Article Preview
TopIntroduction
Evidence from several studies in different fields (e.g., education, psychology, biomedical etc.) revealed that articles reporting a favorable or statistically significant relation/difference were published more than were studies of similar quality that show negative or “no-difference” results (Dickersin, 1993; Emerson et al., 2010). Negative results refer to finding nothing of statistical significance or casual consequence. Publishing statistically significant findings was found to be 3.4 times higher in studies applying behavioral and social methodologies (Fanelli, 2012). The tendency to publish research articles “that appears to support a hypothesis and to refrain from publishing findings that do not, thereby creating opinions about the truth of the hypothesis that may be unduly optimistic” is called “publication bias” (Medical Dictionary, 2009). This is a prevalent phenomenon “arising from bias in submitting, reviewing, accepting and publishing scientific results” (Moller & Jenions, 2002, p. 580). Since its first recognition by Sterling in 1959, publication bias has been named differently (e.g., file drawer-effect, positive outcome-bias, etc). While the names differ, they all convey bias in dissemination of knowledge (Emerson et al., 2010).
A bias towards the positive and avoidance of negative outcomes may have serious implications on the integrity of knowledge and advancement of research for a number of reasons. Firstly, we as researchers tend to construct hypothesis based on the analysis of the available literature. The presence of publication bias inevitably diminishes the credibility of hypothesis testing because it is based on biased information. Secondly, publication bias has effects on meta analyses and systematic reviews, which are the propellers of scientific progress. If meta analyes include only the published studies with significant outcomes, then outcomes of meta analyses are likely to be flawed and misleading since negative data are not taken into account (Rothstein, Sutton, & Borenstein, 2005). Last but not least, if we are doing scientific research to learn and explain the truth, then we must not ignore the realities of life. If “no effect/relation” is a frequently transpiring fact of life, then almost always finding and reporting positive outcomes contradicts with reality.
Several causes, relating to authors, editors, reviewers may account for publication bias. Among other things, we have reasons to believe that current practices, particularly in relation to null hypothesis testing in which “p value” has been taken as the only tool to verify accept/reject decisions could be responsible. Misuses of the p-value and consequently falsely rejecting a null hypothesis at a face value of “p” still appears to be a frequent practice (Cashen & Geiger, 2004; Deng, 2005; Egger et al., 2000; Mone et al., 1996; Wilkinson et al., 1999). There is an increasing concern that in modern research false findings may be the majority (Yüksel, 2017). Surprisingly however publication bias and its reasons have received little attention in tourism, travel and hospitality literature. Based on limited research on the topic, the present review primarily aims to examine whether the share of the published studies reporting significant outcomes is larger than studies reporting insignificant outcomes (i.e., failing to reject the null hypothesis). The second objective is to explore whether researchers in the TTH field follow statistical recommendations and refer to the suggested statistical power levels and effect sizes in order to correctly reject/support a null hypothesis.