please share more details about your model and estimation.

please include two estimation functions you used and the estimation results you obtained, so we can better understand the situation you are facing

One assumptions of the regression model is that regressors are fixed in repeated samples. Imagine the football team. In the game they produce one result, in another game another result. The OLS estimator are the same as likelihood estimators, for example. Perhaps with an example, your question will be clearer

Maurício Assuero Lima Freitas The assumption that the regressors are fixed in repeated samples can not be justified in economics or the social sciences or in many other disciplines where the data are observed and are, almost always, not the outcome of a properly designed randomized survey or experiment.

One has a population, variables y and x1, x2, ...,xk, u where

y = b1 x1 + b2 x2 +...+bk xk +u

where y and x1, x2, ...,xk can be observed and u can not be observed.

u has the following properties -

Under these conditions, the OLS estimates are consistent and have optimal properties. See Wooldridge (2010), Econometric Analysis of Cross Section and Panel Data: Second Edition. MIT, or Hayashi (2000), Econometrics, Princeton.

Introductory textbooks often impose stronger conditions than are necessary.

Le Ma Long Sorry this probably does not help. It is not clear what you mean by two different econometric models. For example, if you have a model

y = b1 x1 + b2 x2 + b3 x3 +u

where economic considerations indicate that x3 may be correlated with u. If this is so OLS estimation will produce biased regressions and instrumental variable estimation will produce unbiased results. Estimation by OLS and instrumental variables will produce different results. If the results are different the Instrumental variable approach is appropriate. Instrumental variable estimation is valid regardless of whether there is correlation or not. OLS is valid when there is no correlation.

Similar concerns arise when considering panel data and using fixed effects or random effects.

If you have an economic model and data, you can use econometrics to check that the model is not misspecified. If you find that the data and model are consistent you can then estimate your model and check the strength of any causal relationships in the model. Your results are then conditional on the validity of your model. The econometrics does not verify the truth of the model.

If you have an alternative model which is also data consistent but provides different estimates of the causal effects these revised effects are again conditional on the truth of the second model. In general, econometrics does not provide a way of deciding which is the appropriate model. You must return to your models and use economic theory to evaluate the results of both models.

Yes, you can different results. That means at least one of the model does to fit what your are modeling. You have refer to the economic theory to find the best model. The assumptions behind your models may differ.

I would rather be surprised if you get similar results. The two models may have different assumptions, estimation methods, etc. That’s why researchers keep on developing new estimation methods. So I advise you to check whether model assumptions are satisfied or not since this is crucial for model choice. Also go through the literature and check if the estimation methods yield reliable parameter estimates and standard errors. If all is okay, select the best fit model using model selection criteria or objective tests.

Well noted with many thanks, Sir. Joy Yang Emmanuel Gabreyohannes Nassibou Bassongui John C Frain Maurício Assuero Lima Freitas .Your excellent advice will help me improve more in the study process.