Dear all viewers,
I have three questions about the use of the clogit command.
My project is about location choice of one group of new firms inside EU15
countries and Norway and Switzerland.Time span of the firm's
establishment is from 1995 to 2003.Choice of each new firms' location depends on
the locational characteristics(like GDP and corporate tax
rate).Following is a list of all explanatory variables I use.
leatr: log of effective aveage corporate tax rate from Devereux and Griffith
(2003).
lmcentrality: log of national GDP adjusted by distances between countries.
loecdlcost1:log of labour cost from OECD.
lunempl:log of unemployment rate from EUROSTAT.
lpatents:log of patent applications submitted to Eurppean Patent Agency.
lutility: log of utility prices(weighted value of light fuel and electricity).
lagglpharma2: log of output of pharmaceutical firms in each country.
leatragglp~2: log of interaction of effective tax rate and national
pharmaceutical products output.
For each 177 new firms, the explanatory variables is one year previous to the
year the firm was established to tackle the issue of
simultaneity.
The questions are:
1. Following the introduction of the clogit in Stata manual version 8, I
generated 15 country dummies(choices,from Belgium to Switzerland
and Austria is reference country) as done by the manual to demostrate
transportation mode choice problem. But when I regress choice on
these 15 country dummies and country characteristcs,none of explanatory
variables are significant! But if I ignore country dummies,
coefficients of explanatory variables become significant and most of them have
expected sign.So could I ignore country dummy?Are they
really necessary in my specification?
2. Because of the data availablity I have to use several different sources to
construct the variable---lagglpharma2---there is a significant time
inconsistence across the year for each country and but data distribution is
consistent for each year across countries.For example the output level in 1998
may be higher than that in 1999(normally should be lower because growth) but
across countries in each year,large country has high output and small country
has low output.........I did my best to make data comparable for different years
and countries by adjusting them with deflator and price index.Will this
inconsistence cause problem?But I think anyway it is not a panel data
model,right?
3.Coefficient of interaction term of effective tax rate and national output
level is 4.669221 with a P-value equals 0.007.Does this suggest firms tend to
locate in high agglomeration-high tax area,which might support the view that
high tax rate guaranties better infrastructure and public administration,which
in turn attracts more business?
I am sorry for this lengthy post and thank you in advance.
Kevin Zhang
Department of Economics in Trinity College Dublin
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
