Laura Coroneo and Fabrizio Iacone, "Comparing Predictive Accuracy in
Small Samples Using Fixed-Smoothing Asymptotics", Journal of Applied
Econometrics, Vol. 35, No. 4, 2020, pp. 391-409.

There are 1 Matlab function and 6 data files (4 for the application
with the US Survey of Professional Forecasters and 2 for the
application with the ECB Survey of Professional Forecasters).

All files are ASCII files in DOS format. They are zipped in the file
ci-files.zip. Unix/Linux users should use "unzip -a".

---------------------------------------------------
               MATLAB function    
---------------------------------------------------

The Matlab function "dm_fsa.m" performs the tests for equal predictive
accuracy using fixed-smoothing asymptotics.

INPUTS: 
- d         loss differential (T x 1)
- wceband   bandwidth for Weighted Covariance Estimate (default = T^(1/2))
- wpeband   bandwidth for Weighted Periodogram Estimate (default = T^(1/3))
- tsize     level of significance of the two-sided test (5% (default) or 10%)

OUTPUTS: 

- test  2x1 array with test statistic values: 
        1) Test statistic with Weighted Covariance Estimate and Bartlett kernel
        2) Test statistic with Weighted Periodogram Estimate and Daniell kernel

- reject   2x1 array (0 = null of equal predictive accuracynot rejected;
                      1 = null equal predictive accuracy rejected) for the:
           1) Fixed-b test
           2) Fixed-m test

---------------------------------------------------
Data for the US Survey of Professional Forecasters       
---------------------------------------------------

The files "RGDP_US.csv", "PGDP_US.csv", "UNEMP_US.csv" and
"TBILL_US.csv" contain the data for the SPF application for real
output growth, inflation, the unemployment rate and the three-month
Treasury bill, respectively. 

The data source is the Federal Reserve Bank of Philadelphia,
https://www.philadelphiafed.org/research-and-data/real-time-center/survey-of-professional-forecasters/data-files/error-statistics.
The original data file names are:

- Data_SPF_Error_Statistics_RGDP_3_AIC; 
- Data_SPF_Error_Statistics_PGDP_3_AIC; 
- Data_SPF_Error_Statistics_UNEMP_1_AIC; 
- Data_SPF_Error_Statistics_TBILL_1_AIC. 

Each data file organized in eleven columns: time, SPF made in the
current quarter, SPF made 1 quarter ahead,... SPF made 4 quarters
ahead, random walk forecast made in the current quarter, random walk
forecast made 1 quarter ahead, ... random walk forecast made 4
quarters ahead, realised value. Sample: 1987.q1 to 2016.q4. 

---------------------------------------------------
Data for the ECB Survey of Professional Forecasters   
---------------------------------------------------

The files "RGDP_EU.csv" and "HICP_EU.csv" contain the data for the ECB
SPF application for GDP growth and HICP inflation. 

The data source is the European Central Bank website. 

ECB SPF forecasts are downloaded from:

- https://www.ecb.europa.eu/stats/ecb_surveys/survey_of_professional_forecasters/html/table_hist_rgdp.en.html
for real GDP growth;

- https://www.ecb.europa.eu/stats/ecb_surveys/survey_of_professional_forecasters/html/table_hist_hicp.en.html
for inflation. 

Real time European data are downloaded from
https://sdw.ecb.europa.eu/browseSelection.do?node=9689716, the data
file names are:

- RTD.Q.S0.S.G_GDPM_TO_C.E for real GDP
- RTD.M.S0.N.P_C_OV.A for inflation. 

Each data file is organized in eight columns: time, ECB SPF made 1
year ahead, ECB SPF made 2 years ahead, ECB SPF made 4 years ahead,
random walk forecast made 1 year ahead, random walk forecast made 2
years ahead, random walk forecast made 5 years ahead, realised value.
The random walk forecasts and the real time real GDP growth are own
computation from the data. Sample: 2006.q1 to 2016.q4.