The LSDV (or CV) is inconsistent for a dynamic panel data model with individual e§ects, whether the efects are fixed or random. 

The biais of the LSDV estimator in a dynamix model is generaly

known as dynamic panel bias or Nickellís bias (1981). I suggest you the following classical papers:

Nickell, S. (1981). ìBiases in Dynamic Models with Fixed

Efects,îEconometrica, 49, 1399ñ1416.

Anderson, T.W., and C. Hsiao (1982). ìFormulation and

Estimation of Dynamic Models Using Panel Data,îJournal of

Econometrics, 18, 47ñ82

Best regards.

@Ramiro Esqueda Walle....Thanks for your reply.

To avoid Nickel bias in your case, I suggest using difference or system GMM instead of FE estimator

@Sherif Maher Hassan...Thanks for replying. Now I am using system GMM estimation.

But I am getting insignificant results when variables are transformed into logs. What could be the possible reason?