In sample h-step ahead forecasts, where h>1.

[varun sinha]

Hi,

I am looking to recreate the results of the paper 'Temporal Instability in the Unemployment-Inflation Relationship' by King, Stock and Watson (1995). In this paper, they estimate a VAR(12) using monthly inflation and unemployment data from 1954M1 to 1973M12. They compute 6 month ahead, 12 month ahead and 24 month ahead forecasts for inflation and unemployment for the period 1954M1 to 1973M12. I understand that this can be done by iterating the VAR forward.

I have monthly data from 1957M2 to 1973M12 for unemployment and inflation and I have estimated a VAR(12) including these two variables. The VAR(12) was thus estimated over the period 1958m2 to 1973m12. I ran the code:

var dlcpi lur, lags(1/12)

dlcpi is the inflation rate and lur is the log unemployment rate.
Here is a sample of the dataset I am using:

clear
input float(Time dlcpi lur)
-35   .004687309 1.3609766
-34  .0021560192 1.3083328
-33  .0025093555 1.3609766
-32   .002503157  1.410987
-31  .0039207935  1.458615
-30  .0028419495 1.4350846
-29   .003187418  1.410987
-28  .0014135838 1.4816046
-27            0 1.5040774
-26  .0031728745 1.6292405
-25   .002109766 1.6486586
-24   .005953312  1.757858
-23  .0020928383  1.856298
-22   .005905867 1.9021075
-21   .002421856   2.00148
-20            0   2.00148
-19 -.0010373592 1.9878744
-18 -.0006918907  2.014903
-17    .00172925   2.00148
-16 -.0010373592 1.9600948
-15            0 1.9021075
-14  .0013828278 1.8245493
-13  .0006904602 1.8245493
-12  .0013797283 1.7917595
-11 -.0003447533 1.7749523
-10  -.001034975 1.7227666
 -9  .0003452301 1.6486586
 -8  .0020682812 1.6292405
 -7   .002407551  1.609438
 -6  .0013730526 1.6292405
 -5  .0010287762 1.6486586
 -4   .002395868  1.704748
 -3   .003412962  1.740466
 -2            0  1.757858
 -1  .0020422935  1.667707
  0 -.0013611317 1.6486586
  1  .0013611317  1.568616
end
format %tm Time

I understand that to make one step ahead forecasts for inflation I would use the following code:

predict f_inflation1, equation(dlcpi)

If I wanted to generate 6 month ahead forecasts, eg if I wanted to show what the inflation forecast would be for 1958m8 in 1958m2, would the correct code be:

predict f_inflation6 if Time > tm(1958m7), equation(dlcpi)

For 12 month ahead forecasts, would the correct code be:

predict f_inflation12 if Time > tm(1959m1), equation(dlcpi)

Finally, for 24 month ahead forecasts, would the correct code be:

predict f_inflation24 if Time > tm(1960m1), equation(dlcpi)

Thank you,
Varun