The Impact Of Measurement Error On Estimation And Inference In The Solor Growth Model

Mankiw,Romer&Weil(1992)tested the validity of both basic and augmented Solow Growth Models.While the basic model did not have human capital accumulation as an explanatory variable,the augmented model did.However,I observe that their measure for human capital accumulation-average percentage of working age population that have enrolled in secondary school education-is not perfect and therefore measured with error.The reason is that MRW data does not consider the variations in the duration of schooling for all countries.As a result,I replicated MRW paper and adjusted for measurement errors using the method proffered by Erickson et al.(2014).This paper finds that after adjusting for measurement error in Basic Solow Model,the coefficient of savings rate is much higher than what MRW obtain for all samples and the coefficient of population growth is more negative than what MRW obtain for all samples.The implied share of physical capital is about twice as large as MRW and is much larger than the share of capital obtained from stylized facts.On the other hand,as opposed to the basic Solow model,adjusting for Measurement Error in Augmented Solow Model led to the rejection of the theoretical restrictions of the Solow model for the Non-oil sample.In contrast,we found the results for the intermediate sample may be robust to some form of measurement error.Moreover,the implied shares of physical and human capital for the intermediate sample do not differ dramatically from the original study.In both cases of Measurement Error adjustment,the Sargan-Hansen test indicates that there is no evidence that the conditions for the adjustment do not hold for Penn World Table(PWT)version 4.0.This points the need to re-examine the results obtained by other authors using the non-oil sample for growth regressions.