Weird xtabond2 results for ynamic panel model

[wesso dadoyan]

Hi,

I have run a panel regression model usingxtabond2 by basically using 1 and 2 lags for the firm specific variables (using level and difference) . the regression is as follows and the ourput contained many omitted coefficients.I am wondering what went wrong. The regression is as follows

xtabond2 wcapqw    epu zscorew    dividendw    financialcostw    growthw    profitabilityw    cashfloww    sizew    oiadpqw    revtqw    interestcoveragew    leveragew    reqw gdpg i.q_date,gmm(wcapqw,eq(diff) collapse lag(1 1)) gmm(wcapqw,eq(level) collapse lag(1 1)) gmm(wcapqw,eq(diff) collapse lag(2 2)) gmm(wcapqw,eq(level) collapse lag(2 2)) gmm(zscore,eq(diff) collapse lag(1 1)) gmm(zscore,eq(level) collapse lag(1 1)) gmm(zscore,eq(diff) collapse lag(2 2)) gmm(zscore,eq(level) collapse lag(2 2)) gmm(dividendw,eq(diff) collapse lag(1 1)) gmm(dividendw,eq(level) collapse lag(1 1)) gmm(dividendw,eq(diff) collapse lag(2 2)) gmm(dividendw,eq(level) collapse lag(2 2)) gmm(financialcostw,eq(diff) collapse lag(1 1)) gmm(financialcostw,eq(level) collapse lag(1 1)) gmm(financialcostw,eq(diff) collapse lag(2 2)) gmm(financialcostw,eq(level) collapse lag(2 2)) gmm(growthw,eq(diff) collapse lag(1 1)) gmm(growthw,eq(level) collapse lag(1 1)) gmm(growthw,eq(diff) collapse lag(2 2)) gmm(growthw,eq(level) collapse lag(2 2)) gmm(profitabilityw,eq(diff) collapse lag(1 1)) gmm(profitabilityw,eq(level) collapse lag(1 1)) gmm(profitabilityw,eq(diff) collapse lag(2 2)) gmm(profitabilityw,eq(level) collapse lag(2 2)) gmm(cashfloww,eq(diff) collapse lag(1 1)) gmm(cashfloww,eq(level) collapse lag(1 1)) gmm(cashfloww,eq(diff) collapse lag(2 2)) gmm(cashfloww,eq(level) collapse lag(2 2)) gmm(sizew,eq(diff) collapse lag(1 1)) gmm(sizew,eq(level) collapse lag(1 1)) gmm(sizew,eq(diff) collapse lag(2 2)) gmm(sizew,eq(level) collapse lag(2 2)) gmm(oiadpqw,eq(diff) collapse lag(1 1)) gmm(oiadpqw,eq(level) collapse lag(1 1)) gmm(oiadpqw,eq(diff) collapse lag(2 2)) gmm(oiadpqw,eq(level) collapse lag(2 2)) gmm(revtq,eq(diff) collapse lag(1 1)) gmm(revtq,eq(level) collapse lag(1 1)) gmm(revtq,eq(diff) collapse lag(2 2)) gmm(revtq,eq(level) collapse lag(2 2))gmm(interestcoveragew,eq(diff) collapse lag(1 1)) gmm(interestcoveragew,eq(level) collapse lag(1 1)) gmm(interestcoveragew,eq(diff) collapse lag(2 2)) gmm(interestcoveragew,eq(level) collapse lag(2 2)) gmm(leveragew,eq(diff) collapse lag(1 1)) gmm(size,eq(level) collapse lag(1 1)) gmm(size,eq(diff) collapse lag(2 2)) gmm(size,eq(level) collapse lag(2 2)) gmm(reqw,eq(diff) collapse lag(1 1)) gmm(reqw,eq(level) collapse lag(1 1)) gmm(reqw,eq(diff) collapse lag(2 2)) gmm(reqw,eq(level) collapse lag(2 2)) iv(epu,eq(level)) iv(gdpg,eq(level)) iv(i.q_date,eq(level))twostep robust

The relevant output is as follows:

I am wondering how to proceed from here given the omitted coefficients and the extremely high Hansen test.

[Phil Bromiley]

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. Screen shots are not encouraged. Also, instead of running you command in one line, break it into multiple lines so we can read it. It is almost unreadable as you presented it. You might also get more of a response if you use xtabond rather than the user written version.

I don't use xtabond. However, with all these omitted parameters, I'd wonder if you don't have a colinearity problem. What happens if you use xtreg or xtivreg instead? It also seems odd to me to have all those collapse statements, but that could just be my lack of understanding of xtabond2. Often, it is best to start simple and build up.