Fooled by Pseudo-Randomness

Abstract

Random numbers are important to conducting surveys, simulations, and other economic and statistical analysis. Despite their importance, many people using popular statistical software packages to generate random numbers are often uninformed about how the software creates a (supposedly) random number. This lack of knowledge can inadvertently lead the user to use “random” numbers that fail important tests for randomness and are therefore not really random. Unbeknownst to many users, the random number generators in some commercial software programs are seriously flawed; they simply fail to generate numbers that are randomly drawn, resulting in serious unintended consequences for the study being performed. In this paper, we introduce the concept of random numbers in layman’s terms. We describe a number of different methods for drawing random numbers in several popular statistical software packages frequently used by economists. We then discuss errors commonly committed by economists and analysts unfamiliar with the algorithms used by their software package of choice. Next we describe flaws contained in the random number functions in some popular software packages. Finally, we discuss better random number functions also found in commonly used software.