Prediction after estimation empty due to missings.

[Hannah Pareik]

Hello!

I have time series survey data and want to analyze the effects of a policy that was introduced late 2000.
The pre period is hence from 1996-early 2000, the post period omits late 2000 and runs from 2001-today.
I then want to compare the effects of the 2008 financial crisis to the 2000 stock market bubble, relative to the pre-policy period.

I hence want to use the four years after 2007 (2008-2011) to predict 2001-2004.
I then want to compare the 2001-2004 prediction to the actual responses.

The difference should then be driven by the policy.

This is the dummy crisis08:

gen crisis08 = 0
replace crisis08 = 1 if year > 2007 & year < 2012
replace crisis08 = . if year > 2011
replace crisis08 = . if year > 2000 & year < 2008
replace crisis08 = . if dateq == yq(2000, 4)

The estimation should hence only compare 1996-2000 with 2008-2011. The years in between are treated as missings.

reg dv crisis08 age male urban educ married working, vce(robust)
predict pred_values if inrange(year, 2001, 2004)
gen diff = dv - pred_values if inrange(year, 2001, 2004)
summarize diff if inrange(year, 2001, 2004)

After running this code "pred_values" is empty, as in the estimation crisis08 omits the years 2001-2004.

I tried to fix it like this:

gen post08 = 0
replace post08 = 1 if year > 2007

and then:

reg dv post08 $c if !inrange(year, 2001, 2007), vce(robust)
predict pred_values if inrange(year, 2001, 2004)
gen diff = dv - pred_values if inrange(year, 2001, 2004)
summarize diff if inrange(year, 2001, 2004)

Is this the correct approach?

Thank you!