour costs, in some versions). The implication for forecasting inflation
is that any information relevant for forecasting the output gap or unit labour costs should
help forecast inflation. There are open economy variants of the NKPC in which import
prices also enter (e.g. Batini, Jackson and Nickell, (2005). A hybrid form of the NKPC
adds lagged inflation with the sum of expected and lagged inflation close to 1, known as
the ‘accelerationist’ restriction.
One problem for the NKPC is inflation heterogeneity. Figure 1 illustrates this with
annual inflation rates for non-durable goods, durable goods and services in the US (see
also Peach, Rich and Antoniades, 2004). Durable goods have experienced negative infla-
tion for the last decade, while non-durable goods only occasionally experienced price falls
over a 12-month period. Prices of services have not fallen in any 12-month period in the
last four decades. Food and oil products, prominent among non-durables tend to have
highly flexible and volatile prices, while some service sector prices are set in annual con-
tracts. Moreover, rents, which are a key service sector component, adjust slowly to house
prices and interest rates, which are variables outside the NKPC framework. Empirical tests
against more general models than closed or open economy versions of the NKPC usually
find that the model restrictions are rejected.2There is also empirical evidence against the
rational expectations hypothesis embodied in the NKPC, using inflation forecasts from
surveys of households (Forsells and Kenny, 2002).
1The NKPC is a modern version of the expectations-augmented Phillips curve, that is, a relationship between
inflation, the output gap or other cyclical demand measure, and expected inflation. The classic reference is Woodford
(2003), though Clarida, Gali and Gertler (1999) did much to popularize the NKPC. Roberts (1995) discusses its
earlier variants.
2Examples are Bårdsen, Jansen and Nymoen (2004), Boug, Cappelen and Swansen (2006), Mavroeidis (2005)
and Rudd and Whelan (2007). In Bardsen et al. (2005, ch. 8) it is also shown that the NKPC model loses in simple
inflation forecast competitions to models incorporating relevant equilibrium correction terms.
©John Wiley & Sons Ltd and the Department of Economics, University of Oxford 2012
To continue reading
Request your trial