Just some notes:
1) both -xtivreg- and -xtivreg2- do allow for more that one endogenous variable (as long as you have one more instrument). Obviously, all endogenous variable must have the same instruments...
2) lagged endogenous and predetermined are terms that sound to me as connected with Arellano-Bond (see -xtabond- and -xtabond2-). But I don't know any panel command to estimate a 5-equation system.
3) how would we identify the panel variable in reg3? you can read more on correcting the standard errors, using cross-sectional commands to manage panel data, etc... at
http://www.kellogg.northwestern.edu/faculty/petersen/htm/papers/se/se_programming.htm
4) I am not an expert but, given your model, it seems easy to reduce it to a two-equation system by hand, given that each of the indipendent variables in the last three equations appears in one equation only.
Nicola
At 02.33 24/02/2007 -0500, Rajiv Sabherwal wrote:
>Hello,
>
>I had earlier sent the message below to statalist. However, a few
>minutes ago, it struck me that the equations I have may not be a
>"simultaneous equation model" at all. This is based partly on Kit
>Baum's message on http://www.stata.com/statalist/archive/2004-03/
>msg01120.html.
>
>In our model (given below), none of the endogeneous variables affect
>each other at that time. All the variables on the right-hand sides of
>all equations are either exogneous of lagged-endogeneous (i.e., pre-
>determined).
>
>Therefore, it would seem that simple panel regressions, rather than
>simultaneous models, would be appropriate. Am I missing something?
>
>Thanks, and best wishes,
>
>Rajiv
>
>On Feb 23, 2007, at 1:25 PM, Rajiv Sabherwal wrote:
>
>> Hello,
>>
>> A PhD student and I are trying to test a model containing
>> simultaneous equations such as the following:
>>
>> y1(t) = a1 + b11*y1(t-1) + b12*y2(t-1) +b13*y3(t-1) + b14*y4(t-1) +
>> b15*y5(t-1) + b16*x1 + b17*x2
>> y2(t) = a2 + b21*y2(t-1) + b23*y3(t-1) + b24*y4(t-1) + b25*y5(t-1)
>> + b26*x1 + b27*x2
>> y3(t) = a3 + b31*x3 + b32*x4 +b33*x5
>> y4(t) = a4 + b41*x6 + b42*x7
>> y5(t) = a5 + b51*x8 + b52*x9
>>
>> We have panel data, with 5000 units observed at 50 periods each,
>> and some time-invariant exogeneous variables characterizing each
>> unit. None of the exogeneous variables vary over time, but all the
>> endogeneous variables do.
>>
>> We believe random effects would be appropriate as we also need to
>> study the effects of the time-invariant exogeneous variables. We
>> are contemplating using reg3 (I couldn't find an appropriate xt
>> command, as xtivreg and xtivreg2 seem to be for a single
>> endogeneous variable). Would reg3 be appropriate for this
>> situation? Also, how would we identify the panel variable in reg3?
>>
>> I have searched the internet and Stata discussion lists, but could
>> not find an answer. I did find the following paper, which uses
>> reg3, but with fixed effects and no time-invariant exogeneous
>> variables.
>>
>> Why Firms Want to Organize Efficiently and What Keeps Them from
>> Doing so: Inappropriate Governance, Performance, and Adaptation in
>> a Deregulated Industry, Jack A. Nickerson; Brian S. Silverman,
>> Administrative Science Quarterly, Vol. 48, No. 3. (Sep., 2003), pp.
>> 433-465.
>>
>> Please advise. Thanks a lot, and best wishes,
*
* 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/