The p-value is a statistical term used in medical research to suggests whether a result might be statistically significant or not. But despite the p-value being an "objective" measure of statistical probability, it has introduced a new concern in research, namely p-hacking. P-hacking is a type of bias where researchers collect or select data until nonsignificant results become significant. It also happens when researchers try a multitude of statistical analyses and cherry-pick those with significant results. Researchers do this because there is a strong incentive to publish significant results, since evidence suggests that journals prefer to publish significant results over non-significant ones. Chapter 2 is thus concerned with how p-hacking and other types of statistical hacking can negatively affect medical science.

Research in consumer research has rightfully been criticized for p-hacking, hypothesizing after the results are known, and other practices that lead to overestimation of the reliability and replicability of published results. Remediation has centered on more closely approximating the ideal hypothetico-deductive (i.e., confirmatory) method. There has been a push toward forming, and registering, selective hypotheses before running experiments, testing only those hypotheses, and testing each hypothesis with a single, preplanned analysis. We argue that doing better confirmatory experiments is not the (whole) solution and that HARKing and running multiple analyses are not the problem per se. The problem is that we misrepresent exploratory research as confirmatory. Forcing exploratory research into a hypothetico-deductive straitjacket leads to bad hypothesis testing. The straitjacket also leads to bad exploration, crowding out essential, good exploration that deserves space in our journals. We propose a recipe for more honest consumer research, in which authors report exploratory studies meant to generate hypotheses followed by truly confirmatory studies that test those hypotheses.

Publication bias and p-hacking are threats to the scientific credibility of experiments. If positive results are more likely to be published than null results conditional on the quality of the study design, then effect sizes in meta-analyses will be inflated and false positives will be more likely. Publication bias also has other corrosive effects as it creates incentives to engage in questionable research practices such as p-hacking. How can these issues be addressed such that the credibility of experiments is improved in political science? This chapter discusses seven specific solutions, which can be enforced by both formal institutions and informal norms.