Lessons and Conclusions

Introduction

This chapter highlights the major findings and limitations of the text. They are concisely stated so that the conclusions are clear and areas of improvement are carefully identified. It is hoped that future researchers might be able to improve on the limitations and experiment with a larger sample of countries/datasets to help enhance the quality of growth empirics, particularly for development policy. The application of alternative empirical methodologies, together with improvement in model specification and tests of robustness should provide a clear-cut guidance for researchers, applied growth economists and policy makers.

Following a brief survey of the growth trends and the two prominent theories of economic growth, the text started with the empirical investigations actually with the analysis of the growth accounts. In fact, growth accounting is the first step into such an inquiry, Barro (1988). The analysis then resorted to Senhadji’s (2000) approach in determining the sources of growth and found that this methodology was a useful development, notwithstanding the need to conduct the tedious growth accounting exercises. Although GAE in itself is not unmanageable, it is the validity of the data on TFP obtained through the GAE that matters. Whilst this can be improved somewhat with the use of improved methods of estimating the TFP growth, (but not totally corrected), Senhadji’s approach better aligns to the theory because it models TFP growth as opposed to annual or some average rate of output growth commonly noted in the conventional growth regressions. This is a huge relief. As an improvement over Senhadji’s approach (to overcome the need for a GAE), Rao and Singh (2007) show a direct way of estimating the impact of the hypothesized factors on TFP or on the long-run growth rate of output. Their measure of the rate of growth is theoretically consistent to the sole determinant of growth in the Solow model. These two approaches imply is that exogenous productivity or economic growth could actually depend on some known forces but its effects can be tested within the Solow framework.

In this book, the growth effects of a limited set of variables are estimated using time series and panel data methods and in alternative chapters, the results of these methodologies are compared on the basis of their empirical performance and consistency to the theory. The obtained results provide support to Rao and Singh’s extension but do not discredit Senhadji’s approach because it also performed well with time series data. While Barro (1990b) method produced equally good results with large panel average datasets, its use with short-period panel average rate of growth is contestable in many of its applications, including what is shown in the analysis in the text. In addition, the well-known method cannot be applied in time-series analysis involving country-specific studies, which seems to now have become more important. To test robustness of these findings, the sensitivity analyses are also conducted. These give further confidence in the use of the new extension and the results obtained from the alternative methods. The robustness tests show equivocal results on the robustness of the growth factors, their estimated growth effects and that of Rao and Singh’s specification.