Direct comparison
P-hacking vs HARKing: Key Differences
P-hacking manipulates analysis until p<0.05; HARKing rewrites hypotheses after seeing results. Compare both and how pre-registration defends against each.
Written and maintained by CASRAI Editorial Board
Last updated
Ask CASRAI · free to try
Ask about P-hacking vs HARKing: Key Differences
Ask your first 2 questions free below. Subscribers get 150 a day for $29 a month.
An AI assistant specialized in research administration. It cites the sources behind every answer, labels web answers and says when it can't answer.
Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.
Works on this site and inside Claude, Cursor and the AI tools you already use.
Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays free to read, with no account and no card.
How do P-hacking, HARKing compare side by side?
The table below compares P-hacking, HARKing across 9 procurement-relevant dimensions, from what it manipulates through can co-occur?.
Side-by-side comparison
| Dimension | P-hacking | HARKing |
|---|---|---|
| What it manipulates | The data analysis - which tests are run, when data collection stops, which exclusions are applied | The stated hypothesis - rewriting it after the results are known |
| What stays fixed | The hypothesis | The analysis that was actually run |
| Origin of the term | Popularized by Simmons, Nelson & Simonsohn, 'False-Positive Psychology' (2011) | Coined by Norbert Kerr, 'HARKing: Hypothesizing After the Results are Known', Personality and Social Psychology Review (1998) |
| Typical mechanism | Multiple comparisons, optional stopping, selective covariate inclusion, outlier exclusion | Post-hoc hypothesis presented as a priori in the introduction/discussion |
| Where it happens in the paper | Methods/results - the analysis pipeline itself | Introduction/discussion - the narrative framing |
| Detectable via | Excess of p-values just below .05 across the literature (p-curve analysis); discrepancy between pre-registration and reported analysis | Discrepancy between a pre-registered hypothesis and the published introduction; implausibly precise 'predicted' post-hoc findings |
| Primary structural defense | Statistical Analysis Plan (SAP) / pre-specified analysis, Registered Reports | Pre-registration of hypotheses, Registered Reports |
| Classified as | Questionable research practice (QRP), not research misconduct | Questionable research practice (QRP), not research misconduct |
| Can co-occur? | Yes - a p-hacked result is often also HARKed into the introduction as a predicted finding | Yes - see above |
Common questions
Common questions about P-hacking vs HARKing
Can a single study involve both p-hacking and HARKing?
+
Yes. A researcher may run multiple undisclosed analyses until one is significant (p-hacking), then write that specific comparison into the paper's introduction as the original hypothesis (HARKing). The two are frequently discussed together for exactly this reason.
Is HARKing always dishonest?
+
It is widely treated as a questionable research practice because it misrepresents the confirmatory/exploratory status of a finding, even when the underlying data and analysis are reported accurately. The problem is the false impression of a priori prediction, not fabricated results.
Does pre-registration eliminate both problems completely?
+
It substantially reduces both by creating a timestamped, checkable record, but it is not self-enforcing - a pre-registration can be ignored or a study can be run without pre-registering an exploratory arm. Registered Reports, which add independent peer review of the pre-registered protocol, are generally considered a stronger defense than self-filed pre-registration alone.
Is exploratory analysis itself a problem?
+
No. Exploratory analysis is a legitimate part of research. The issue is disclosure - exploratory findings should be labelled as exploratory, not presented as confirmatory tests of a hypothesis that was actually formed afterward.
