IV regressions - modifying dates

[Ama Perera]

Hi,

I am working on a panel data set with 20 countries and quarterly data. The dependent variable is price and the independent variable is risk. I am using quarterly election as an IV in IV regressions. This is a dummy variable which takes the value of 1 for election quarters and 0 otherwise. I want to run instrumental variable regressions;

01. by using three quarters before plus election quarter as the IV
02. quarter right before the election quarter as the IV

A snapshot of the data file is included below.

The current code that I use to run regressions using current election quarter data is;

ivreg2 price $controls i.countrynum i.yearqtr (risk = election), savefirst savefprefix(first)

Can someone please advise on how to modify this code to run the above 2 regressions?

* Example generated by -dataex-. To install: ssc install dataex
clear
input str97 location double(year qdate price risk) byte election
"Australia" 1970 41    .04297956780566525 .3646308 0
"Australia" 1970 42    .05162054093048929  .655308 0
"Australia" 1970 43    .04565936262218839        0 0
"Australia" 1971 44    .04049916884918492  .337344 0
"Australia" 1971 45   .049496729488366364  .449169 0
"Australia" 1971 46    .07607787494393325  .304909 0
"Australia" 1971 47    .07383885597122553        0 0
"Australia" 1972 48     .0903276229663641  .373925 0
"Australia" 1972 49     .1292809534922108        0 0
"Australia" 1972 50    .11936403114460226 .2447681 0
"Australia" 1972 51    .14367467361375175 1.230959 1
"Australia" 1973 52    .11256845289860817  .354045 0
"Australia" 1973 53   .023396646749443528  .226142 0
"Australia" 1973 54  -.022503139633368807        0 0
"Australia" 1973 55   -.07171322256754664        0 0
"Australia" 1974 56  -.059487461301409694 .4249894 0
"Australia" 1974 57 -.0009579086563914485 .4659109 1
"Australia" 1974 58   -.01590767621940592        0 0
"Australia" 1974 59   .012655398230987824  .474046 0
"Australia" 1975 60   .004346961030980312 .3881988 0
end
format %tq qdate

Thanks a lot.

[Joro Kolev]

You need to -xtset- your data, and then lead (F.) and lag (L.) operators become available.

[Ama Perera]

You need to -xtset- your data, and then lead (F.) and lag (L.) operators become available.

Thanks for your response.

so three quarters before election quarter would be:
ivreg2 price $controls i.countrynum i.yearqtr (risk = l3.election), savefirst savefprefix(first)?

Thanks.

[Joro Kolev]

Thanks for your response.

so three quarters before election quarter would be:
ivreg2 price $controls i.countrynum i.yearqtr (risk = l3.election), savefirst savefprefix(first)?

Thanks.

Yes. Something like:

xtset country qdate, quarterly

ivreg2 price $controls i.countrynum i.yearqtr (risk = l3.election), savefirst savefprefix(first)

This will include only the said quarter. You can also include all quarters up to certain lag with something like

ivreg2 price $controls i.countrynum i.yearqtr (risk = l(1/3).election), savefirst savefprefix(first)

see -help tsvarlist-.

[Ama Perera]

Yes. Something like:

xtset country qdate, quarterly

ivreg2 price $controls i.countrynum i.yearqtr (risk = l3.election), savefirst savefprefix(first)

This will include only the said quarter. You can also include all quarters up to certain lag with something like

ivreg2 price $controls i.countrynum i.yearqtr (risk = l(1/3).election), savefirst savefprefix(first)

see -help tsvarlist-.

Thanks a lot.

can you please let me know how to create an interaction term between all quarters upto third lag with the election variable?

I tried gen new_var = l(1/3).election*election
but it doesn't work.

[Joro Kolev]

Try directly using the time series operators in conjunction with interactions terms, like this

 
 ivreg2 price $controls i.countrynum i.yearqtr (risk = l(1/3)#election), savefirst savefprefix(first)

[Joro Kolev]

If Stata does not understand the syntax, you need to use the -generate- statements lag by lag

E.g.,

gen l3electionXelection = l3.election*election
gen l2electionXelection = l2.election*election

etc.