IKEA effect

The mechanism is narrower than “work makes value”. The premium appears only when the work is completed successfully, which points at competence rather than effort as the engine: finishing is evidence you can act on the world and get the result you intended, and the artifact is the receipt. Mochon, Norton and Ariely found that self-reported feelings of competence mediated the valuation gap (Sobel's Z = 2.08, p < .05) and that affirming the self beforehand removed it — if you already feel capable, your bookcase has nothing left to prove. Franke, Schreier and Kaiser arrive in the same place from the design side under a different name: the premium their customisation toolkits produced was mediated by a feeling of accomplishment (Sobel's z = 3.822, p < .001), while perceived process costs — their sunk-cost proxy — were unrelated to willingness to pay (F(1, 115) = 0.032, n.s.). The published conceptual replication favours a third account, psychological ownership, and a developmental study concluded that none of the three fits its data. Treat the mechanism as unsettled and the completion requirement as the part you can design against.

Norton, Mochon and Ariely (2012) ran four experiments with real money on the table. In Experiment 1A (N = 52 undergraduates), participants who assembled a plain black IKEA storage box bid more for it under an incentive-compatible Becker-DeGroot-Marschak procedure (M = $0.78, SD = 0.63) than participants who inspected an identical pre-assembled box (M = $0.48, SD = 0.40), t(50) = 2.12, p < .05 — the 63% premium the effect is usually quoted with. Liking moved too, 3.81 against 2.50 on a 7-point scale, t(50) = 3.58, p < .001, and participants rated the boxes as utilitarian rather than hedonic (M = 3.13 on a 9-point scale, below the midpoint at t(51) = 7.36, p < .001), so this is not a story about having something to show off. In Experiment 1B (N = 106), builders valued their own origami frogs and cranes at M = $0.23 (SD = 0.25), nearly five times what uninvolved bidders would pay for the same creations (M = $0.05, SD = 0.07) and statistically indistinguishable from what bidders paid for expert-folded origami (M = $0.27, SD = 0.26), F(2, 100) = 5.34, p < .01. Amateur output prices like an expert's, but only to the person who made it — and a further 14 builders' own bids (M = $0.19) matched what they predicted the average student would bid (M = $0.21), t(13) = .45, p > .65, so they were not merely overbidding to keep their work.

The completion boundary is the paper's main contribution. In Experiment 2 (N = 118, Lego sets of 10 to 12 pieces), participants bid on their own set and on their partner's; only in the build condition were bids for their own significantly higher — about twice as high — t(39) = 3.08, p < .01. Building the set and then taking it apart erased the gap, t(39) = 1.20, p = .23, and the pre-built condition showed no gap either, t(37) = 1.30, p = .20. In Experiment 3 (N = 39), builders who finished an IKEA box bid M = $1.46 (SD = 1.46) against M = $0.59 (SD = 0.70) from participants stopped before the last two steps, t(37) = 2.35, p < .05, even though the unfinished group held every piece needed. Self-rated do-it-yourself disposition raised valuations overall (β = .35, p = .06) but did not interact with condition (β = .09, p > .60): novices showed the effect as much as enthusiasts, so this is not a feature for power users.

Mochon, Norton and Ariely (2012) ran the mechanism in both directions. Builders of a Lego car outbid non-builders (M = $1.20, SD = 1.35 against M = $0.57, SD = 0.76; N = 79, 75 analysed, t(73) = 2.44, p < .05), with competence mediating the gap. Their second experiment (N = 135) is the sharper result: without self-affirmation the effect appeared ($0.72 against $0.46, t(52) = 1.99, p = .05), but after participants wrote about the value they ranked most important it vanished ($0.49 against $0.58, t(63) = .71, p = .48), interaction F(1, 115) = 3.90, p = .05. Run backwards, threatening competence with unsolvable maths problems raised the share who chose to assemble an IKEA bookcase themselves from 33% to 58% (N = 75, χ²(1) = 4.72, p < .05) and the share choosing a self-assembly table over a pre-assembled one from 27% to 74% (N = 41, χ²(1) = 8.79, p < .01). Asked in the abstract, though, 92% of a separate group (N = 51) said they would pay more for pre-assembled products, χ²(1) = 36.26, p < .001: nobody wants the work in prospect, which is why the build is offered and never required.

The strongest adjacent evidence comes from Franke, Schreier and Kaiser (2010) in Management Science, five studies with real-money auctions run on commercial customisation toolkits. Study 2 held the design constant and varied only the process: bidders who rebuilt a target T-shirt design in Customink's configurator paid €6.85 (SD = 3.70) against €4.75 (SD = 3.04) off the shelf, over 40% more, F(2, 108) = 3.512, p < .05, η²p = .06. A third group that configured and then bid on the identical off-the-shelf shirt showed no premium (€5.26), which rules out mere exposure, mood and attention. Study 3 (116 students, a ski manufacturer's live toolkit) had one group design a ski graphic freely and the other pick from 28 of the manufacturer's standard designs, then bid on the graphic rather than the ski: €74.42 (SD = 56.85) against €45.89 (SD = 43.58), F(1, 116) = 9.153, p < .01, η²p = .07. Study 4 (n = 129) found the premium larger when the target design was attractive (a €4.90 gap) than unattractive (€1.94), interaction F(1, 129) = 4.446, p < .05. Study 5 (n = 66, watch faces) found a toolkit with free design produced €30.34 (SD = 23.19) against €19.21 (SD = 10.22) for one offering only predefined attributes, F(1, 66) = 6.502, p < .05, η²p = .09. How much the user gets to decide, and whether the result is any good, sets the size of the effect.

The replication record is thinner than the citation count suggests. Sarstedt, Neubert and Barth (2017) report a conceptual replication of both Ariely-group papers supporting the robustness of the original effect and pointing to psychological ownership as the mechanism, though the article is paywalled and its sample sizes and effect sizes could not be checked. Marsh, Kanngiesser and Hood (2018) found the bias emerging in children around age 5, but also that it was not moderated by how much effort 5-to-6-year-olds invested and that ownership alone did not explain it, concluding that the existing theories are incompatible with their results. Oishi and colleagues (2023) confirmed the effect behaviourally in a neuroimaging study, t(23) = 3.018, p = 0.007, d = 0.64 — but only after excluding 4 of 30 participants whose willingness to pay for self-assembled products was not higher, which both inflates that estimate and shows that a meaningful minority show no effect at all. Set against that, the original effects come from small undergraduate samples, several p-values sit just under .05, the stakes are cents, and there is no preregistered high-powered replication or meta-analysis to point at. The direction is reproducible across labs, products and currencies; the magnitudes and the mechanism are not.

Drop the cake-mix anecdote, in which manufacturers supposedly removed powdered egg so housewives would add a fresh one and feel involved. Norton and colleagues open with it, citing a popular history, but food historians dispute it: fresh-egg mixes were patented as early as 1935, two decades before Ernest Dichter's 1950s motivational research for General Mills, a popular fresh-egg line was already selling by 1951, and the reformulation has mundane explanations in shelf life and cake quality. It is folklore, not a pattern with evidence behind it.