Galvão, Ana Beatriz C., "Structural Break Threshold VAR for Predicting
US Recessions using the Spread", Journal of Applied Econometrics, Vol.
21, No. 4, 2006, pp. 463-487.

All files are ASCII files in DOS format. The data files are zipped in the
file galvao-data.zip, and the program files are zipped in the file
galvao-progs.zip. Unix users should use "unzip -a".

Gauss programs:

  gendgpsp.prg generates data from the DGPs of Table I
  (employed in Figure 1)

  mcestim.prg checks bias and efficiency of proposed estimation
  methods. It generates results of Table 2.

  montecarlo2.prg evaluates the selection procedure with efficiency bounds,
  It generates results of Table 3.

  newestte.prg estimates models for output growth and spread data (Table 4).
  It also computes the bounds presented in Table 5. Data are in data03.csv.

  newfor.prg does the out-of-sample analysis with real time (coefficients of
  Figure 2) and forecasting results of Tables 6 and 7. It also generates 
  in-sample probability forecasts of Figures 3 and 4. Real-Time data are 
  in dataev.csv. Spread is in datar.csv.

Data files: 

data03.csv (first column is the output growth and second is the spread;
	   1953:Q2-2002:Q4). Spread is the difference between 10-year T bond
	   and the 3-month T bill interest rates. They are obtained from
	   www.stls.frb.org/fred.

datar.csv (first column has spread and the other colummns are real output
          vintages since 1986Q1 until 2003Q3),

dataev.csv (equal 1 when in recession (1953:2 to 2003:Q4); first column:
	   recession is two consecutive quarters of negative growth in the
	   next five quarters; second column: turning point rule at t with
	   no censoring). 

Real output growth series in real time are obtained from
www.phil.frb.org/econ/forecast/reaindex.html.