Stock returns and investors' mood: Good day sunshine or spurious correlation?

Abstract

This paper examines the validity of statistical significance reported in the seminal studies of the weather effect on stock return. It is found that their research design is statistically flawed and seriously biased against the null hypothesis of no effect. This, coupled with the test statistics inflated by massive sample sizes, strongly suggests that the statistical significance is spurious as an outcome of Type I error. The alternatives to the p-value criterion for statistical significance soundly support the null hypothesis of no weather effect. As an application, the effect of daily sunspot numbers on stock return is examined. Under the same research design as that of a seminal study, the number of sunspots is found to be highly statistically significant although its economic impact on stock return is negligible.

Significance Testing in Accounting Research: A Critical Evaluation Based on Evidence

“…First, we recommend that accounting researchers understand the statistical implications of large or massive sample sizes for their hypothesis testing. It certainly improves the power but the research strategy of maximizing the power of the test (prevalent in modern accounting and finance research studies) is statistically flawed and may lead to spurious statistical significance if the conventional level of significance is maintained (see, e.g., Kim, ). As Dyckman and Zeff () argue, a clear justification should be given for the selection of the sample range, and there should be a ‘stopping rule’ in data collection.…”

“…As Dyckman and Zeff () argue, a clear justification should be given for the selection of the sample range, and there should be a ‘stopping rule’ in data collection. The sample size required for hypothesis testing, which ensures a desired level of statistical power (1– β ) at the chosen level of significance ( α ), can be calculated under the assumption of normality (see, e.g., Kim, ; Kelley and Maxwell, ).…”

“…It does not mean that a researcher should collect and use as many data points as available to achieve a more accurate research outcome. In fact, such a research design is statistically flawed and can produce a spurious statistical significance (see Kim, ). Recently, Harford () and Hand () also raise serious concerns that big data analysis can provide misleading predictions.…”

“…See Kim () for a more detailed analysis as to why this research design can provide spurious statistical significance.…”

Tackling False Positives in Business Research: A Statistical Toolbox With Applications

“…The latter is particularly problematic for business studies, where large or massive samples are routinely adopted. In this case, the power is often practically one and the probability of Type II error is 0, rendering the statistical test severely biased towards Type I error if the value of α is set at a conventional value such as 0.05, with the consequence of frequently rejecting true null hypotheses (see, for example, Kim, ).…”