Projection Bias: Why You Shop Wrong When You're Hungry

You're at the supermarket, and you're hungry. You pile things into the trolley with a kind of desperate optimism — the cheese, the crisps, the ready meals, the dessert that seems absolutely essential right now. You get home. You're not hungry anymore. You look at the trolley's contents with a mixture of amusement and mild regret. Where did all this come from? It came from projection bias: the systematic error of assuming that how you feel now is a reliable guide to how you will feel later.

Loewenstein and the Architecture of Projection

The formal account of projection bias was developed by George Loewenstein, Ted O'Donoghue, and Matthew Rabin in a 2003 paper in the Quarterly Journal of Economics that has become one of the most cited works in behavioural economics. Their core insight was deceptively simple: people predict their future utility as a weighted average of their current state and their "true" stable preferences — but they systematically over-weight the current state. The result is a predictable set of forecasting errors that appear across a wide range of choice contexts.

Loewenstein and colleagues modelled this formally: if your true future preferences would give you utility U*, and your current state gives you utility feelings of U_current, you will forecast your future utility as something like U* + α(U_current − U*), where α is the projection parameter. When α = 0, you forecast perfectly. When α = 1, you assume the future will feel exactly like the present. Most people sit somewhere in between but systematically closer to 1 than to 0. You know intellectually that your hunger will pass; you just can't quite feel what it will be like when it does.

The Hungry Shopper and the Cold-State Forecaster

The grocery shopping case is one of the most intuitive illustrations. Research has repeatedly found that hungry shoppers spend more, buy more calorie-dense food, and make more impulsive purchases. This is partly about self-control failures in the moment, but it is also — more interestingly — about forecasting. Hungry shoppers genuinely believe they will want all this food when they get home. They are predicting that their future self will be as hungry as their current self, which is obviously wrong. The physical state colours the prediction.

The reverse also holds. Loewenstein and colleagues conducted a field study in which participants were asked, in winter, how much they would enjoy a bicycle trip to the beach — and they consistently underestimated it. Standing cold in December, it was genuinely difficult to imagine the visceral pleasure of warm sand and sunshine. The cold state projected onto the imagined warm experience and muted it. People bought fewer beach passes, signed up less for summer activities, and in general systematically undervalued warm-weather experiences when they were making decisions in cold weather.

This directly parallels the hot-cold empathy gap — the difficulty of understanding our behaviour in emotionally or physiologically "hot" states when we are currently in a "cold" state (and vice versa). Projection bias and the empathy gap are closely related but distinct: projection bias is about predicting future preferences, while the empathy gap is about understanding the motivational force of different states. They are often co-present in real decisions.

Cars and Convertibles

One of the most elegant real-world demonstrations comes from a study of convertible car purchases by Loewenstein and colleagues. They found that people buying convertibles on warm, sunny days — days when driving with the top down seemed obviously wonderful — were systematically less satisfied with their purchase months later than people who bought convertibles on overcast days. The sunny-day buyers had over-weighted the current-state pleasure of imagining open-top driving; the grey-day buyers made a colder calculation that happened to better approximate average conditions.

The same pattern appears in housing decisions. People touring houses in summer consistently rate garden space more highly than people touring identical houses in winter. Buyers who purchase in summer report being somewhat less satisfied with their garden year-round than their pre-purchase predictions suggested. The current-state enjoyment of sunlight and green space was projected onto an imagined future in all weathers — a future that, inevitably, includes a lot of rain.

Addiction and Escalation

Loewenstein's work extends to consumption habits with a particularly sobering application: addictive behaviour. People who are not currently addicted systematically underestimate how much they will want a substance or behaviour if they become habituated to it. This is projection bias running in the wrong direction: the current non-craving state is projected onto the imagined future addicted state. "I won't get hooked" is, in part, a projection error — an inability to imagine how different the future craving state will feel from the current neutral state.

Conversely, people who are currently addicted overestimate how much they will want the substance after quitting. The current craving state is projected onto the imagined future clean state. This contributes to the perceived hopelessness of quitting — the future without the substance seems worse than it will actually be. Both forecasting errors are systematically wrong in ways that worsen decision-making around addiction.

Career Choices and the Projection Trap

Projection bias appears in longer-horizon decisions as well. People choosing careers during periods of high motivation overestimate how motivated they will feel doing routine tasks years into a role. Students excited about a subject in undergraduate study over-commit to careers in that area, underestimating the regression to baseline that typically follows. People who are bored and restless in their current job over-estimate how different — and therefore more exciting — a new role will be; they project their current state of novelty-seeking onto a future that, once routine sets in, will feel much like the old one.

This interacts with planning fallacy in career and project planning. The planning fallacy captures the systematic underestimation of time and cost; projection bias adds a related error — the overestimation of how good it will feel to be done, and the underestimation of how tedious the middle portion will be. The excited current self projects enthusiasm onto the imagined future slog.

Retail and the Subscription Economy

Projection bias is an engine of commercial revenue. Gym memberships are the canonical example: sign-up rates are highest in January, when people are in a motivated, resolution-making state, and over-predict how often they will attend. The business model of most gyms depends on members paying for attendance they will never make. DellaVigna and Malmendier (2006) found that gym members who chose monthly contracts paid an average of 70% more per visit than they would have on a pay-per-visit basis — consistent with systematic over-prediction of future attendance driven partly by projection from the current motivated state.

Streaming services, app subscriptions, and "free trial" conversions all exploit a related mechanism. You sign up in a moment of genuine interest and project that interest forward. You do not adequately account for the regression to baseline that tends to follow novelty. The subscription billing continues long after the enthusiasm has faded, because cancellation requires effort and because — at each renewal moment — the current mild-use state is again projected forward.

Can You Correct for It?

The partial antidote to projection bias is the same as for many forecasting errors: reference class thinking. Instead of imagining how you will feel, ask how you have typically felt in similar situations after the novelty wore off. Instead of asking "how motivated will I be to go to the gym?", ask "how often have I historically exercised regularly when I didn't have a specific commitment forcing me to?". Past behaviour in similar states is a much better predictor than current-state enthusiasm.

A second strategy is deliberate time-shifting: force yourself to make the decision from an imagined future state. Before buying the convertible on a sunny day, ask yourself: how will I feel about this in February, on a rainy commute? Before signing up for a new subscription while excited, schedule a calendar reminder to evaluate it in six weeks. The act of mentally simulating the non-excited state provides a partial corrective to the projection.

That said, projection bias is robust because it operates at a level below deliberate reasoning. Knowing about it does not eliminate it — as Loewenstein and colleagues noted, people who had experienced a state multiple times (and thus had good experiential data) still showed significant projection from their current state. The error is sticky because it is embedded in how we access and simulate future states, not just in our conscious reasoning.

Sources & Further Reading

• Loewenstein, G., O'Donoghue, T., & Rabin, M. "Projection Bias in Predicting Future Utility." Quarterly Journal of Economics 118, no. 4 (2003): 1209–1248.
• DellaVigna, S., & Malmendier, U. "Paying Not to Go to the Gym." American Economic Review 96, no. 3 (2006): 694–719.
• Conlin, M., O'Donoghue, T., & Vogelsang, T. J. "Projection Bias in Catalog Orders." American Economic Review 97, no. 4 (2007): 1217–1249.
• Loewenstein, G., & Schkade, D. "Wouldn't It Be Nice? Predicting Future Feelings." In Well-Being: The Foundations of Hedonic Psychology, ed. Kahneman, Diener, & Schwarz (1999).
• Wikipedia: Projection bias