An examination of specification error in modern United States growth processes

This dissertation involves an empirical reexamination of US growth with the purpose of explaining growth usually attributed to advances in productivity. First, retaining the assumption of exogenous technological progress, I attempt to improve upon existing empirical models through new functional form assumptions. Next, I employ recent models of endogenous growth. Later chapters explore the issues of nonstationarity and international dependence. A significant generalization of the Gumbel Exponential distribution is developed and applied to the statistical modeling of economic growth. My chief objective is to characterize more accurately recent growth experience so that we may determine the most effective policy actions.

Current empirical studies of growth behavior have concentrated on a cross sectional approach. I believe, in addition, much can be learned about individual growth processes through a time series approach. This approach avoids many complicated issues in cross sectional analysis including changes in institutions within and between countries. Better understanding the nature of growth in a particular country and relating this process to other nations should yield valuable insight into the nature of growth, convergence and divergence and provide implications for public policy.

Many empirical studies have downplayed the crucial issue of examining the data in order to find the most appropriate econometric model specification. Through misspecification testing, we can identify and avoid faulty assumptions. Instead of viewing our data set as uncooperative, we should value the rich information our data contain. If our usual specification assumptions are invalid, more information can be extracted from our series through the inclusion of additional variables or through a Maximum Likelihood approach based upon an alternative distribution.

This is the approach I follow in reexamining commonly utilized US input and output series. Utilizing the statistical and graphical abilities of the computer packages GAUSS and MATLAB, I am able to examine both graphically and analytically the validity of various assumptions about the underlying distributions of the data. With this approach, I can show that the Solow Residual contains a great deal of additional information about the dynamic pattern of growth of macroeconomic aggregates.