The Scarcity Bias

Two experiments provided empirical support for the scarcity bias, that is, when the subjective value of a good increases due to the mere fact that it is scarce. We define scarcity as the presence of limited resources and competition on the demand side (i.e. not enough for two people). In Experiment 1, 180 students were divided into two conditions. The same good was abundant in one condition but scarce in the other one. The scarcity condition involved a partner (competitor) to create scarcity, while the abundant condition did not. Results showed that more participants chose a good when it was scarce than when it was abundant, for two out of four sets of items (ballpoints, snacks, pencils, and key rings). Experiment 2 employed 171 participants and a WTA (willingness to accept) elicitation procedure of the subjective value of the good. Results showed that the scarce good was given a higher WTA price by those participants choosing it, than by those who did not, compared to the WTA of the abundant good, despite the fact that both types of participants assigned a lower market price to the scarce good, as compared to the abundant one. Deux expériences ont confirmé empiriquement le biais de rareté (la valeur subjective d'un bien s'accroît en raison de sa seule rareté). La rareté est définie par une demande conflictuelle car confrontée à des ressources limitées (tout le monde ne pourra pas être satisfait). La première expérience fit appel à 180 étudiants répartis sur deux conditions. Le même bien était abondant dans l'une des conditions et rare dans l'autre. La condition de rareté impliquait la présence d'un concurrent pour créer la pénurie, concurrent inexistant dans la condition d'abondance. Les résultats ont montré que davantage de participants se sont portés sur un bien quand il était rare que lorsqu'il était abondant, cela pour deux des quatre séries d'items (stylos à bille, casse‐croûte, crayons, porte‐clefs). 171 sujets participèrent à la seconde expérience qui utilisa une procédure de mise en évidence de la valeur subjective du bien (WTA: prix jugé acceptable pour se procurer le bien). Il apparaît que le bien rare bénéficiait de la part de ceux qui le choisissait d'un prix WTA supérieur à celui octroyé par ceux qui le délaissaient, par comparaison au WTA du bien abondant, et cela en dépit du fait que les deux catégories de sujets accordaient un prix de vente inférieur au bien rare par rapport à celui qui était largement disponible.

This paper examines the attraction effect produced by perceived scarcity on consumer behaviour ([10]).1 The focus is on a mechanism of attraction that scarcity produces on the consumer's preference model. This mechanism has nothing to do with the so‐called Veblen's goods ([18]), that is, commodities which stand as status symbols. Furthermore, we are not interested in discussing the effect played by scarcity within the market mechanism and on prices.

Scarcity is here considered as a local‐based attribute of goods—where the idea of attribute is intended in [4] view (1966)—linked to demand‐side competition. Assuming a specific place and moment in time, a consumer subjectively perceives that a given good possesses the attribute of scarcity, and therefore models her preference map according to this attribute. The perception of scarcity is sustained by a competitive pressure on the demand side, and the consumer infers from this competition that the scarce good should possess some inner intangible property.

Imagine a situation where a set of goods are displayed on a shop shelf in such a way that suggests that one of the goods displayed is scarce. The hypothesis here tested from an experimentally based perspective, is that the scarce good has more chance of being chosen than the abundant ones. It follows that, more generally, we are concerned with the empirical testing of the assumption that a context‐dependent attribute of goods can influence in a specific way the preference modelling of the consumers. In our view, scarcity can act as a direct positive attractor on the consumers' preference structure, without necessarily being mediated by any social or cultural factor.

The literature on consumers' behaviour has treated scarcity as an attribute from which the consumer infers other attributes such as price ([7]), or uniqueness ([16]). More generally, psychological and marketing research has repeatedly found that scarcity affects the consumer's perception of goods by enhancing attractiveness ([17]) and desirability (see [6], for a review).

For example, a psychological theory that suggests an explanation for scarcity effects on the perceived utility of a good, is "uniqueness theory" ([16]). According to this theory, consumers may desire scarce goods as a way to differentiate themselves from others. Consistent with this explanation, studies have found that subjects high in need‐for‐uniqueness show a stronger preference for scarce commodities than those low in need‐for‐uniqueness ([2]; [14]; [8], [9]).

An alternative explanation refers to price appreciation ([5]). According to this theory, people presumably associate scarcity with higher prices ([7]) and, since high‐priced goods are status symbols ([18]), and price is often used as a cue to good quality ([12]), then people perceive scarce goods as a vehicle for acquiring a higher status symbol, or as being of higher quality.

Our study analyses how the attraction for a good is increased by the mere fact that this good is presented as scarce. The same good is abundant in one condition, but scarce in the other. Moreover, the scarcity condition involves a partner (competitor) to create scarcity, while the abundant condition does not. Our central hypothesis is that more participants will choose the good when it is scarce than when it is abundant. Furthermore, we will show how those people who choose the scarce good also give it a higher willingness to accept (WTA) price, as compared to those who do not choose the good, despite the fact that both types of participants assign a lower market price to the scarce good with respect to the abundant good.

Our work is different from the research cited previously because we used a between‐subjects design with real choices and tangible products with economic value, which are actually distributed among the participants, and we included a competitive pressure on the demand side.

The aim of Experiment 1 was to compare the frequency of choices for the same bundle of goods (AB) in two conditions: when good B was abundant, and when it was scarce. Through a control condition, that we term abundance condition, we were able to obtain the preference structure of the statistical population, to which the sample of subjects who participated in the scarcity condition belonged. Through the experimental condition, that we term scarcity condition, we were able to test whether this preference structure changed when we introduced scarcity—represented by the presence of only one good B and demand‐side competition. We hypothesised that in a condition of scarcity, and in the presence of demand‐side competition, the participants would change their preference structure: the B‐type good would be chosen more frequently under the scarcity condition than under the abundant condition.

Participants were randomly recruited among the undergraduate students of the Faculty of Economics of the University of Trento. One hundred and eighty participants (57% males; mean age 23) were divided into two experimental conditions: abundance and scarcity. Seventy‐six participants were assigned to the abundance condition, while 104 participants were assigned to the scarcity condition.

Under both conditions, the task assigned to the participants was to choose two items among a set of four goods. Items were actually distributed to the students. In the abundance condition, participants chose alone, while in the scarcity condition they played with a partner (competitor). In the abundance condition, the sets of goods included two goods of type A and two goods of type B (AABB). Participants played alone and were free to build up their favourite set of goods: AA, AB or BB. In the scarcity condition, the sets of goods included three goods of type A and one good of type B (AAAB). Each of the two participants had to choose sequentially from the same set of goods, and this meant that the final chosen set could be either AA or AB. Only the first player was sure to be able to obtain both goods (AB) if she liked, whereas the second player could choose AB only if her opponent had not chosen good B at the first move. As a consequence of this experimental setting, the second players in the scarcity condition had a different choice set from participants in the abundance condition. For this reason, only the data from the first players in the scarcity condition were used to compare with data from the abundant condition. This can be seen as a partial limit of our design, but it is an internal consequence of the nature of the problem of scarcity itself—an end result of demand‐side competition combined with scarcity is that the consumer is in some way constrained in her choices—and this cannot be overcome.

We used four goods plus a filler good. The four goods were: ballpoints, snacks, pencils, and key rings. The filler good was a plastic line. Figure 1 shows the goods used in the experiment and the filler good. The filler good was introduced with the aim of making the underlying logic of the experiment less transparent. Each good was represented by a couple of items, item A and item B, which were different enough as to stimulate the building of a deterministic structure of preferences. At the same time, the items of the pair were not so different as to induce too strong a dominance of good A (always slightly—20–30%—more expensive than B) over good B. On the other hand, the filler good was represented by a pair of identical items that differed only in colour (blue or yellow). In this way, we hoped to disguise the underlying logic of the pair structure (dominant item vs. weak item) of the four target goods. It is worth noticing that all the goods used in the experiment belong to a category of commonly used commodities, and none of them could in any way be interpreted as a status symbol, at least for the time and place where the experiments were carried out.

Graph: 1 The types of goods used in the experimental set.

In the abundance condition, the participants entered a room where the goods were displayed on five tables divided by a stud wall. The layout of the goods on each table was randomised for each participant: AABB; ABAB; BBAA; ABBA; BAAB. The participants had to examine each set of goods for not less than 20 seconds, and could choose with no time restrictions. The 20‐second time to examine each set was introduced in the experimental design because we wanted to be sure that the participants did not make a random choice. If they needed more time, they could take it. Each participant was asked to choose two goods from the displayed ones. The goods chosen were given to the participants immediately after the end of the experiment.

In the scarcity condition, participants stood facing each other. A coin was tossed to determine who was the first player; then the game started. The first player chose first, then the second player, then the first again, and finally the second again. After finishing with the first set of goods, the participants changed their opponent and played with a new set of goods. No participant played twice with the same opponent, and none of them played only against opponents of the same gender. At the end of the experiment, the participants received the goods that they had chosen, plus a participation fee (5€). Both the display of the goods on the table, and the display order of the sets of goods were randomised, with the exception of the filler good which was always shown to the participants as their third choice. Finally, after the choice, the participants were asked to give a market value to each of the goods with which they were presented.

Participants in the scarcity condition had one choice less than the participants in the abundance condition, but we needed to compare the choices in the two conditions. In the abundance condition, each participant could choose out of three different possible combinations of goods: AA, AB and BB. For each of these combinations the following preference ordering can be built:

Looking at the preference structure, we should consider only the first four systems of preference orderings out of the total of six shown: ({AA AB BB}, {AA BB AB}, {AB AA BB}, {AB BB AA}), because in the scarcity condition the participants could not choose BB. Therefore, all the players who chose BB in the abundance condition were, for the sake of precaution, inserted in group AB AA—we call this criterion "restricted comparative criterion" (RCC). Adding the participants who preferred BB to the group of AB means using a restrictive criterion, which allows stricter control over the starting assumption, i.e. that the percentage of those who prefer AB to AA is higher in the scarcity condition (B scarce) than in the abundance condition (B and A equally available). The results obtained in the abundance condition are reported in Table 1.

1 
Frequencies of the Three Choice Patterns (AA, AB and BB) of the Four Goods and the Filler Item for Participants in the Abundance Condition

The AA bundle dominates the choice pattern for three goods out of four (ballpoints, snacks and key rings), without considering the filler good (line). Participants, hence, expressed a clear AA AB preference structure for three goods, and an inverse AB AA preference structure for one good, pencils. It follows that when we analyse the data obtained in the scarcity condition, we should take into account that in the case of pencils AB dominates AA and, therefore, in order to retain our hypothesis it will be sufficient that the frequency of AB in the scarcity condition is higher than the frequency of AB in the abundance condition.

In the scarcity condition, the sequences of choices resulting from each pair of players were the following: AAAB, AABA, ABAA, BAAA. For example, the AAAB sequence resulted from the first player choosing A, then the second player choosing A, the first player choosing A again, and the second player choosing the remaining B. In this case the first player ends up with AA and the second player with AB. In order to compare the frequency of choices of the scarce good B between the scarcity and the abundance conditions, we used only the choices of the first players, and coded as "AB" the sequences BAAA and AABA, and as "AA" the other sequences: AAAB and ABAA. In the first two sequences, indeed, the first player preferred B when both A and B were available while in the other two sequences one can be sure that the first player would actually chose AA only when the second player left her the good B as an option for her second choice (i.e. in sequence AAAB). Adding AAAB to ABAA as AA is therefore to be interpreted as a restricted criterion (equivalent to RCC) and the results obtained are more robust.

The results obtained from the scarcity condition are reported in Table 2 (first player only). As noted earlier, the first player is the only one comparable to the choices made in the abundance condition, since she is free to choose the preferred set of goods.

2 
Frequencies of the Two Choice Patterns (AA, AB) of the Four Goods and the Filler Item for First Players in the Scarcity Condition

AB became the dominant preference for ballpoints and key rings in the scarcity condition, whereas the dominant preference did not change for snacks and pencils. A comparison between the scarcity condition and the abundance condition is shown in Table 3. In the abundance condition the choices AB and BB have been added and labelled "AB", according to the above‐mentioned RCC, for the sake of comparison.

3 
Observed Choice Patterns and Chi‐square Values of the Comparison between the Two Conditions

1 §   First mover only.

The choice AA, which is the favourite bundle in the abundance condition for ballpoints, key rings, and snacks, is not the preferred bundle for ballpoints and key rings in the scarcity condition for the first mover, while it remains the most chosen set for snacks in all conditions. More precisely, and looking at Table 3, when B is abundant the percentages of AB are lower than when B is scarce for ballpoints and key rings. The percentage of ballpoints AA in the abundance condition is 55.2, while it decreases to 44.2 in the scarcity condition; the percentage of key rings AA in the abundance condition decreases from 60.5 to 36.5 in the scarcity condition. Pencils did not obey the general rule of AA preferred to AB in the abundance condition, so it is not surprising that AB is still the preferred option also in the scarcity condition. Nonetheless, the percentage of choices in favour of AB in the scarcity condition is slightly higher than the same percentage in the abundance condition (AB AA 64.5 abundance condition; 67.3 scarcity condition). On the other hand, applying a different criterion (from RCC) for the case selection, that is to say looking only to the sub‐samples of participants who have chosen AAAB (AA) or BAAA (AB), the values for this good change in an even stronger way in the expected direction. Using this different criterion the percentage of participants who choose AB increases from 64.5 per cent (abundance) to 96.9 per cent (scarcity), χ2(1) = 12.31; p=.0001.

The chi‐square tests computed for each good show that the choices in the abundance and scarcity conditions are different only for key rings. On the other hand, if an aggregate variable of all four goods (ballpoints, key rings, snacks, pencils) is computed, we observe that the preference for AB in the abundance condition significantly increases when the choice is made in a scarcity condition, and the difference is significant at p=.034.

In summary, we observe a change in the pattern of choices that supports our hypothesis in three goods out of four, even if only key rings reach statistical significance. There are different explanations for the absence of statistical significance of two out of three goods (pencils and ballpens). As regards pencils, however, we have seen that scarcity has a significant effect if we adopt a less strict criterion to aggregate the data. As regards ballpens, there is no apparent reason that explains the lack of statistical significance.

Snacks are the only good that falsifies our hypothesis. The bundle AA was preferred to AB in the abundance condition, as supposed, but the participants' choices did not shift at all to AB when we introduced scarcity. In fact, the choices remained the same in the two conditions. Most likely, this is due to a peculiarity of the bundle. The snack bundle is made up of two products which are very different from each other. One is an elaborate chocolate snack, while the other is a very basic type of chocolate snack (simple chocolate candy). The AA bundle for snacks records the highest percentage (65.8%) of preferences in the abundance condition among the goods analysed, and the qualitative (and market price) difference between the two alternatives was the largest.

The assessed market price and the statistics for each good are reported in Tables 4 and 5. Table 4 reports the statistics separately for those participants who chose the scarce good (AB) and for those who chose the abundant good (AA). As expected, the mean assessed market price of good A is always higher than the assessed market price of good B, for both groups AA and AB.

4 
Descriptive Statistics of the Participants' Assessed Market Price in Euro Cents (First Player Only, Scarcity Condition)

5 
Statistics of the Mean Participants' Assessed Market Price Differences between Group AA and AB in Euro Cents (First Player Only, Scarcity Condition)

In the first four columns of Table 5 one can see that the assessed market prices of the abundant good (A) and of the scarce good (B), respectively, are not significantly different between the group AB and the group AA, except for pencils. This confirms that pencils was not a good bundle, because it consisted of two items too similar to each other as discussed earlier. In the last four columns of Table 5 are reported the differences and the p‐values between the mean assessed market price of the abundant good (A) and the scarce good (B) separately for group AA and group AB. One can notice that all the differences are positive and statistically significant. This means that the assessed price of good A is significantly higher than the assessed price of good B, independently of the choice of the participants. In line with our assumptions, in the presence of competition, a percentage ranging from 35 to 63 per cent of the first players exhibit incoherence: they prefer B, a scarce good, over A, an abundant good, even if they systematically perceive A to be more expensive than B.

In Experiment 1, we observed the impact of demand competition on the preference for the scarce good. Participants' preferences for AB always increased (from a minimum of 1% to a maximum of 23%) when scarcity was introduced. This result is even more striking if we consider that participants assessed that the market price for the scarce good (B) is always lower than that of the abundant good (A).

This pattern of results supports our prediction on the role of scarcity: it works as an attractive mechanism that increases the value that the subjects attribute to the good. More precisely, what scarcity increases is not the assessed market value of the good, but the subjective value of the good. To test this assumption, we designed Experiment 2, in which participants were asked to choose among the set of goods as in Experiment 1 but, afterwards, they were also asked to indicate a price at which they would be willing to give back the scarce good to the experimenter (willingness to accept, WTA).

In Experiment 1 we asked participants to assess the market price of each good. This information was useful to measure the perceived value of the good. It is true, however, that the subjective monetary value can differ from the assessed market price. For example, I might assess a market price of 10 Euro for a good, but I would not be willing to give up that good at that price. I might be willing to give up the good at a lower or higher price than the assessed market price, depending on the subjective value I attribute to that good. For our purposes, it is important to have a measure of the subjective monetary value of the scarce good, that is, the price at which one is willing to give back the scarce good to the experimenter (WTA). Since we predict that scarcity works as an attractive mechanism that increases the value that the subjects attribute to the good, we expect that the distance between WTA for the scarce good (B) and WTA for the abundant good (A) will be smaller for those that choose AB than for those who do not choose AB. In other words, we expect that those players that choose the AB bundle will give the scarce good B a higher value compared to the abundant good A than those players who do not choose the AB bundle.

We have already found in Experiment 1 that AB individuals set the maket price for the abundant good (A) higher than that for the scarce good (B). If individuals in Experiment 2 give equal or higher WTA prices to the scarce good (B) than to the abundant one (A), we can state that scarcity acts as a generator of value. In terms of equations, we hypothesise that P(A)/P(B) > WTA(A)/WTA(B) for first players who choose B first (pattern BAAA), where P(A) and P(B) are the assessed market prices for goods A and B, respectively. Rationally speaking, P(A)/P(B) should equate with WTA(A)/WTA(B). Any variation from this equality is proof of a scarcity effect acting on the perceived subjective value of A and/or B, that is by decreasing the WTA of A compared to the P(A) or by increasing the WTA(B) compared to the P(B).

This experiment used an "incentive‐compatible" (IC) WTA procedure of the type used by Becker‐deGroot‐Marschak (BDM) ([1]), and defined in the literature as revealing the subject's true preferences. In the BDM mechanism, in fact, subjects give back the item in exchange for the random price tossed if it is higher than their WTA price.2

Experiment 2 used the same design as Experiment 1, except that only the scarcity condition was implemented. One hundred and seventy‐one3 (52% males, mean age 22) undergraduate students of the Faculty of Economics at the University of Trento were recruited to play one against the other. Participants in Experiment 2 had not participated in Experiment 1.

The procedure was the same as in Experiment 1, except that only one good was used (key rings), and the players were asked a WTA price for the goods they chose. The WTA procedure consisted in declaring a monetary value (minimum 0.05 Euro and maximum 4 Euro) for each of the two key rings they had chosen (A and B). The participants were told that a random device would be tossed and if the random price was higher than their WTA price they would have to give back the key ring to the experimenter, and receive the random price tossed.

Forty‐eight first players (56%) chose AA and 38 chose AB (44%). These percentages are similar to those found in Experiment 1, even if less pronounced in favour of scarcity. We did not have a control condition in Experiment 2, so it is difficult to say whether these percentages are significantly different from the control condition.

To test our first hypothesis regarding WTA prices, we first compared WTA(B) of the first players who chose B as their first choice (sequence BAAA) with the WTA(B) of the second players who were left with nothing other than B to choose (sequence AAAB). The means are: WTA(B)BAAA= 125.8 (n= 24) and WTA(B)AAAB= 69.1 (n= 22). The Mann‐Whitney‐Wilcoxon test reveals that this difference is significant, Z=−1.936; p=.053. We can compare these values also with the WTA(A) for the same groups: WTA(A)BAAA= 132.5 (n= 24) and WTA(A)AAAB= 139.1 (n= 22). The Mann‐Whitney‐Wilcoxon test shows that this difference is not significant, Z=−0.212; p=.832 (see Figure 2). To test the first hypothesis more directly we also ran a 2 × 2 ANOVA with a first factor, type of group (players who chose B as first choice vs. second players who were left with B) between subjects and a second factor, WTA (WTA for A vs. WTA for B) within subjects. As expected, we found a significant interaction between the two factors, F(1, 44) = 10.30; η2=.190. p=.002. In support of our hypothesis, we found that those who chose the scarce good also attributed a higher subjective value to it, compared to those who chose the abundant good.

Graph: 2 Mean willingness to accept prices for the scarce good and the abundant good when the scarce good is chosen first (BA) or when the same good is left as last choice (BA).

To test our second hypothesis regarding WTA price in comparison with the assessed market price from Experiment 1, we computed the ratio k= P(A)/P(B) and the ratio h= WTA(A)/WTA(B) for the first players who chose B (sequence BAAA). As shown in Figure 3, WTA prices increase significantly less than market prices. The k ratio is 2.70 and the h ratio is 1.35, t(42) = 1.82; p=.076. This means that the distance between the assessed market prices of the two goods is greater than the distance between the WTA for the two goods. In other words, when participants have to set a selling price for the scarce good that they have chosen, they set a price which is equal to that of the abundant good, despite the fact that they had previously attributed a lower market price to the scarce good.

Graph: 3 Mean assessed market price and WTA for first players who choose B first.

Experiment 2 was designed to test the assumption that scarcity works as an attractive mechanism that increases the value that the subjects attribute to the good. More precisely, the assumption is that scarcity increases the subjective value of the good, but not the subjectively assessed market value. Results, indeed, showed that those who chose the scarce good also attribute to it a higher subjective value compared to the abundant good. This experiment also showed that P(A)/P(B) was greater than WTA(A)/WTA(B); therefore, when participants are asked a WTA price for the scarce good (that they have chosen first), they set a price which is equal to that of the abundant good, despite the fact that they had previously assessed a lower market price for the scarce good than for the abundant one.

In this study, we show how the use of tangible products with economic value, real choices, and a competitive pressure on the demand side, allows one to ecologically investigate people's preferences in the context of scarcity. Furthermore, the specific kind of goods used in the experiments, the general frame adopted, and the coherency between the subjective perception of the market values declared by the participants and the real market prices, allows us to state that scarcity does work as an attractor even when it is detached from any status symbol effect.

The results obtained cannot be considered entirely conclusive because when real choices and real products are used the different attributes of these interact in unpredictable ways and tend to reduce the effects. Nevertheless, the results show that individual preferences for a product increase when it is presented as scarce, in three out of four sets of products studied, even if only two reach significance. As we have argued above, it is not the type of product that determines whether one can profitably use scarcity to increase or not its attractiveness, but it is rather the bundle of products, hence the relative attractiveness of the two products, that determines this pattern of results, at least in the experimental setting.

We have also shown that the increase in preferences for the same good when it is presented as scarce is accompanied by an increase in WTA prices, that is in the price at which one is willing to sell the good to the experimenter. This result strengthens our previous argument that scarcity does change the subjective value attributed by the participant to the product. We labelled this effect the scarcity bias since it goes against the principle of stochastic dominance, and can induce a distortion, more properly, a variation, in the agent's preference structure, acting on the perception of irrelevant context cues (scarcity) and not of fundamental cues ("objective" attributes of the product).

As stated before, however, we think this bias should not be considered irrational tout court since it is extremely useful in the context of competition for resources, and helps development and survival of the species. Very often, indeed, scarcity is accompanied by higher product quality. Further research is needed to address the different contexts in which the bias is more or less strong. For example, one could study the subjective attractiveness of goods. One could present a larger set of goods to participants which they have to rate according to subjective attractiveness. In a second stage, one could employ the choice task and measure the differences in the subjective attractiveness ratings of chosen goods.