James M. Nason and Gregor W. Smith, "Measuring the Slowly Evolving
Trend in US Inflation with Professional Forecasts", Journal of Applied
Econometrics, Vol. 26, No. 1, 2021, pp. 1-17.

There are two data files, cpi_pi_May2019_JAE.txt and
out_pi_May2019_JAE.txt. These are ASCII files in DOS format. They are
zipped in the file ns-data.zip. Unix/Linux users should use "unzip -a".

The source of both datasets is the the Real-Time Data Research Center
(RTDRC) of the Federal Reserve Bank of Philadelphia:

https://www.phil.frb.org/research-and-data/real-time-center/survey-of-professional-forecasters

Both datasets consist of average inflation predictions taken from the
Survey of Professional Forecasters (SPF) and realized inflation.

The RTDRC compiles 0- (or nowcast), 1-, 2-, 3-, and 4-quarter ahead
average inflation predictions of the Survey of Professional
Forecasters (SPF).  

Realized inflation is the fifth definition of CPI or PGNP/PGDP
deflator inflation in the SPF forecast error statistics spreadsheets
tabulated by the RTDRC.

The data were downloaded fom the RTDRC-SPF website in May 2019.


A) cpi_pi_May2019_JAE.txt consists of 150 rows and six columns.  

The 150 rows are quarterly data from 1981Q3 to 2018Q4.  However, the
paper estimates on a sample from 1981Q4 to 2018Q4.

The six columns are 0-, 1-, 2-, 3-, and 4-quarter ahead average SPF
inflation predictions and realized inflation.


B) out_pi_May2019_JAE.txt consists of 201 rows and six columns.  

The 201 rows are quarterly data from 1968Q4 to 2018Q4.  However, the
paper estimates on a sample from 1969Q1 to 2018Q4.

The six columns are are 0-, 1-, 2-, 3-, and 4-quarter ahead average
SPF inflation predictions and realized inflation.

There are several observations coded -99.0 for the 4-quarter ahead
average SPF inflation predictions early in the sample.  These are
missing observations in the PGNP/PGDP-SPF data.


C) The paper only uses the 1-, 2-, 3-quarter, and 4-quarter ahead
average inflation predictions of the SPF and realized inflation.

First, compute x(i,t) = 100*ln( 1 + z(i,t)./100 ), where x(i,t) = the
1-, 2-, and 3-quarter ahead average inflation predictions of the SPF.

Similarly, realized inflation is pi(t) = 100*ln( 1 + z(pi,t)./100 ),
where z(pi,t) = the fifth definition of inflation in the SPF forecast
error statistics spreadsheets.

Next, construct the dependent variables y(i,t) = x(i,t) - pi(t), i =
1, 2, and 3.

In the sticky information model, the lagged dependent variables are
y(i,t-1) = x(i,t-1) - pi(t-1), i = 2, 3, and 4.

Please, see section A.2.a of the appendix to the paper for more
information.  The appendix is available at
https://www.jamesmnason.net/s/NasonSmith_appendix_26April2020.pdf

Jim Nason (jmnason [AT] ncsu.edu)
Gregor Smith (smithgw [AT] econ.queensu.ca)

26 April 2020