Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Simultaneous equation models

    Dear Statalist Forum users,

    I am estimating the relationship between an organizational policy (X) and an organizational performance outcome (Y).

    I have unbalanced panel data, with observations of X in 2002, 2004, 2008, and 2012 with observations of Y in each year from 2001-2013.
    Currently, I run two separate models: a fixed effects model with X predicting Y in year (t+1) and a random effects model predicting the presence of X from Y in year (t-1). Both models include similar control variables and I have chosen fixed vs. random effects for theoretical reasons.

    Someone suggested that I should perhaps use simultaneous equation models instead of modeling the two separately given that I expect a reciprocal relationship. What Stata command should I investigate further? I'm also unsure what the advantage of doing this would be and whether it makes sense to do this given the odd structure of the dataset. I have tried to omit irrelevant details but am happy to explain further if helpful. I have found a few topics on this but had a hard time applying them to my specific situation. Any advice is, as always, much appreciated.

    -Matt



  • #2
    You didn't get a quick answer. You;ll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

    Searching the subject index for Stata will give you many options for simultaneous equation models. You can use reg3 to do 3SLS, adding i.panel gives fixed effects. You can use one of the SEM estimators. Or, you can do equation by equation using ivreg, ivreg2, xtivreg, xtivreg2, eregress, or similar estimators that do equation by equation estimation. These will give you consistent estimates. Moving to 3SLS may give you more efficiency but at the cost that misspecification in any of the equations can render all the equation estimates inconsistent.

    Comment

    Working...
    X