To create shock variable

[Raphael George]

Hi how do I create a shock variable? Suppose I want to create a variable for each year ( for this variable- vc_body_out)which is an indicator of that variable being above or below a threshold of the standard deviation for all the years considered. Kind of similar to how rainfall shocks are created (sd from a historical mean).
[/CODE]

input int year byte(State_code_census Dist_code_census) float vc_body_out
2005 1 1 211
2006 1 1 187
2007 1 1 186
2008 1 1 122
2009 1 1 116
2010 1 1 148
2011 1 1 136
2012 1 1 135
2013 1 1 128
2014 1 1 117
2015 1 1 126
2016 1 1 118
2017 1 1 139
2018 1 1 125
2005 1 2 284
2006 1 2 292
2007 1 2 231
2008 1 2 172
2009 1 2 223
2010 1 2 286
2011 1 2 268
2012 1 2 258
2013 1 2 258
2014 1 2 214
2015 1 2 223
2016 1 2 160
2017 1 2 190
2018 1 2 236
2005 1 3 361
2006 1 3 336
2007 1 3 233
2008 1 3 205
2009 1 3 239
2010 1 3 291
2011 1 3 236
2012 1 3 204
2013 1 3 190
2014 1 3 150
2015 1 3 193
2016 1 3 136
2017 1 3 196
2018 1 3 234
2005 1 4 122
2006 1 4 100
2007 1 4  96
2008 1 4 100
2009 1 4 109
2010 1 4  98
2011 1 4 119
2012 1 4  93
2013 1 4 102
2014 1 4  87
2015 1 4  95
2016 1 4 133
2017 1 4  83
2018 1 4  96
2005 1 5 197
2006 1 5 178
2007 1 5 156
2008 1 5  71
2009 1 5  63
2010 1 5  70
2011 1 5  94
2012 1 5  66
2013 1 5  70
2014 1 5  73
2015 1 5  80
2016 1 5  84
2017 1 5 104
2018 1 5 134
2005 1 6 264
2006 1 6 239
2007 1 6 182
2008 1 6 196
2009 1 6 247
2010 1 6 236
2011 1 6 254
2012 1 6 277
2013 1 6 268
2014 1 6 227
2015 1 6 224
2016 1 6 286
2017 1 6 204
2018 1 6 236
2005 1 7   6
2006 1 7  13
2007 1 7  10
2008 1 7   6
2009 1 7   6
2010 1 7   6
2011 1 7   2
2012 1 7   5
2013 1 7   1
2014 1 7   2
2015 1 7   5
2016 1 7   5
2017 1 7   9
2018 1 7   6
2005 1 8   2
2006 1 8   5
end

[/CODE]

[Luca Calianno]

Hi Raphael,
I hope I understood what you need clearly. The code below should give you the variable you are interested into, change sd_threshold to set the threshold of the standard deviation you prefer.
The code does not consider the fact that your data is longitudinal and not a time series.

local sd_threshold = 1 // change this to have a different threshold of your choice

qui summ vc_body_out
local mean = r(mean)
local sd = r(sd)

local low_threshold = r(mean) - r(sd)*`sd_threshold'
local high_threshold = r(mean) + r(sd)*`sd_threshold' 


gen indicator = (vc_body_out > `high_threshold' | vc_body_out < `low_threshold')

[Raphael George]

Thank you very much. I wanted this indicator based on all the years in a particular state and district. What I mean is, the indicator should tell me whether for a particular year the value of vc_body_out, is greater or lesser than the threshold by considering all the values of vc_body_out in the time period considered in that specific locality (a district in a particular state)
I suppose I have to bysort it by state district. Is the following code correct?

 bys State_code_census Dist_code_census: qui summ vc_body_out local mean = r(mean) local sd = r(sd)

[Luca Calianno]

Now I got it. Sorry for the confusion. I would slightly change the code to make it simpler to understand but you can also create a loop and inside use bysort + if conditions.
I hope this solves your question.

local sd_threshold = 1 // change this to have a different threshold of your choice

bysort Dist_code_census: egen mean = mean(vc_body_out)
bysort Dist_code_census: egen sd = sd(vc_body_out)

gen low_threshold = mean - sd*`sd_threshold'
gen high_threshold =mean + sd*`sd_threshold' 


gen indicator = (vc_body_out > high_threshold | vc_body_out < low_threshold)

[Raphael George]

Thank you so much.
I just have one additional question if that's alright.
A) How am to I create lagged and lead variables for this shock indicator?
A1) Suppose its the year 2011. I want an indicator (in the row of year==2011) to show if there was a shock in

vc_body_out

constructed as earlier (sd from a historical mean) in the year 2010.

A2)Similarily I want to run a falsification excercise to see if future years shocks have an impact on my outcome. So I want to construct an indictaor (in the row of year 2011) to show if there was a shock in

vc_body_out

constructed as earlier (sd from a historical mean) in the year 2012. My codes are this , but when I br to see if the lagged and lead is showing up its not right