Statistical Errors in Business Decisions: USA World Bank Case

Introduction: The Power and Pitfalls of Statistics

From the development of new psychological treatments to the selection of the President of the United States, statistics have been used throughout history to cast predictions that helped advance the scientific, social, and political climate of the world. In addition to their traditional role in disciplines such as psychology, medicine, and political science, statistics have recently been used to aid the research and development of products and services in the business community. Although statistics can be extraordinarily beneficial across fields from medicine to business, they can also be misleading. By choosing an insufficient sample, allowing data to lapse, or making any number of other errors, companies and researchers can produce data they believe is infallible when it is ultimately incorrect.

In the case study involving the U.S.A. World Bank, Bea pointed out several historical instances when this was the case. The first is the 1936 presidential election, in which Literary Digest predicted that Alf Landon would win by a landslide, using flawed data. When Franklin D. Roosevelt won the election, the error was exposed and the publication was widely embarrassed. The mistake lay in the sample. Although Literary Digest had used a much larger sample than the Gallup Poll, their sample skewed heavily toward more affluent respondents — a demographic that tends to vote Republican — and was therefore not representative of the general population. It is easy to understand how the researchers reached their conclusion, given that they polled the sector most likely to vote for the Republican candidate, but the underlying sample was fatally biased.

Similarly, Bea mentioned the 1948 presidential election, in which most polling companies stopped polling based on the substantial lead Thomas Dewey held over Harry Truman two months before Election Day. According to Bea, however, Truman's aggressive campaign strategies in the final month shifted public opinion in his favor. Because the polls were more than a month old, most polling agencies were working with incorrect data. By stopping their polling prematurely, they committed a time fallacy — drawing conclusions from an outdated dataset.

Although these two examples are among the more famous statistical errors, they are far from unique. Statistics in newspapers, journals, and magazines frequently imply connections that may not exist. One cannot assume that correlation equals causation, nor can one assume that a meaningful correlation exists without scrutinizing how the data were gathered. A more modern example involves the 2008 presidential election. According to a poll published that year, 87% of Fox News viewers were likely to support John McCain ("News You Watch Says a Lot" 2008). While this result effectively identifies which political groups watch which networks — which was the poll's intended purpose — it would be a serious misapplication to infer that 87% of all news viewers would vote Republican. Fox News audiences skew heavily conservative, so using this figure to draw conclusions about the broader news-watching public would constitute a clear sample error.

Brief History of Statistics and Business

Understanding both the power and the limitations of statistics is a necessary prerequisite for using them in business decisions. Fortunately, USA World Bank was given the opportunity to reassess its statistical data and make corrections before committing to a product launch. Through an examination of the history of statistics in business and a description of the problems in USA World Bank's data, one can identify solutions that will allow the bank to launch the product most likely to attract new customers and satisfy existing ones — without alienating either small business owners or individual consumers.

Although its origins date back to the seventeenth century, the identity of the philosopher who first articulated the principles of statistics is uncertain. In 1654, letters between mathematicians Blaise Pascal and Pierre de Fermat introduced the concept that would become probability theory. By 1656, German philosophers and mathematicians had incorporated these emerging statistical ideas into mainstream mathematics. The 18th century saw considerable exploration and publication on probability theory, and the concepts of observation and combination in probability gained wider recognition by the century's end. In the 19th century, the field expanded from simple probability into more complex theories, including the method of least squares, probability errors, quartiles, and regression (Verduin 2007). By the 20th century, statistics had become a mainstream mathematical discipline taught in high school and college classrooms. Mathematicians continued to push the field forward, spending much of the century engaged in debate over the correct methods of statistical practice (Verduin 2007).

The history of statistics in business, however, followed a different path — one driven by competition rather than pure inquiry. According to Arsham, "in a competitive environment, business managers must design quality into products, and into the process of making the products" (2008). That quality is achieved through statistical experiments, which identify and remove obstacles at each stage of development, enabling researchers to build products most likely to satisfy customers (Arsham 2008). As Arsham further notes, "today's good decisions are driven by data" (2007). Rather than making choices based on organizational politics or personal intuition, the integration of statistics into business practice allows managers and executives to rely on empirical evidence.

Description of the Problem at USA World Bank

Nevertheless, Arsham cautions business statisticians to use their findings carefully. Statistics is built on variation: opinions shift, and businesses cannot fully control the market. While statistical data gives managers a more informed picture of their target audience throughout product development, it does not eliminate the possibility of statistical error, nor does it make statistical analysis a foolproof predictor of customer or consumer behavior.

Based on the foregoing discussion of statistical pitfalls and the role of statistics in business, USA World Bank can be identified as a company that recognized the value of statistical research in predicting customer preferences. However, it was not a company that fully understood the foundations of statistics, its inherent risks, or the errors that arise from imprecise methodology — at least not until Bea joined the board of directors. While Bea identified several problems with the company's statistical research, additional flaws exist in the data gathered by both Best Market Research and Jim's independent study. The most significant problems are summarized in the sections below.

Conclusion

"News You Watch Says a Lot About How You'll Vote." (2008). Retrieved October 5, 2008, from

United States Census Bureau. (2008). Retrieved October 5, 2008, from

Verduin, Kees. (2007). A Short History of Probability and Statistics. Retrieved October 5, 2008, from http://www.leidenuniv.nl/fsw/verduin/stathist/stathist.htm

"What's the Best Way to Collect My Information?" (1998). Retrieved October 6, 2008, from the United States Department of Education: