Skip to main content
    Path to Science

    Thinking Dictionary

    Confounding variable

    What it is

    A confounding variable, or confounder, is a hidden third factor that affects both of the things you are comparing. It can make two things look connected when they aren't. It can also hide a real connection, or make one look bigger or smaller than it really is.¹

    Here is a simple illustration. Ice cream sales and drownings rise and fall together. Ice cream doesn't cause drowning. Hot weather drives both: more people buy ice cream, and more people go swimming. Hot weather is the confounder.

    The entry on Correlation vs causation introduces it as one of four ways a link can fool you. This entry goes deeper: what exactly a confounder is, how to spot one, and how studies deal with it.

    What makes something a confounder

    Not every third factor is a confounder. To confound a comparison, a factor has to be linked to both things:¹

    1. It is linked to the supposed cause (people who do X are more likely to have it).
    2. It also affects the outcome (it changes Y on its own).

    If a factor is only linked to one of the two, it can't create a false connection between them. And a factor that is itself caused by the supposed cause is a different issue, not a confounder.

    Why it matters

    Confounders are behind many misleading health headlines. "People who do X are healthier" often means that the people who do X differ in other ways too. They may have more money, be younger, or have other healthy habits.

    A famous example is hormone replacement therapy. Earlier studies that simply observed women suggested it protected the heart. A large trial in which chance decided who got the treatment found a higher risk of heart disease instead.² One explanation that has been proposed is that the women who chose the therapy were healthier and better off to begin with.³ The full story, which is more complicated, is in the Correlation vs causation entry.

    Here is a lighter example. Across countries, the amount of chocolate people eat goes along with the number of Nobel prize winners per person.⁴ Nobody seriously thinks chocolate wins Nobel prizes. A country's wealth is an obvious candidate confounder: richer countries can afford more chocolate and more research.

    How studies deal with confounders

    • Randomization. In a randomized controlled trial, chance decides who gets the treatment. That makes the groups alike, on average, in every way, including factors nobody thought to measure.⁵˒¹ This is the strongest protection, and it is the main reason such trials exist. See the Randomized controlled trial entry.
    • Matching. Researchers compare people who are alike on key factors, for example people of the same age and income, where one group did X and the other didn't.
    • Adjustment. Researchers use statistics to account for factors they measured, such as age or income. But this only works for confounders they thought of and measured. Unknown ones slip through.¹

    In everyday life

    • "Coffee drinkers live longer." Imagine it's true on average. Who drinks coffee? Perhaps people who are healthy enough to work, with regular routines. Those factors could affect lifespan on their own.
    • "Children who eat breakfast do better at school." Imagine that families who serve breakfast every day also have more time and money to spend on homework help.
    • "People who own dogs are less lonely." Perhaps. But people who are healthy enough and have homes suited to a dog might differ in other ways too.

    The common trap

    The first trap is thinking "they controlled for confounders, so the link is a cause." Adjustment only covers the confounders that were measured.¹

    The second trap is thinking a big study avoids confounding. A larger study reduces random error. It does nothing about a confounder. A huge study with a confounder just gives you a precise wrong answer.

    Some ways out:

    • Ask: "How might the people who do X differ from those who don't?"
    • For each difference, ask whether it could affect the outcome too.
    • Check whether the study was randomized. If it wasn't, check which confounders it measured.

    Try it

    Find a news story that says "people who do X are healthier," or happier, or richer. Write down three ways the people who do X might be different from everyone else, such as income, age, or other habits.

    For each one, ask: could this affect the outcome on its own? If yes, it is a candidate confounder. Then look at the study. Did the researchers measure and adjust for it? Was it randomized?

    Related concepts

    Correlation vs causation · Randomized controlled trial · Indicators of causation · Sample size · Random vs systematic error · Reverse causation · Simpson's paradox

    Sources

    1. Greenland, S., Pearl, J., & Robins, J. M. (1999). Confounding and collapsibility in causal inference. Statistical Science, 14(1), 29–46. https://doi.org/10.1214/ss/1009211805

    2. Writing Group for the Women's Health Initiative Investigators. (2002). Risks and benefits of estrogen plus progestin in healthy postmenopausal women. JAMA, 288(3), 321–333. https://doi.org/10.1001/jama.288.3.321

    3. Lawlor, D. A., Davey Smith, G., & Ebrahim, S. (2004). Commentary: The hormone replacement–coronary heart disease conundrum: Is this the death of observational epidemiology? International Journal of Epidemiology, 33(3), 464–467. https://doi.org/10.1093/ije/dyh124

    4. Messerli, F. H. (2012). Chocolate consumption, cognitive function, and Nobel laureates. New England Journal of Medicine, 367(16), 1562–1564. https://doi.org/10.1056/NEJMon1211064

    5. Schulz, K. F., Altman, D. G., & Moher, D. (2010). CONSORT 2010 statement: Updated guidelines for reporting parallel group randomised trials. BMJ, 340, c332. https://doi.org/10.1136/bmj.c332