Econometric Issues To Consider and Model Selection for Some Hypothesis Tests

This chapter discusses standard regression assumptions such as uncorrelated residuals, constant variance of the residuals that are independently distributed, normally distributed residuals with zero mean, and the use of stationary variables. Given concerns about violations of standard regression assumptions in Chapter 3 regressions, this chapter introduces equations less likely to violate such assumptions. Rather than use either total real GDP or per capita real GDP in log level form, many if not most of the regression equations in this chapter and subsequent chapters use the first difference in the log of the output variable (or the ‘change in’ the output variable) on the left side of the regression equation and the right side of the regression equation. The goals are to avoid spurious regressions and to reduce if not eliminate serial correlation. Such violations of standard regression assumptions could invalid t-statistic and F-statistic hypothesis tests. This chapter also discusses model selection for many of the hypothesis tests in Chapter 5 .