Symmetrically Normalized Instrumental-Variable Estimation Using Panel Data

Abstract

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __In this paper we discuss the estimation of panel data models with sequential moment restrictions using symmetrically normalized GMM estimators. These estimators are asymptotically equivalent to standard GMM but are invariant to normalization and tend to have a smaller finite sample bias.They also have a very different behaviour compared to standard GMM when the instruments are poor. We study the properties of SN-GMM estimators in relation to GMM, minimum distance and pseudo maximum likelihood estimators for various versions of the AR(1) model with individual effects by mean of simulations. The emphasis is not in assessing the value of enforcing particular restrictions in the model; rather, we wish to evaluate the effects in small samples of using alternative estimating criteria that produce asymptotically equivalent estimators for fixed T and large N. Finally, as an empírical illustration, we estimate by SN-GMM employment and wage equations using panels of UK and Spanish firms.

Inequality and growth in advanced economies: an empirical investigation

“…Similar results are found with the quintiles. The results in columns (2)(3)(4)(5) show that a higher share of education attained by the majority of the population as well as a lower concentration of education in the top quintile led to a more beneficial effect on the growth rates of the developing countries. In contrast, the greater concentration of education among the elite favored the per capita income growth rates in the non-European high-income OECD economies; when the dummy for the advanced countries is equal to one and that for Europe is equal to zero, the estimated effect of the 5 th quintile is 0.025.…”

“…Results show that the negative influence of human capital inequality on per capita income growth rates, reported by the Gini coefficient in column (1), is also found with measures of the education distribution by quintiles. Columns (2)(3)(4)(5) show that whereas a greater share of education in the hands of the middleincome group of the population ( 3 rd quintile) had a beneficial effect on growth, a greater concentration of education in the upper part of the distribution (top 20%) discouraged growth. These effects are stronger during the period 1965-1985 than during the period 1985-2005.…”

“…Furthermore, the role of human capital inequality in economic growth is part of most of the models that analyze the effect of inequality on growth under imperfect credit markets (e.g., [29,39]). 4 Methodologically, this paper uses the system GMM estimator to control for country-specific effects. The reason is that the traditional first difference GMM estimator used by Forbes [27] may not be appropriate when variables are highly persistent, as is the case with income and education inequality measures.…”

“…Thus, they do not capture the fact that a year of French and a year of Kenyan education is not the same 4. A comprehensive empirical analysis of the channels through which human capital inequality may influence human capital accumulation and growth rates is examined by Castello-Climent[18].…”

International integration and the determinants of regional development in China

“…In constructing our data set, we have used new income and investment data reported by Hsueh and Li (1999) Chow (2006) point out that Chinese official statistics are by and large reliable because of the assigned responsibility of the officials preparing them, of their being used in government decision making that is open to public scrutiny, and in many published articles in referred journals. Rawski (2001) argues that two decades of reform have produced impressive achievements in the realm of economic statistics the range, depth, and quality of statistical documentation 21 Monte Carlo results on the finite sample properties of the GMM estimator for dynamic panel data models have been reported by Arellano and Bond (1991), Kiviet (1995), Ziliak (1997), Blundell and Bond (1998) and Alonso-Borrego and Arellano (1999), amongst others. 22 The choice of the period makes sense for two reasons.…”