Dean Croushore and Katherine Marsten, "The Continuing Power of the Yield Spread in Forecasting Recessions", Journal of Applied Econometrics, Vol. 31, No. 6, 2016, pp. 1183-1191. All data files are ASCII files in DOS format. They are zipped in the file cm-data.zip. Unix/Linux users should use "unzip -a". The files described below contain the various time series of data used in the article (all at a quarterly frequency), with rows corresponding to dates and columns to variables. The first column in each file provides the dates of the observation. The first row lists the variables. The Survey of Professional Forecasters is the most well-known survey of U.S. macroeconomic forecasts. The survey participants are asked to report the probability that real output will decline in the current quarter and in each of the following four quarters. The mean probability forecasts across all the participants are the main recession forecasts used in this article. The data used are publically available and can be found at https://www.philadelphiafed.org/research-and-data/real-time-center/survey-of-professional-forecasters/historical-data/probability-variables. The following file contains the data with observations from 1968q4 to 2013q2. File #obs #var RECESS.txt 180 11 Variables RECESS1: quarterly forecasts of the probability of a decline in real output in the current quarter RECESS2: quarterly forecasts of the probability of a decline in real output for one quarter ahead RECESS3: quarterly forecasts of the probability of a decline in real output for two quarters ahead RECESS4: quarterly forecasts of the probability of a decline in real output for three quarters ahead RECESS5: quarterly forecasts of the probability of a decline in real output for four quarters ahead RECESS1B: RECESS1 divided by 100 RECESS2B: RECESS2 divided by 100 RECESS3B: RECESS3 divided by 100 RECESS4B: RECESS4 divided by 100 RECESS5B: RECESS5 divided by 100 The Real-Time Data Set for Macroeconomists (RTDSM) consists of vintages of data series for real output (GNP before 1992, GDP since 1992). The data used are publically available and can be found at https://www.philadelphiafed.org/research-and-data/real-time-center/real-time-data/data-files/routput. The following files contain the monthly vintages of the quarterly real output data. The vintages are from 1965m11 to 2013m5 with observations from 1947q1 to 2013q1. The first row of each file lists the vintage variable names. For example, ROUTPUT65M11 is the vintage for 1965m11 and ROUTPUT13M5 is the vintage for 2013m5. The vintages are divided between six data files. File #obs #var first vintage last vintage routputMvQd1.txt 271 51 1965m11 1969m12 routputMvQd2.txt 271 121 1970m1 1979m12 routputMvQd3.txt 271 121 1980m1 1989m12 routputMvQd4.txt 271 121 1990m1 1999m12 routputMvQd5.txt 271 121 2000m1 2009m12 routputMvQd6.txt 272 42 2010m1 2013m5 The first final actual series and the R1FF variables are calculated using the real output vintages. The following files are the results. The Y_ACT_1Q_FF variable has observations from 1968q4 to 2012q4. The R1FF variable has observations from 1955q1 to 2012q4. File #obs #var y_act_1q_ff.txt 178 2 R1FF.txt 233 2 Variables Y_ACT_1Q_FF: real output actuals based on the first final release R1FF: takes the value 1 in quarters when real output declines and the value 0 in quarters when real output increases For calculating the yield spread, we use the interest rate on 10-year U.S. government Treasury bonds minus the interest rate on three-month Treasury bills. The data for these two series are from the FRED database maintained by the Federal Reserve Bank of St. Louis, using quarterly averages. These series span from 1955q1 to 2013q1. The data for the 10-year U.S. government Treasury bonds can be found at https://research.stlouisfed.org/fred2/series/GS10 and the data for the interest rate on three-month Treasury bills can be found at https://research.stlouisfed.org/fred2/series/TB3MS. File #obs #var Spread.txt 234 4 Variables GS10: 10-Year Treasury Constant Maturity Rate TB3MS: 3-Month Treasury Bill: Secondary Market Rate S: equal to GS10 minus TB3MS A probit model is used to create forecasts for the probability of a decline in real output. The model uses the R1FF and Spread variables and is described in the article. These series span from 1968q4 to 2012q4. File #obs #var spread_sh01234_ff.txt 178 6 Variables SPREAD_SH0: quarterly forecasts of the probability of a decline in real output in the current quarter SPREAD_SH1: quarterly forecasts of the probability of a decline in real output for one quarter ahead SPREAD_SH2: quarterly forecasts of the probability of a decline in real output for two quarters ahead SPREAD_SH3: quarterly forecasts of the probability of a decline in real output for three quarters ahead SPREAD_SH4: quarterly forecasts of the probability of a decline in real output for four quarters ahead Part C of the article tests the robustness of results by using alternative R1 recession probabilities. The following file contains the data for the real-time probit forecasts. These series span from 1968q4 to 2012q4. File #obs #var spread_sh01234_rt.txt 178 6 Variables RT_SPREAD_SH0_SE: quarterly forecasts of the probability of a decline in real output in the current quarter RT_SPREAD_SH1_SE: quarterly forecasts of the probability of a decline in real output for one quarter ahead RT_SPREAD_SH2_SE: quarterly forecasts of the probability of a decline in real output for two quarters ahead RT_SPREAD_SH3_SE: quarterly forecasts of the probability of a decline in real output for three quarters ahead RT_SPREAD_SH4_SE: quarterly forecasts of the probability of a decline in real output for four quarters ahead Contact information: Croushore: dcrousho [AT] richmond.edu Marsten: katherine.marsten [AT] richmond.edu