Panel Data Logged Differences Help!!

[Katherine Paradise]

Hello thank you!! I'm doing a research project on whether or not individual CEO campaign contributions affects their 6 month average stock price post election, as compared to their six month average pre election. I have panel data, with multiple observations of pre/post for each company for multiple election years, as seen in the dataex output. I'm trying to run this panel data regression: reghdfe changestockprice totaldem totalrep total_donated towinner, a(companyid year) vce(cluster companyid), but I was informed that it would be a better analysis to make the "change" variable a ln() function to better capture differences across firms. However, sometimes the stock price declined post-election, and the difference is negative. How do I reconcile this with logs? Any and all help is much appreciated!!!


input str8 ticker byte post float(totaldem totalrep companyid changestockprice) long(total_donated towinner tolosers) float year
"AAP" 0 0 1000 1 . 211 1 1 2004
"AAP" 1 0 1000 1 -1.6100006 211 1 1 2004
"AAP" 0 0 1000 1 . 226 1 1 2006
"AAP" 1 0 1000 1 -2.5200005 226 1 1 2006
"AAP" 1 0 1000 1 -9.939999 225 1 1 2008
"AAP" 0 0 1000 1 . 225 1 1 2008
"AAP" 1 0 1000 1 14.23 232 1 1 2010
"AAP" 0 0 1000 1 . 232 1 1 2010
"AAP" 1 0 1000 1 .2099991 235 1 1 2012
"AAP" 0 0 1000 1 . 235 1 1 2012
"AAPL" 0 2300 0 1 . 117 1 1 2002
"AAPL" 1 2300 0 1 -6.25 117 1 1 2002
"AAPL" 0 2300 0 2 . 120 1 1 2004
"AAPL" 1 2300 0 2 -.5 120 1 1 2004
"AAPL" 1 2300 0 2 6.660002 128 1 1 2006
"AAPL" 0 2300 0 2 . 128 1 1 2006
"AAPL" 1 2300 0 2 -5.18 126 1 1 2008
"AAPL" 0 2300 0 2 . 126 1 1 2008
"AAPL" 0 2300 0 2 . 131 1 1 2010
"AAPL" 1 2300 0 2 15.97 131 1 1 2010
"AAPL" 1 2300 0 2 4.1299973 133 1 1 2012
"AAPL" 0 2300 0 2 . 133 1 1 2012
"ABC" 1 3667 10750 2 -5.609997 282 12 35 2002
"ABC" 0 3667 10750 2 . 282 12 35 2002
"ABC" 0 3667 10750 3 . 293 12 35 2004
"ABC" 1 3667 10750 3 5.98 293 12 35 2004
"ABC" 0 3667 10750 3 . 312 12 36 2006
"ABC" 1 3667 10750 3 7.68 312 12 36 2006
"ABC" 0 3667 10750 3 . 315 12 37 2008
"ABC" 1 3667 10750 3 -8.360001 315 12 37 2008
"ABC" 1 3667 10750 3 10.41 323 13 35 2010
"ABC" 0 3667 10750 3 . 323 13 35 2010
"ABC" 0 3667 10750 3 . 326 14 37 2012
"ABC" 1 3667 10750 3 3.369995 326 14 37 2012

[Clyde Schechter]

...but I was informed that it would be a better analysis to make the "change" variable a ln() function...

Either you misunderstood the advice or you were given extremely bad advice.

I imagine that what was intended is that rather than using the change in stock price, you should consider using the logarithm of the ratio of the pre- and post- stock prices. This would seem sensible to me in that it would treat stocks with initially high prices similarly to stocks with initially low prices, and it would also treat a doubling of the stock price as equivalent to a halving of it (but in the opposite direction).