AR(1) and AR(2) tests, xtabond2

[Maria Reyes Retana]

Hello, I am currently working on my thesis and I have a question maybe you can help me with.
I am using the xtabond2 command and the p values of my AR(1) and AR(2) tests are insignificant, I understand that that is not good because the insignificance of AR(1) could mean that the model is missespecified in the other hand AR(2) I think is fine. My results of the two tests are the following:
Arellano-Bond test for AR(1) in first differences: z = 0.65 Pr > z = 0.517
Arellano-Bond test for AR(2) in first differences: z = -0.69 Pr > z = 0.492

My question is if my interpretation is correct because I saw this post (link below) with the same case in wich Roodman replied The "AR()" (really, serial correlation) tests are showing no sign of first-order serial correlation, which is good". I want to know what you think and if I could report those coefficients of AR(1) and AR(2)
https://www.statalist.org/forums/for...15397-xtabond2

Thank you in advance for you answers
María Reyes Retana

[Sebastian Kripfganz]

Just looking at the AR(2) result and ignoring the AR(1) test is dangerous. Here is why: Suppose the error term follows a random walk, then it is serially uncorrelated in first differences for every order. However, a random-walk error term is the worst thing you can image because your lagged variables remain invalid instruments no matter how far you lag them in time.

Roodman's recommendation to include time dummies might help if you have not done so already.

[Maria Reyes Retana]

Sebastian,
Thank you for your advice!