Temptation and Impulse Buying Daniel Houser Professor of

Temptation and Impulse Buying Daniel Houser Professor of Economics Director, Interdisciplinary Center for Economics Science George Mason University, Fairfax, VA

Temptation and Impulse Buying

Impulse Buying Impulse buying is a sudden and powerful urge that arises within the consumer to buy immediately (Beatty and Ferrell 1998; Rook 1987). Impulsive purchasing is defined as involving spontaneous and unreflective desires to buy, without thoughtful consideration of why and for what reason a person should have the product (Rook 1987; Rook and Fisher 1995; Verplanken and Herabadi 2001).

Why Investigate Impulse Buying? • Impulse buying is easier now than it has ever been. – In past times, one would usually need to wait several hours between seeing a product advertised and actually purchasing the product. • This can be useful in reducing the impact of impulsivity on purchase decisions. – Now cash machines, television shopping and the internet make it easy to respond to any urge immediately.

Key Questions in Impulse Buying l l What products are best classified as impulse items? Does the “what” include a “where”? l l Do certain retail environments promote impulse buying? Who engages in impulse buying? l l Individual differences Impulsive vs. non-impulsive “types”

Key Questions in Impulse Buying l Why do impulse purchases occur? l Mood effects l Rook and Gardner 1993 asked people to name the single mood that most often preceded an impulse purchase § l Answers: “Pleasure”, “carefree” and “excited. ” Impulse buying in negative mood is also common, because buying is a way to change the mood. § Consistent with research on self-gifting (Mick and Demoss, 1990).

Key Questions in Impulse Buying l Why do impulse purchases occur? l Willpower effects (Hoch and Loewenstein, 1991) l l Consumer decisions represent an ever-shifting conflict between desire and willpower. Impulse spending occurs when desire for a product outstrips willpower to deny oneself the purchase.

Key Questions in Impulse Buying l How does one model impulse buying, or temptation in general? l Gul and Pesendorfer l l l l 2001 Econometrica 2002 Econometrica 2003 Review of Economic Studies 2003 Review of Economic Dynamics (survey). Strotz (1956) Review of Economic Studies Deckel, Lipman, Rustichini (2001, 2006) Ozdenoren, Salant and Silverman (2006)

Evidence on Impulsivity l l Self-regulatory resources play an important role in affecting many types of behaviors (Baumeister et. al, 1998; Baumeister and Ciarocco 2000; and many others, see Vohs 2006 for review. ) l Overeating l Procrastination l Underachievement Vohs and Heatherton (2000) find that dieters sitting next to a bowl of candy are subsequently less able to perform arithmetic as well as dieters who were seeted away from a bowl of candy. l Exerting self-control on one task renders it harder to exert self-control on a subsequent task.

The famous “marshmallow test. ” • Mischel and Ebbeson, with 4 year old subjects. “Here is a marshmallow for you. I have to leave the lab for 10 minutes. If you can refrain from eating the marshmallow until I return, you can have a second marshmallow. ” • Results put children into three categories: – Some children did wait for the delayed reward. • A predictor of later academic success! – Many children chose to take the lesser reward immediately. – A third group of children waited several minutes, only to end up eating the marshmallow before the researcher returned.

The Marshmallow Experiment Attention to the rewards strongly influenced the outcomes in the experiment. – Children who managed to distract themselves from the marshmallow (or other reward) were much more likely to “pass” the marshmallow test. – Follow up research found that putting the marshmallow inside a desk drawer helped the subjects become much more successful at waiting.

A Simple Theory of Decisions Under Temptation • Goal: Develop a simple framework within which the “Marshmallow” type behavior can be modeled – Ozdenoren, Salant and Silverman (2006)

A Simple Theory • An agent either consumes (d = 1) or does not consume (d = 0) a “tempting” item. – All utility associated with consuming or not consuming the item occurs at the point when the item is no longer a target for consumption • The agent consumed it • It is no longer immediately available to the agent.

Decision Under Temptation • An item is “tempting” to a consumer if her preferences for the item satisfy the following. – First, the “no consumption” (d = 0) utility depends on exposure duration according to U(0) W(t), where U(0) is a real scalar and W(t) (cost of depleted willpower) is real and monotonically increasing. – Second, utility derived from consumption is a scalar U(1) that does not vary with exposure duration and that satisfies U(1) < U(0).

Decision Under Temptation This model captures the intuition that the longer an agent is exposed to a tempting item the less satisfaction they feel in the ultimate decision not to consume it, while consuming it gives them the same pleasure regardless.

Decision Under Temptation Thus, when exposed to a tempting item for duration t, consumer j’s preferences are: V(d = 0, t) = U(0) W(t) V(d = 1, t) = U(1) The consumption decision d* is d* = 0 if V(0, t) > V(1, t); d* = 1 otherwise.

Decision Under Temptation Suppose there are two possible exposure durations, S(hort) and L(ong) Thus, W(S) < W(L). Suppose also that one knows whether she will be exposed to the tempting good for duration S or L.

Decisions Under Temptation Three cases arise: • Case 1. U(1) > [U(0) W(S)] > [U(0) W(L)]. – In this case the agent consumes the tempting good immediately and obtains utility U(1). • Case 2. [U(0) W(S)] > [U(0) W(L)] > U(1). – In this case the consumer does not consume the tempting good and obtains utility U(0) W(S) if the duration is short, and U(0) W(L) if the duration is long. • Case 3. [U(0) W(S)] > U(1) > [U(0) W(L)]. – This agent consumes the tempting good immediately and obtains utility U(1) if the known duration is L, but does not consume the good and earns utility U(0) W(S) if the duration is short.

Predictions • Assuming different individuals are characterized by different cases, the model predicts the following. (i) the frequency of tempting purchases increases as exposure duration increases (ii) some people will not purchase tempting goods even with long exposure (iii) some people will purchase tempting goods even with short exposure

Predictions (iv) If there is uncertainty regarding exposure duration the model predicts delay in consumption as observed in the marshmallow task. – Because the value of consumption is time invariant, a person with Case 3 preferences will always wait to resolve duration uncertainty prior to making their consumption (or no consumption) decision.

Testing the Predictions • Can we discover a naturally occurring economic environment within which: – Temptation plays an important role – The predictions (i) (iv) of the model can be tested

Checking Out Temptation: A Natural Experiment at the Grocery Register Daniel Houser, George Mason University David H. Reiley, University of Arizona Michael B. Urbancic, UC-Berkeley

Grocery store innovations have increased the time and attention customers spend with products. • 1800 s: General stores kept goods behind the counter. – Individual consumers presented their shopping list to the clerk. – Simple product packaging, for the clerk’s benefit only.

• 1916: Self service stores invented. – At first, cramped shelves through which customers navigated one way through a predetermined pattern. – Consumer packaging became important. • 1936: Shopping carts invented. Carts (along with automobiles) increased the feasible size of grocery purchases. Customers could spend much more time comfortably browsing. Previously, only hand carried baskets were available.

• Behavioral psychologist John Watson made a second career of consulting on product placement in stores, including “impulse items” at the checkout counter.

Predictions to Test • Assuming different individuals are characterized by different cases, the model predicts the following. (i) the frequency of tempting purchases increases as exposure duration increases (ii) some people will not purchase tempting goods even with long exposure (iii) some people will purchase tempting goods even with short exposure (iv) If there is uncertainty regarding exposure duration the model predicts delay in consumption just as observed in the marshmallow task.

We made over 2800 direct observations of customers at the checkout aisle to test predictions (i)-(iv). • During spring 2002, undergraduate research assistants watched and recorded 2827 observations of customers in grocery checkout aisles in Tucson, Arizona. • Three stores: – Store 1: a large national grocery chain, located in a middle income area of the city. (2042 observations) – Store 2: a more upscale chain store, located in a wealthier part of town. (423 observations) – Store 3: a local, independent grocery in a lower income neighborhood. (326 observations)

Observations included an array of descriptive and quantitative statistics. • Location, day of the week, time of day • Length of time spent in line (until checkout began) • Whether or not an impulse item was purchased (binary variable; multiple impulse items counted the same as a single item—at least one impulse purchase) • Some demographic data (always gender & kids, sometimes race & age)

Following are some descriptive statistics of the observations: Distribution of Sex and Purchases by Store (N = 2827) Store 1 Total Purchases Store 2 Total Store 3 Purchases Total Aggregate Purchases Total Purchases Males 872. 67* 53 (6. 1%) 155 6. 5 (4. 2%) 216. 833 20 (9. 2%) 1244. 5 79. 5 (6. 4%) Females 1169. 33 114 (9. 7%) 268 14. 5 (5. 4%) 145. 167 21 (14. 5%) 1582. 5 149. 5(9. 4%) 2042 167 (8. 2%) 423 21 (5. 0%) 362 41 (11. 3%) 2827 229 (8. 1%) Total Distribution of Observations Where Children Were Present and Purchases by Store (N = 2827) Store 1 Total Purchases Store 2 Total Store 3 Purchases Total Purchases Aggregate Total Purchases 52 6 (11. 5%) 4 0 (0%) 7 2. 5 (35. 7%) 63 8. 5 (13. 5%) With Females 141 27 (19. 1%) 24 4 (16. 7%) 22 6. 5 (29. 5%) 187 37. 5(20. 1%) Overall 193 33 (17. 1%) 28 4 (14. 3%) 29 9 (31. 0%) 250 46 (18. 4%) With Males *Note: In each of the above tables groups of customers of mixed gender were treated as an appropriately proportioned fractional sex observation. Though the percentages above suggest that females were more likely to make impulse purchases, t-tests show that there is no significant difference due to gender. Instead, this effect reflects the fact that 74. 8% of observations with children involved female customers.

As in the marshmallow test, customers often waited before picking up an impulse item in the checkout aisle. The behavior seen above is consistent with temptation theory.

The data directly suggest that time spent in line may affect the incidence of impulse purchases. Time in Line for All Observations and Given Purchase, by Store (N = 2827, times in mm: ss) Store 1 All Store 2 Given Purchase All Store 3 Given Purchase All Aggregate Given Purchase All Given Purchase Min Time 0: 03 0: 29 0: 26 0: 09 0: 21 0: 03 0: 21 Max Time 18: 35 14: 03 16: 24 7: 20 12: 42 6: 45 18: 35 14: 03 Mean Time 4: 29 5: 40 2: 28 2: 44 2: 38 3: 10 3: 56 4: 57 Mean time in line given purchase is a full minute (25. 8%) longer than the mean for all observations, suggesting that longer wait times influence positively the frequency of impulse purchases. Could the direction of causation be the opposite of what we think? No, because we measure time until the cashier begins to ring up one’s purchases. Spending time to pick up an item would not increase my wait time, though it might possibly increase the wait time of those who come after me.

Logistic regressions confirm the positive effect of time in line on the frequency of impulse purchases. FEMALE Reg III KIDS STOR 2 STOR 3 TIME FEM * TIME KIDS * TIME STR 2* TIME STR 3 * TIME FEM * KIDS Constant 0. 332 1. 134** 0. 174** 0. 015 0. 035 3. 473** 0. 316 0. 369 0. 048 0. 060 0. 065 0. 246 0. 497 1. 092** 0. 230 1. 044** 0. 220** 0. 036 0. 027 0. 113 0. 064 3. 839** 0. 318 0. 376 0. 441 0. 365 0. 052 0. 060 0. 066 0. 126 0. 090 0. 245 0. 481 1. 001* 0. 229 1. 040** 0. 220** 0. 036 0. 029 0. 112 0. 064 0. 128 3. 829** 0. 324 0. 516 0. 441 0. 366 0. 052 0. 060 0. 067 0. 126 0. 090 0. 489 0. 285 * Significant at the 5% level ** Significant at the 1% level Standard errors in italics • Dependent variable: Purchase of an impulse item (0/1). • Note the positive coefficient on time in line. • The presence of kids also tends to increase the purchase probability. • Kids and females tend to reduce the impact of time on purchase relative to males, though these effects are not statistically significant. • Store 3 has more impulse purchases. • Probit and linear-probability specifications produce qualitatively similar results.

implications for both academics and the grocery industry. l l This study quantifies the effect of increased time in line on impulse purchases. l A measurable, real-world implication of temptation theory. l Though we did not attempt to measure intention, the choice data suggest that some purchasers changed their decisions and behavior over time due to temptation. l Future research might benefit from choice data with surveys of impulseitem purchasers. These results may have concrete applications for grocers, especially since impulse items often earn stores their highest profit margins. l Stores may wish to staff checkout aisles so that customers spend slightly longer before reaching the register (though not if it drives customers to competing stores). l Since kids tend to increase impulse purchases, stores may wish to encourage the presence of children with their parents on shopping trips. l Or, can stores distinguish themselves by having checkout lanes without these items?