This repositories makes available the code to produce all figures and tables in the paper: An, Collodel and Loungani (2021). "When (where and why) forecasters get it wrong?". Unpublished manuscript. Please cite us if you refer to our paper.
We assemble a large panel of real GDP forecasts from multiple sources and with these new data study the performance of different forecasters over time and the determinants of forecast errors. We find that a) forecasters are extremely precise during normal periods, but never predict the onset and extent of a recession b) forecasts from different sources are extremely correlated and c) that political economy considerations play an important role in understanding forecast errors.
Umberto Collodel ([email protected])
Zidong An ([email protected])
Prakash Loungani ([email protected])
R
Download all files in the folder
https://www.dropbox.com/sh/33sosx2ltq6vif4/AADbEPy8EGOPjfPsJh9fVWdRa?dl=0
The folder contains:
- IMF World Economic Outlook Data
weo_rgdp: GDP growth forecasts contained in Spring and Fall WEO publication, one sheet for each year/vintage
weo_january_ypdate: GDP growth forecasts contained in January WEO publication, one sheet for each year
weo_july_update: GDP growth forecast contained in July publication, one sheet for each year
- World Bank GEP Data
GEP_forecast_vintages_06232020: GDP growth forecasts contained in Summer and Winter GEP publication, available both in long and wide format
- Consensus Data
gdp_2008_2019_firstweek: GDP growth forecasts contained in April and September version of Consensus survey, one sheet for each year/vintage - mainly advanced economies available in the first week
gdp_2008_2019_secondweek: GDP growth forecasts contained in April and September version of Consensus survey, one sheet for each year/vintage - mainly emerking markets and low income economies in the second week
- MONA Data
mona_2002-2020_macro: forecasts for main macroeconomic variables for inception of program and subsequent reviews
Note: the script retrieves European Commission forecasts from AMECO directly scraping the website and downloanding them, this is why they are not available as raw data.
-
Main Source
-
Preparation datasets
-
Short-term analysis (current-year and year-ahead): bias and efficiency
-
Comparison with official and private sector of short-term forecasts
-
Medium-term analysis: bias and efficiency
-
Programs and forecast errors
The main sourcing file runs the entire project. The main file of each section cleans the global environment, installs and loads the packages required and sources all the scripts in the section. The function file contains custom functions to generate tables and produce graphs for the section. Individual files run the function and export the output.
The data and codes are under the MIT license. This means that you can use everything as you please for research or commercial purposes, as long as you refer back to us.
If you find irregularities or bugs, please open an issue here.