LogicalMANIPULATIVE

Selective Baseline

What it is

Choosing the reference point against which a change is measured — the comparison year, the index date, the “before” of a before-and-after — after seeing which one makes the change look largest, smallest or reversed.

How it works

Every percentage change is a change from something, and the something is a choice. Start from a trough and any recovery is a boom; start from a peak and any normalization is a collapse; index two series to a date on which one was unusually low and it “outperforms” from then on. Worked: US gasoline averaged roughly 2.40 dollars a gallon in January 2021, peaked at about 5.02 in June 2022 and stood near 3.10 at the end of 2023; “up 30 percent since the inauguration” and “down nearly 40 percent from the peak” are both true of the same pump. The lever is that the baseline is presented as natural — pre-pandemic, since the election, since the law passed, last year — and the audience never learns that the natural-sounding date was picked last, after the alternatives were tried. The device differs from the cherry-picked window, which sets the visible span; the baseline sets the zero of the comparison, and it is most powerful when the chosen point was an extreme, since regression to the mean then guarantees a dramatic change in the wanted direction. Honest practice fixes the baseline in advance for a stated reason and shows the level, not just the change.

Real-world examples

  • •Through 2022 and 2024 gasoline prices were described by Republicans as up sharply since January 2021, a month of pandemic-depressed demand, and by Democrats as down sharply from the June 2022 peak; each baseline was chosen for what it did to the number.
  • •US crime claims in the 2024 campaign measured from 2019 (“up since before the pandemic”) or from the 2020-2021 peak (“down sharply”), and the same speaker could use both baselines for different offenses in one speech.
  • •Fund marketing that indexes performance to March 2009, the bottom of the financial crisis, shows spectacular returns for almost any equity fund; the same fund indexed to October 2007 shows something else.
  • •Before-and-after advertising in weight loss and skincare uses a “before” taken at the client's worst moment, so that ordinary regression toward their usual state is presented as the product's effect.
  • •Retail and travel companies reporting “record growth” in 2021 measured against 2020, a year of lockdowns; growth measured against 2019 was often flat or negative, and both figures appeared in the same earnings season depending on the company's story.

Ethical guidelines

  • ●State the baseline, say why it was chosen, and choose it before you know the result; “pre-pandemic” or “full cycle” are reasons, “it looks best” is not.
  • ●Show the result against at least one other defensible baseline, and show the level of the series so the reader can see where the baseline sits on it.
  • ●Do not use an extreme — a peak, a trough, a lockdown year — as a baseline without saying it was extreme.
  • ●Use the same baseline for your own record and your opponent's.

How to defend against it

  • ►Ask “compared to when?” and “why then?” for every percentage change; if the baseline was a peak or a trough, ask for the change from a normal year.
  • ►Ask for the level, not just the change: the price now, the rate now, the count now, next to the same number five and ten years ago.
  • ►Try one alternative baseline yourself — a year earlier, a year later — and see whether the claim survives; a real change survives, a manufactured one flips.
  • ►Notice baseline-switching within one argument, where one number is measured from 2019 and the next from 2021; that is two selections, not one analysis.
  • ►For before-and-after claims, ask when the “before” was taken and whether the person or the series was at an extreme at that moment.

From the Defense Playbook

Ask for Absolute Numbersseconds

When someone quotes a relative change ("cuts your risk by 50 percent", "doubles the danger"), ask what the risk was before and after in plain counts out of the same number of people, because the same fact sounds enormous as a ratio and modest as a difference.

Check the Dateseconds

Before reacting to a story, photo, or video, find out when it was first published or recorded, because genuine but old material recirculated as if it were new is one of the cheapest and most common forms of misinformation.

Fallacy Audit of Your Own Caseminutes

Before you publish, present, or press an argument, go through it as a hostile reader would, looking for the fallacies and statistical shortcuts you would attack in an opponent, and fix or cut what you find.

Decide Metrics Before Testingminutes

Before an A/B test or pilot starts, write down the primary metric, the harm metrics that must not get worse, the sample size or stopping rule, and the planned analysis, so that the result can tell you something you did not already want to hear.

Every playbook entry states how strong its evidence is and when not to use it. Browse the full playbook.

References

  1. Huff, D. (1954). How to Lie with Statistics. W. W. Norton
    Choosing the base of a percentage change to inflate or deflate it.
  2. Best, J. (2001). Damned Lies and Statistics: Untangling Numbers from the Media, Politicians, and Activists. University of California Press
    Advocates' selection of comparison points and the habit of asking where a statistic starts.
  3. Cairo, A. (2019). How Charts Lie: Getting Smarter about Visual Information. W. W. Norton
    Index charts and baselines chosen to produce a story, with examples from more than one political side.
Last reviewed
Suggest a correction

Detect Selective Baseline in any text

Paste any message, email, or article into our free Manipulation Detector to see if Selective Baseline or other techniques are being used on you.

Try the Detector FreeCreate Free Account