Save More Today or Tomorrow: The Role of Urgency in Precommitment Design

Abstract

To encourage farsighted behaviors, previous research suggests that marketers should invite consumers to precommit to adopting these behaviors “later.” However, the authors propose that people will draw different inferences from different types of precommitment offers, and that these inferences can help explain when precommitment is (and is not) effective at increasing adoption of farsighted behaviors. Specifically, the authors theorize that simultaneously offering consumers the opportunity to adopt a farsighted behavior now or later (i.e., offering “simultaneous precommitment”) may signal that the behavior is not urgently recommended; however, offering consumers the opportunity to adopt that behavior immediately and then, only if they decline, inviting them to adopt it later (i.e., offering “sequential precommitment”) may signal just the opposite. In a multisite field experiment (N = 5,196), the authors find that simultaneously giving consumers the chance to increase their savings now or later reduced retirement savings. Two preregistered lab studies (N = 5,080) show that simultaneous precommitment leads people to infer that taking action is not urgently recommended, and such inferences predict less adoption of recommended behaviors. Importantly, offering sequential precommitment increases inferred urgency, predicting greater adoption. Together, this research advances knowledge about the limits and potential of precommitment.

Keywords

precommitment inference making farsighted decisions choice architecture field experiment

Consumers often face decisions about whether and when to engage in behaviors that have immediate costs and long-term benefits (e.g., saving for retirement, updating malware protection software, undergoing preventative health screenings, receiving a vaccine). When invited to adopt such farsighted activities immediately, people often decline because the immediate costs loom large relative to the distant benefits (Frederick, Loewenstein, and O’Donoghue 2002). Marketers and policy makers seeking to encourage farsighted choices frequently offer “precommitment” as a solution: By inviting people to precommit to adopt the behavior in the future, it makes the costs feel less aversive, which extant theory and evidence suggest should increase adoption (Milkman, Rogers, and Bazerman 2009, 2010; Read and Van Leeuwen 1998; VanEpps, Downs, and Loewenstein 2016). For example, companies like Apple and Zoom invite customers who are due for a large software update to install their updates later, Wikipedia offers the option to precommit to donate later, and stickK (a popular goal-setting website) allows users to begin goal pursuit on a future date. Across hundreds of employers, precommitment to saving is offered through the Save More Tomorrow program, which allows employees to commit now to start saving for retirement in the future (Benartzi and Thaler 2013; Thaler and Benartzi 2004). While previous research suggests that precommitment should unambiguously increase overall take-up of farsighted behaviors, we propose that the effects of precommitment on the adoption of farsighted behaviors may be more nuanced, and we present evidence consistent with our theorizing.

The current research begins with the premise that there are different ways that a marketer can offer consumers the opportunity to precommit to a farsighted behavior. Previous research has focused on single-option choice sets, asking how offering the option to adopt a farsighted behavior at a delay (vs. now) impacts take-up (e.g., Rogers and Bazerman 2008). However, when precommitment is offered in practice, marketers commonly offer each consumer both the option to adopt the behavior “now” and the option to adopt the behavior “later.” Marketers do this presumably because it allows firms to satisfy consumers’ heterogeneous preferences (i.e., attracting consumers who prefer to delay in addition to those who prefer to start immediately) while expediting the adoption of offered behaviors. In this article, we propose the first theory to examine two common strategies that marketers use to offer both the precommitment option and the option to adopt the behavior immediately. We term these two strategies “simultaneous precommitment” and “sequential precommitment” (illustrated in Figure 1). We further identify a novel mechanism that helps explain the differential effects of simultaneous and sequential precommitment on farsighted decisions.

Figure 1.

Summary of the Three Choice Sets Studied in This Article.

When offering simultaneous precommitment, marketers present the option to adopt a farsighted behavior now and the precommitment option side by side. When Zoom has a software update, for instance, it offers the option to update it now or at a future date, and these options are offered simultaneously, side by side. Similarly, when Wikipedia solicits donations, it simultaneously offers consumers the options to donate “now” or “later.” Indeed, in surveys we conducted, the majority of industry professionals (73.8% of N = 229) and marketing professors (62.4% of N = 85) predicted that offering simultaneous precommitment to encourage a policy-relevant behavior—enrollment in a retirement plan—would lead individuals to save more compared with only offering the option to enroll now (see Web Appendix A for details). However, in the current research, we find that simultaneous precommitment does not necessarily increase adoption of farsighted behaviors and can even backfire, leading people to delay important behaviors they would otherwise have engaged in immediately. Such delays can be costly, as waiting longer to begin saving reduces accumulated wealth, waiting longer to update software increases the likelihood of malware attacks, waiting longer for a health screening reduces the likelihood of detecting a disease early enough to cure it, and waiting longer for a vaccine reduces the likelihood of having protection at the time of disease exposure. We theorize and show that simultaneous precommitment signals that the marketer who designed the adoption opportunity does not view the offered behavior as very urgent.

When offering sequential precommitment, marketers first give consumers the option to adopt the farsighted behavior immediately, and then, only if the initial offer is declined, do they offer the option to adopt the behavior later. For example, in one of the original implementations of the Save More Tomorrow retirement savings program, employees were only offered the option to precommit to save in the future if they had already rejected an offer to start saving immediately (Thaler and Benartzi 2004). In contrast to simultaneous precommitment, we theorize that sequential precommitment heightens urgency: By offering immediate adoption before the option to delay the action, the marketer is signaling that they prefer the action be taken sooner rather than later. We further show that this inference about urgency can help explain why sequential precommitment effectively increases adoption of farsighted behaviors.

In the remainder of this article, we first develop our theory, position it in the literature, and motivate our hypotheses. We then present evidence supporting our theory from one large-scale field experiment, one vignette-based laboratory experiment, and one incentive-compatible laboratory experiment. We end with a discussion of the implications of our research.

Theory

Previous Theories of Precommitment

Herein, when we state that people are offered “precommitment,” we mean that they are offered an option to commit now to adopt a behavior at a future point in time. We consider precommitment to be a specific type of “commitment device” (Rogers, Milkman, and Volpp 2014) because, broadly speaking, when people are offered a commitment device, they are offered the option to commit to restrict a future choice set (Ashraf, Karlan, and Yin 2006; Schwartz et al. 2014). Our research specifically examines precommitment.

Multiple streams of research have found that offering consumers a choice to precommit to a farsighted behavior increases take-up. For instance, when making choices for the future (vs. for now), people are more likely to select healthy foods (Milkman, Rogers, and Bazerman 2010; Read and Van Leeuwen 1998; VanEpps, Downs, and Loewenstein 2016), rent educational films (Milkman, Rogers, and Bazerman 2009; Read, Loewenstein, and Kalyanaraman 1999), and support policies that bolster environmental sustainability (Rogers and Bazerman 2008).

The most common explanation for this pattern is that many people tend to exhibit present bias (O’Donoghue and Rabin 1999). Present-biased consumers struggle to make farsighted decisions because they overweight the immediate costs associated with such behaviors (e.g., the taste disadvantages of healthy food, the discomfort of a vaccine or colonoscopy) while steeply discounting the future benefits (e.g., longevity). Previous research has argued that precommitment offers should be attractive to present-biased consumers; if the farsighted behavior (e.g., eating a healthy diet, receiving a shot) starts or occurs in the future, the disutility of incurring the costs will be heavily discounted—making the behavior seem less aversive (Milkman, Rogers, and Bazerman 2008). Accordingly, as long as a decent share of consumers are present-biased, offering them a chance to adopt a farsighted behavior in the future (i.e., to precommit) should, on average, increase overall take-up of the farsighted behavior.

In addition to present bias, theories about resource slack and pain of payment lead to a similar prediction that precommitment should increase adoption of farsighted behaviors. That is, people expect to have more discretionary resources (i.e., “resource slack”) in the future than they do in the present (Zauberman and Lynch 2005), and people find spending resources less painful when they have more resources (Morewedge, Holtzman, and Epley 2007). As a result, people should anticipate that, compared with adopting a farsighted behavior now, precommitting to adopt it in the future will be less painful and thus more attractive.

When a consumer is offered both the option to adopt a behavior “now” and the option to adopt the behavior “later,” we still assume that some combination of the aforementioned mechanisms identified in previous work should make the precommitment offer attractive to some extent. However, our focus is on an additional mechanism that has been neglected by extant theory about precommitment and may counterbalance these previously studied benefits.

Precommitment and Inferred Urgency

We argue that to understand when precommitment fails, it is necessary to consider people's inferences about the marketer offering the precommitment option. Generally speaking, previous research has shown that consumers make inferences about marketers’ motives and recommendations (Kardes, Posavac, and Cronley 2004), particularly based on the options they offer and the way those options are arranged (Benartzi 2001; Lieberman, Duke, and Amir 2019; Krijnen, Tannenbaum, and Fox 2017). For example, people assume that the option that marketers set as a “default” is what they recommend (Brown and Krishna 2004; McKenzie, Liersch, and Finkelstein 2006).

Extending this work, we propose that people make inferences about the urgency of marketers’ implicit recommendations. In previous marketing research, urgency has typically been defined as an objective characteristic of tasks (often referred to as “task urgency”; Zhu, Yang, and Hsee 2018). Tasks with upcoming deadlines, for instance, have greater task urgency than those with more distant deadlines. We argue that urgency can also describe a person's subjective judgment that it is better to take action sooner rather than later. Importantly, consumers may make sense of how urgent a marketer thinks it is for them to take action based on the marketer's implicit and explicit recommendations. For instance, if a financial advisor recommends that their clients start saving immediately, the clients may perceive that the financial advisor considers saving an urgent priority. In the current research, we examine how consumers make inferences about the urgency of marketers’ recommendations from the presentation of choices marketers offer. We define “inferred urgency” as the inference by a consumer that a marketer recommends adoption of a behavior sooner rather than later. Next, we theorize about how the design of a precommitment offer affects consumers’ inferred urgency and then hypothesize about the consequences of precommitment design for consumer choice.

We specifically propose that consumers make inferences about the urgency of a marketer's implicit recommendation from the order in which a precommitment option is presented. Previous research has shown that consumers hold “position-based beliefs” (Valenzuela and Raghubir 2009). Consumers assume that options presented at eye level, for instance, are placed there by the retailer because they are more popular than the options on the bottom of the shelf (Valenzuela and Raghubir 2009). Building on this work, we argue that consumers do not just hold position-based beliefs about where options are placed in a display but also make inferences based on when marketers present options in a sequence. We specifically propose that consumers will view the temporal ordering of options as an intentional decision by the marketer that signals how strongly the marketer recommends certain options relative to salient alternatives.

We argue that the two common ways marketers design precommitment offers—simultaneously and sequentially—send contrasting signals about the urgency of a marketer's recommendation. When a marketer offers consumers simultaneous precommitment, the options to adopt the behavior “now” and “later” are presented side by side in the same menu. In presenting these options side by side, the marketer does not signal a clear preference between the options. Without additional information, consumers may naturally infer that a marketer endorses both options equally (Fox, Ratner, and Lieb 2005; Tannenbaum, Fox, and Goldstein 2013). To consumers, it is as if the marketer is saying, “Doing it either now or later will suffice.” When a marketer offers consumers sequential precommitment, however, they are offering the option to adopt a given behavior “now” first, and only after their offer is rejected do they offer the option to adopt it “later.” We propose that this presentation implies an ordinal ranking of the marketer's recommendations. It would be natural for consumers to infer that the marketer is not outright endorsing the “later” option but rather treating it like a contingency plan to ensure that if consumers do not adopt what the marketer is offering now, they will at least adopt the behavior at some point in the future. To consumers, it is as if the marketer is saying, “You should do this as soon as you can!” We formally hypothesize,

H₁: Compared with not offering a precommitment option, offering simultaneous precommitment decreases consumers’ inferences about the urgency with which a behavior's adoption is recommended, whereas sequential precommitment increases the inferred urgency.

Inferred Urgency and Farsighted Decisions

When people infer that a behavior is urgently recommended, this should subsequently influence their decisions. Previous work has shown that consumers’ choices are often influenced by their inferences about marketers’ recommendations (Smith, Goldstein, and Johnson 2013). For instance, labeling health care plans “gold,” “silver,” and “bronze” conveys to consumers what marketers consider to be the best, middle, and worst plans, which alters consumers’ insurance choices, even when the labels are assigned arbitrarily (Ubel, Comerford, and Johnson 2015). When people infer that a default option is recommended by the marketer, they are typically more likely to choose that option (McKenzie, Liersch, and Finkelstein 2006). Other work has shown that when marketers communicate the urgency of a task by highlighting an upcoming deadline, people are more motivated to do it (D’Adda, Galliera, and Tavoni 2020; Zhu, Yang, and Hsee 2018).

Bridging this work, we argue that when consumers infer that a behavior is urgently recommended, they should be more likely to adopt the behavior. In the context of precommitment offers, our research focuses on inferred urgency's influence on two choice outcomes: immediate adoption and overall adoption of a farsighted behavior. Both of these outcomes have important consequences for consumer well-being, and the two outcomes together present a comprehensive evaluation of how precommitment design affects engagement in farsighted behaviors.

Immediate adoption

Immediate adoption refers to whether consumers commit to adopt a behavior immediately (i.e., choosing the “do it now” option). When studying farsighted behaviors, immediate adoption is particularly important to examine because the benefits of these behaviors typically accumulate over time. For instance, saving earlier (vs. later) in life results in greater accumulated savings, updating software sooner (vs. later) increases likelihood of stopping a malware attack, getting screening exams sooner (vs. later) increases likelihood of catching a disease early enough to cure it, and receiving a vaccine sooner (vs. later) increases the likelihood of being protected at the time of disease exposure.

When consumers infer that a behavior is urgently recommended by a marketer, they believe that the marketer recommends that they adopt the behavior sooner rather than later, which should in turn increase immediate adoption of the behavior. Given that simultaneous precommitment signals lower urgency (compared with no precommitment or sequential precommitment), and given that this lack of urgency should reduce immediate adoption, we expect simultaneous precommitment to decrease immediate adoption of farsighted behaviors.¹ More formally,

H₂: Offering simultaneous precommitment decreases the immediate adoption of farsighted behaviors compared with offering sequential precommitment or making no precommitment offer.

Overall adoption

Overall adoption refers to whether consumers commit to adopt a behavior at any point in time (i.e., choosing either the “do it now” option or the “do it later” option). When consumers infer that a marketer recommends that they promptly adopt a behavior as soon as possible, this should in turn increase overall adoption (i.e., prompting them to commit to do it, either immediately or at a future time).

The predicted impact of simultaneous precommitment (relative to no precommitment) on overall adoption is unclear due to competing mechanisms. On the one hand, simultaneous precommitment may capitalize on people's preference for delaying the adoption of farsighted behaviors (as found in previous research), which should increase overall adoption by getting people who otherwise would not sign up to choose the precommitment option. On the other hand, it reduces inferred urgency, which should curb adoption of the farsighted behavior. Therefore, we do not make predictions about the effect of simultaneous precommitment (vs. no precommitment) on overall adoption of farsighted behaviors, because the effect will depend on the relative strength of these opposing mechanisms.

However, sequential precommitment should unambiguously increase overall adoption of farsighted behaviors relative to not offering a precommitment option, because it both signals greater urgency and capitalizes on people's preference to pursue farsighted behaviors at a time delay. Further, compared with offering simultaneous precommitment, offering consumers sequential precommitment should also clearly boost overall adoption: both types of precommitment leverage people's preference for delaying the costs associated with farsighted activities, but we expect sequential precommitment to signal a greater sense of urgency than simultaneous precommitment. Therefore,

H₃: Sequential precommitment increases overall adoption of farsighted behaviors compared with making no precommitment offer or offering simultaneous precommitment.

Mediation via inferred urgency

Given our predictions that simultaneous and sequential precommitment impact consumers’ inferences about how urgently action is recommended, and that heightened inferred urgency spurs immediate and overall adoption of farsighted behaviors, we formally hypothesize,

H₄: Inferred urgency mediates the effects of simultaneous and sequential precommitment (vs. not offering a precommitment option) on both the immediate and overall adoption of farsighted behaviors.

The four hypotheses are summarized in Table 1.

Table 1.

Summary of the Theory's Key Predictions.

	Effects of SimultaneousPrecommitment	Effects of Sequential Precommitment
Measure	(vs. No Precommitment)
Inferred urgency	Decrease (H₁)	Increase (H₁)
Immediate adoption of farsighted behaviors	Decrease (H₂) because of decreased urgency (H₄)	No effect, because sequential precommitment and no precommitment are identical up to the point of the immediate adoption decision^a
Overall adoption of farsighted behaviors	Ambiguous effect, because decreased urgency (H₄) may be offset by counteracting mechanisms^b	Increase (H₃) because of increased urgency (H₄)

For more information, see footnote 1.

For more information, see the “Theory” section on overall adoption.

Theoretical Implications

Overall, the current article aims to make three main contributions. First, we contribute to research on precommitment and farsighted decision making by drawing a theoretical and practical distinction between simultaneous and sequential precommitment. Although commonly used in practice and thus worthy of systematic investigation, these types of precommitment have not previously been distinguished from each other and rigorously studied.

Second, we contribute to research on inference making (Kardes, Posavac, and Cronley 2004) and information leakage (McKenzie, Liersch, and Finkelstein 2006) by uncovering a novel, consequential inference that people draw from choice sets: the inferred urgency of the marketer's implicit recommendation. We theorize about why people form inferences about the urgency of recommendations and why inferred urgency can spur people to take prompt action. Further, we argue and show that inferred urgency can help explain when offering precommitment increases adoption of farsighted behaviors and when it does not.

Third, we present a large, real-world experimental test of precommitment, arguably one of the most commonly used “nudge” interventions. Our research suggests that a seemingly small difference in the way a popular idea is implemented (e.g., the simultaneous vs. sequential presentation of a precommitment option) can change its effects. We discuss generalizable lessons for scaling promising marketing strategies in the field.

The remainder of this article is organized as follows. We begin our investigation with a field experiment studying simultaneous precommitment, and then we present two additional well-powered, preregistered laboratory studies testing all four of our hypotheses. The results of each study are summarized in Table 2. Our preregistrations, materials, nonproprietary data, and code are available at https://researchbox.org/434.

Table 2.

Summary of Key Results.

Study	Sample and Measures	Details	NoPrecommitment	SimultaneousPrecommitment	SequentialPrecommitment	Mediation via Inferred Urgency
Study 1	Sample	N = 5,196 university employees	n = 2,600	n = 2,596
	Immediate adoption	Whether employees enrolled in a savings plan immediately (=1) or not (=0)
	Immediate adoption		.075 (.26)	.059 (.23)
	Overall adoption	Whether employees enrolled in a savings plan immediately or at a delay (=1) or notat all (=0)
	Overall adoption		.123 (.33)	.116 (.32)
Study 2	Sample	N = 2,682 participants from Prolific	n = 901	n = 895	n = 886
	Inferred urgency	Inference about the urgency of the employer’s recommendation (1–7 response scale)
	Inferred urgency		4.88 (1.52)	4.55 (1.63)	5.22 (1.30)
	Immediate adoption	The number of employee benefits participants enrolled in immediately (ranging from 0–3)
	Immediate adoption		1.86 (.78)	1.49 (.89)	1.93 (.85)	✓
	Overall adoption	The number of employee benefits participants enrolled in immediately or at a delay (ranging from 0–3)
	Overall adoption		1.86 (.78)	1.84 (.81)	2.14 (.82)	✓
Study 3	Sample	N = 2,398 participants from MTurk	n = 806	n = 794	n = 798
	Inferred urgency	Inference about the urgency of the researcher’s recommendation (1–7 response scale)
	Inferred urgency		3.87 (1.79)	3.57 (1.89)	4.23 (1.81)
	Immediate adoption	Whether participants enrolled in a financial assessment immediately (=1) or not (=0)
	Immediate adoption		.320 (.47)	.282 (.45)	.331 (.47)	✓
	Overall adoption	Whether participants enrolled in a financial assessment immediately or at a delay (=1), or not at all (=0)
	Overall adoption		.320 (.47)	.480 (.50)	.574 (.49)	✓

Study 1: How Simultaneous Precommitment Affects Retirement Savings Decisions in the Field

To test our theory's main predictions about the impact of offering simultaneous precommitment in the field, we report on the results of two conditions from an experiment involving real savings decisions.² A companion paper (Beshears et al. 2021) compares a third condition from this field experiment with one of the conditions examined in our article to explore a separate research question (see additional details in the following sections and Web Appendix B).

Methods

Four U.S. universities (labeled Universities A, B, C, and D to preserve their anonymity) collaborated with us on our field experiment.³ Each university began by identifying a retirement savings plan in which it hoped to increase employees’ contributions. All universities then identified employees who were not enrolled in this “targeted plan” and therefore had a contribution rate of zero, encouraging them to sign up to save in the targeted plan. One university (University D) also identified employees who were contributing to the targeted plan, but not at the level necessary to take full advantage of their employer's matching contributions; these employees were encouraged to save more, rather than to start saving. Table 3 presents more information about the targeted plans offered by the four universities, and Table W2 in Web Appendix C details other (nontargeted) savings plans.

Table 3.

Descriptions of Targeted Plans.

University	Eligibility	Employer Contributions
A	All employees on the university's payroll with FICA deductions	None
B	All employees whose annual contribution limit to the targeted plan is at least $200	None
C	All paid employees OR students with a stipend	None
D	Eligibility for Employee Contributionsi. Regular full-time staff (with monthly or weekly pay cycles) ORii. Full-time faculty and academic support staff in a benefits-eligible title ORiii. Limited-service staff scheduled to work at least 35 hours per week for a minimum of 9 months per year (with monthly or weekly pay cycles)	Automatic Employer Contribution Rates (Regardless of Whether the Employee Contributes)i. 1.5% (employee age < 30)ii. 3% (employee age 30–39)iii. 4% (employee age ≥ 40)
D	Eligibility for Employer ContributionsAll employees who are eligible for employee contributions (described above), are age 21 or older, and have at least one year of prior service	Matched Employer ContributionsDollar-for-dollar match on employee contributions up to 5% of employee's salary

Notes: This table is identical to the one presented in Beshears et al. (2021) because it describes the same retirement plans in the same universities.

One retirement plan record keeper shared by all four universities sent out mailings in early October of 2013 to university employees’ homes. The mailings provided employees with an opportunity to increase their savings contributions by filling out and mailing back a simple form on a prestamped, preaddressed postcard. If an employee checked a box indicating they wanted to save and then signed and returned the postcard, that employee would be enrolled in the plan at a preselected contribution rate with their contributions allocated to a preselected fund. At all four universities, the preselected fund on the mailing was a life cycle fund, which provided a diversified portfolio with a mixture of equity, bond, and money market funds tailored to the employee's age. The preselected contribution rate was 3% of the employee's pay for Universities A–C and 5% for University D. If an employee who was already contributing to the targeted plan elected to save more (only relevant to University D), their contributions would increase to the preselected rate with the contributions allocated according to their existing asset allocation.

Employees were randomly assigned to different experimental conditions, which determined the exact mailing they received (for additional details on our stratified random assignment process, see Figure 2 and Web Appendix B). In this article, we only analyze employees who were randomly assigned to receive either a No Precommitment mailing or a Simultaneous Precommitment mailing (N = 5,196; M_age = 43.10 years, SD_age = 12.05; 52.25% female, 47.75% male). The No Precommitment mailing offered employees the opportunity to immediately increase their contribution rate to the targeted plan. The Simultaneous Precommitment mailing offered employees the opportunity to increase their contribution rate to the targeted plan either immediately or after a time delay (e.g., “in two months”) ranging from two to six months. Both mailings are displayed in Figure 3.

Figure 2.

Randomization.

Figure 3.

Mailings from Field Experiment (Study 1).

The experiment included another group of employees who were randomly assigned to receive a different type of Simultaneous Precommitment mailing, which offered them an opportunity to increase their contribution rate to the targeted plan either after a labeled temporal landmark (e.g., their birthday, the first day of spring, Thanksgiving, Valentine's Day) or immediately. A companion paper (Beshears et al. 2021) compares the enrollment decisions of employees who received these distinctive Simultaneous Precommitment mailings linked with temporal landmarks and employees who received the standard Simultaneous Precommitment mailings studied here. The objective of Beshears et al. (2021) was to test whether inviting people to boost their contribution rate after a “fresh start” date (e.g., a birthday, the first day of spring; following Dai, Milkman, and Riis [2015]) increases savings over and above inviting people to increase contributions at an equidistant future time point (e.g., in two months). Beshears et al. do not report results from the No Precommitment condition studied here because their article solely explores the effect of inviting savings following fresh start dates and not the effects of offering precommitment. See Web Appendix B for more information about Beshears et al.

This experiment's randomization was stratified by university because the universities varied on important features such as the targeted plans’ characteristics. Randomization was also stratified by birth month (within each university) because employees’ birthdays partially determined which mailing they received; only those whose birthday fell within November 2013–March 2014 had the opportunity to be randomized to receive the option to save more after their birthday.

In all conditions, mailing recipients who wanted to increase their contribution rates had to send back their response card by November 1, 2013. If they chose to save at a higher rate immediately via the mailer, their contribution rate would increase to the preselected rate in November 2013. If they chose to save more at a delay (e.g., in five months), their contribution rate would automatically increase to the preselected rate at the predetermined time (e.g., in March 2014).

Data

Our university partners first pulled a cross-sectional snapshot of information about all plan-eligible employees in August 2013, including their current contributions to the targeted plan, contributions to all other nontargeted savings plans, birth date, hire date, termination date, salary, and position (faculty vs. staff). Our conditions are reasonably well balanced across baseline employee characteristics, with the only statistically significant difference being that the mean salary of employees in the No Precommitment condition (M = $56,505.19, SD = $35,234.21) is slightly less than that of employees in the Simultaneous Precommitment condition (M = $58,505.26, SD = $36,111.88; p = .043; see Table W1 in Web Appendix B). To ensure that the slight imbalance detected on this dimension is accounted for, our regressions control for baseline employee characteristics, including salary decile.

After the study concluded, our university partners provided information on each employee's contributions to the targeted plan and all other retirement savings plans as well as their pay for each pay cycle from August 2013 through June 2014.⁴ We measured the impact of our mailing by observing changes in employees’ retirement plan contributions (made by mail, by phone, or online).

Variables

To comprehensively measure the effects of offering simultaneous precommitment on savings, we created three outcome variables, described next.

Immediate adoption

Immediate adoption is a binary variable that takes on a value of one for people who increased their contribution rate to the targeted plan immediately after receiving our mailing and zero for others. We constructed this variable by examining whether an employee's contribution rate in November 2013 (the first month our mailings could have triggered increased contributions) was higher than their rate in September 2013 (the month right before our mailings were sent out).⁵

Overall adoption

Overall adoption is a binary variable that takes on a value of one for people who increased their contribution rate to the targeted plan by the end of our study period and zero for others. We constructed this variable by examining whether an employee's contribution rate in June 2014 (the last month in which we received data on employees’ contributions and pay) was higher than their contribution rate in September 2013.⁶

Table 5.

Robustness Checks for the Effect of Offering Simultaneous Precommitment in Study 1.

Robustness Check		Results
Robustness Check		Immediate Adoption	Overall Adoption	Average Savings Rate
Measuring Decisions with Cards Mailed back to the One University that Tracked Responses. Here, we measure immediate and overall adoption by analyzing the choices made on returned response postcards at the one university that shared this information (N = 2,029). See Table W8.		−.016*(.009)	−.008(.009)
Measuring Overall Adoption Differently. To measure overall adoption in a way meant to capture employees’ direct responses to our mailers, we only counted someone as enrolling if the first time their contribution rate increased (relative to their rate in September 2013) matched one of the month(s) offered to them in our mailing. See Table W9.			−.003(.007)
Running Logistic and Fractional Logit Regressions. Here, we rely on logistic regression to analyze immediate and overall adoption and fractional logistic regression to analyze average savings rates. The coefficients reported are in terms of log odds ratios (and thus have different interpretations than those from the other robustness checks). See Table W10.		−.359***(.118)	−.099(.092)	−.163**(.081)
Dropping Those with Missing Data. Here, we drop data from employees who have missing data for salary or contributions in all months in the study period (rather than assuming zeroes). See Table W11.		−.020***(.007)	−.010(.009)	−.0015**(.001)
Dropping Those with Missing Data in Key Months. Here, we drop data from employees who have missing data for salary or contributions in one of the key months used to calculate immediate and overall adoption (rather than assuming zeros). See Table W12.		−.021***(.008)	−.009(.010)
Varying the Cutoff for a Contribution Rate Increase. Here, we use different thresholds to measure immediate and overall adoption. Since we constructed the adoption variables by assessing whether increases in contribution rates occurred, here we ensure our results are not spuriously driven by how we rounded contribution rates. See Table W13. We counted a contribution rate as having increased if the increase was…	≥.1% of salary	−.020***(.007)	−.009(.009)
	≥.01% of salary	−.021***(.007)	−.008(.009)
Including Limited Controls. Here, we control for the interaction between university and birth month (due to the stratified random assignment employed) but drop other controls (i.e., gender, age, tenure, salary, and faculty status). See Table W14.		−.017**(.007)	−.007(.009)	−.0012*(.001)

*p ≤ .10. **p ≤ .05. ***p ≤ .01.

Notes: Each row corresponds to a robustness check testing the effects of simultaneous precommitment (vs. no precommitment). Unless otherwise specified, the models are similar to the primary model specified in the “Analysis Strategy” section. In the right three columns, we report the coefficients from the regressions in the relevant row, with heteroskedasticity-robust standard errors in parentheses and significance indicated by asterisks. Full descriptions of each robustness check are reported in Web Appendix E.

Average savings rate

To capture employees’ cumulative retirement savings (adjusted for their salary) during our study period, we calculated every employee's average savings rate by taking the total number of dollars the employee contributed to the targeted plan from November 2013 through June 2014 and dividing it by the employee's total pay during the same period. This outcome variable ranges from 0 to 1, representing the percentage of an employee's total pay that was contributed to the targeted savings plan during our study period.

Analysis Strategy

To estimate the causal impact of the Simultaneous Precommitment mailing (compared with our No Precommitment mailing that only invited people to save now), we relied on the following ordinary least squares (OLS) regression specification:

outcom e_{i} = α + β simultaneous precommitmen t_{i} + γ^{'} X_{ij} + Σ_{j} (δ_{j} I [universit y_{i} = j] + {ζ^{'}}_{j} X_{ij} I [universit y_{i} = j]) + ε_{i},

(1)

where i indexes employees and j indexes universities. We estimated this regression once with each of the aforementioned outcome variables. The coefficient on the indicator for simultaneous precommitment is the estimate of the causal impact of the Simultaneous Precommitment condition relative to the No Precommitment condition. To increase statistical power, we estimated a single treatment effect across universities instead of separate treatment effects for each university, but this decision does not invalidate the interpretation of the coefficient as a causal effect, since randomization was stratified by university. X_ij is a vector of controls: gender, age decile, tenure decile, salary decile, faculty status, and birth month, where decile breakpoints are calculated separately for each university. I[university_i = j] is an indicator variable that takes a value of one when employee i is associated with university j and a value of zero otherwise.⁷ The δ_j and ζ_j coefficients allow the intercept term and the coefficients on the control variables to vary by university, accounting for differences across universities in their average responsiveness to our mailings and differences across universities in the relationship between the control variables and the outcome variable. Given that retirement savings decisions are largely determined by socioeconomic circumstances, controlling for employees’ demographics, income, and employer characteristics in analyses of interventions designed to increase savings rates can dramatically enhance statistical power and is consistent with standard practice in retirement savings research (for a review, see Choi [2015]; see Web Appendix D for more information). We report heteroskedasticity-robust standard errors. For the binary outcomes, we report linear probability regressions here rather than logistic regressions for the ease of interpretation, but the results do not substantively differ when we estimate logistic regressions (see this and all other robustness checks summarized in Table 5 and described in detail in Web Appendix E).

Results

Immediate adoption

Consistent with H₂, employees’ probability of immediately increasing their contributions to the targeted plan was 1.9 percentage points lower in the Simultaneous Precommitment condition than in the No Precommitment condition (p = .006; Table 4, Model 1). In terms of its relative effect, this represents a 25.3% decrease, relative to the 7.5% of employees who immediately increased their contribution rate to the targeted plan in the No Precommitment condition.

Table 4.

The Effect of Offering Employees Simultaneous Precommitment in Study 1.

	Model 1	Model 2	Model 3
	Immediate Adoption	Overall Adoption	Average Savings Rate
Simultaneous precommitment	−.019***	−.009	−.0014**
	(.007)	(.009)	(.001)
University FEs	Yes	Yes	Yes
Controls	Yes	Yes	Yes
University FEs × controls	Yes	Yes	Yes
R-squared	.07	.08	.11
Observations	5,196	5,196	5,196

*p ≤ .10. **p ≤ .05. ***p ≤ .01

Notes: Model 1 reports an OLS regression where the dependent variable is a binary variable reflecting whether an employee immediately increased their contribution rate to the targeted savings plan. Model 2 reports an OLS regression where the dependent variable is a binary variable reflecting whether an employee increased their contribution rate to the targeted savings plan by the end of our study period. Model 3 reports an OLS regression where the dependent variable is an employee's average savings rate in the targeted savings plan during our study period. Standard errors robust to heteroskedasticity are reported in parentheses. Controls include gender, age decile, tenure decile, tenure decile, salary decile, faculty status, and birth month.

Overall adoption

Importantly, when we look at the full study period through June 2014, there was not a statistically significant difference between conditions in employees’ probability of increasing their contributions to the targeted savings plan (p = .301; Table 4, Model 2).

Given that simultaneous precommitment decreased the number of people who immediately increased their contribution rate to the targeted saving plan but did not significantly affect the total number of savers, we can conclude that simultaneous precommitment led some people to delay saving who otherwise would have started saving immediately.⁸ When people delay saving, they save less overall than they would have if they had started saving earlier. We next formally tested whether offering simultaneous precommitment ultimately reduced accumulated savings.

Average savings rate

Table 4, Model 3 indicates that receiving the Simultaneous Precommitment mailing (rather than the No Precommitment mailing) decreased average saving rates to the targeted plan by .14% of pay (p = .046). In relative terms, receiving the Simultaneous Precommitment mailing caused a 16.5% decrease in savings in targeted plans during our study period compared with the No Precommitment mailing (for which the average savings rate was .85% of employee pay).

In additional models reported in Web Appendix E, we examined how the treatment effect on each of the outcome variables varied by university. The effects of simultaneous precommitment on both immediate and overall adoption did not significantly vary by university, though the decrease in average savings rates was more prominent in two universities.

Discussion

Contrary to previous research demonstrating the benefits of precommitment, we find in a large field experiment conducted with four employers that inviting consumers to save more “now or later” (i.e., offering simultaneous precommitment) did not increase the share of people contributing to a retirement savings plan. In fact, offering simultaneous precommitment (relative to simply inviting consumers to save now) on average decreased overall retirement wealth because some people opted to delay increasing their contribution rates to the savings plan and thus saved over a shorter time horizon than they would have otherwise.

As mentioned previously, in a companion paper (Beshears et al. 2021), the Simultaneous Precommitment condition from this field experiment is used in conjunction with an additional experimental condition not analyzed here to examine the effects of offering precommitment shortly after fresh start dates. That paper finds that offering precommitment after fresh starts (e.g., birthdays, the first day of spring) increases overall adoption and average savings rates compared with offering precommitment after an equidistant future time point that is not associated with such a temporal landmark. The primary contribution of Beshears et al. (2021)—to demonstrate the impact of linking opportunities to save with dates that feel like fresh starts—is theoretically distinct from the key contribution of our Study 1, which is to show the impact of simultaneous precommitment (vs. not offering precommitment). For a summary of how Beshears et al. (2021) and Study 1 in the current article differ in terms of their conditions, research questions, and contributions, see Table W15 in Web Appendix F.

Next, we present the results of a posttest to examine whether our theory about inferred urgency could potentially explain why the precommitment offer in our field experiment backfired.

Posttest: Initial Evidence on Inferred Urgency

We theorize that when an immediate enrollment option and a precommitment option were presented side by side in our field experiment (i.e., via simultaneous precommitment), employees may have inferred that their university's human resources department did not urgently recommend retirement savings. In a preregistered online experiment (N = 1,499 Amazon Mechanical Turk [MTurk] workers), we confirmed that the Simultaneous Precommitment mailing from our field experiment conveyed a less urgent recommendation to save than our No Precommitment mailing (p < .001; for more details on this posttest, see Web Appendix G). In the same posttest, we also confirmed that people believed a Sequential Precommitment offering—that is, sending a second mailing inviting employees to save later only if they neglected an initial mailing with the offer to save immediately—conveyed a more urgent recommendation to save than either our No Precommitment mailing (p = .060) or the Simultaneous Precommitment mailing (p < .001).

Together, the findings from this posttest offer tentative support for a mechanism that might explain the results we observed in the field. That is, simultaneous precommitment may have reduced immediate adoption of savings and failed to increase overall adoption because the mailing signaled that saving was not urgently recommended. This posttest also reveals that offering sequential precommitment signals that saving is highly urgent. We next present two large laboratory experiments designed to deductively test our full theory about how simultaneous and sequential precommitment may differentially impact the inferences consumers draw about the urgency of marketers’ recommendation and, in turn, consumers’ farsighted decisions.

Study 2: Inferred Urgency and Farsighted Decisions

We conducted a preregistered laboratory experiment in which participants decided (hypothetically) whether and when to enroll in three different benefits programs offered by a new employer. This study tests all four of our hypotheses.

Methods

As preregistered, we recruited workers on Prolific who were fully employed at a firm other than Prolific and passed one brief attention check. A total of 2,682 participants satisfied these selection criteria and completed the study (M_age = 34.7 years, SD_age = 9.1; 58.5% male, 40.3% female, 1.1% nonbinary, .1% preferred not to say).

All participants were asked to imagine that they were offered three benefits programs as a new, full-time employee at Company X: a retirement savings plan, a life insurance program, and a health savings account. They were told that all programs were optional, and money would be deducted from their take-home pay for each program they enrolled in. Participants were required to correctly answer two comprehension check questions before proceeding, and they were allowed to keep trying until they got these comprehension check questions right. Participants were then randomly assigned to one of three conditions: the No Precommitment condition, the Simultaneous Precommitment condition, or the Sequential Precommitment condition.

In the No Precommitment condition, participants only had the option to enroll in each benefits program immediately. Specifically, they read: “If you check the ‘Enroll now’ box to enroll in a program, Company X will start providing you with the given benefit now and begin deducting from your paycheck as soon as possible.” Then they indicated which program(s) they would enroll in by marking the corresponding checkbox(es).

In the Simultaneous Precommitment condition, participants had the option to enroll in each benefits program either immediately or in six months. The instructions explained, “You can choose ‘Enroll now,’ which means Company X will start providing you with the relevant benefit now and begin deducting from your paycheck as soon as possible. Or you can choose ‘Enroll in 6 months,’ which means Company X will start providing you with the relevant benefit in 6 months and begin deducting from your paycheck in 6 months.” After reading these instructions, participants indicated which program(s) they would enroll in by marking the corresponding checkbox(es). For any programs they selected, they then decided when to enroll (either now or in six months).

In the Sequential Precommitment condition, participants were first given the option to enroll in each program immediately, and then for the programs they did not enroll in immediately, they were offered the option to enroll in six months. The condition looked identical to the No Precommitment condition through the first (immediate) enrollment decision. Participants who did not immediately enroll in all three programs were next told, “Imagine that the day after you submitted your enrollment decisions, Company X follows up and sends you another online enrollment form.” Participants were then given the option to enroll in the remaining programs in six months by checking the corresponding checkbox(es).

Next, participants in all conditions answered two questions about Company X: “To what extent do you think Company X recommends that employees enroll in the benefits programs as soon as they can?” and “To what extent do you think Company X urgently recommends that employees enroll in the benefits programs” (1 = “Not at all,” and 7 = “Very much”). We adapted these items from previous research on information leakage (McKenzie, Liersch, and Finkelstein 2006) and task urgency (Zhu, Yang, and Hsee 2018) to assess participants’ inferences about how urgently Company X recommended they sign up for the benefits programs. The two items were collapsed into a single measure of inferred urgency because they are highly correlated (r = .70, p < .001).

Finally, we included a set of questions to assess whether decision difficulty could be an alternative mechanism for our predicted results. Specifically, we adapted four items from an existing decision difficulty scale (Goodman et al. 2013): “To what extent [did you find the choice difficult/were you overwhelmed/were you frustrated/were you annoyed] when you were making your enrollment decision?” (1 = “Not at all,” and 7 = “Very much”). The items are highly correlated (α = .85) and were averaged into one measure of decision difficulty. To establish the discriminant validity of these two mechanism measures, we used an exploratory factor analysis and confirmed that the four decision difficulty items loaded on a separate factor than the two urgency items (see Web Appendix H for details). We also checked that the composite score of decision difficulty and that of inferred urgency are only correlated at r = .15. At the end of our study, participants were asked about their age, gender, education, and income.

We focus on two preregistered outcome variables in this study. The first is immediate adoption, which measures the number of benefits programs participants elected to enroll in immediately (i.e., by selecting the “Enroll now” option). The second outcome of interest is overall adoption, which measures the total number of benefits programs participants elected to enroll in (i.e., by selecting either the “Enroll now” or “Enroll in 6 months” option).

Results

For analyses that include all three conditions, we relied on OLS regressions with heteroskedasticity-robust standard errors where the key predictors are indicators for the Simultaneous Precommitment and Sequential Precommitment conditions, with the No Precommitment condition serving as the reference group. All mediation analyses use 5,000 bootstrapped samples to estimate 95% confidence intervals (CIs) around the indirect effects.

Inferred urgency

Providing support for H₁, compared with those in the No Precommitment condition (M = 4.88, SD = 1.52), participants in the Simultaneous Precommitment condition rated Company X's implicit recommendation to enroll in its benefits programs as less urgent (M = 4.55, SD = 1.63; b = −.33, p < .001, d = .21), whereas participants in the Sequential Precommitment condition rated Company X's recommendation to enroll as more urgent (M = 5.22, SD = 1.30; b = .34, p < .001, d = .24).

Immediate adoption

Confirming H₂ and replicating the results from our field experiment, participants in the Simultaneous Precommitment condition signed up for fewer benefits immediately (M = 1.49, SD = .89) than did participants in the No Precommitment condition (M = 1.86, SD = .78; b = −.37, p < .001, d = .44). Moreover, consistent with H₄, this reduction in immediate adoption was significantly mediated by a drop in the inferred urgency of the recommendation to enroll (b = −.009; 95% CI = [−.019, −.001]).

Overall adoption

Consistent with the results of our field experiment, there was not a statistically significant difference in overall adoption between the Simultaneous Precommitment condition (M = 1.84, SD = .81) and the No Precommitment condition (M = 1.86, SD = .78; b = −.017, p = .659, d = .02). Inferred urgency significantly and negatively mediated the relationship between simultaneous precommitment and overall adoption (b = −.015, 95% CI = [−.026, −.006]), providing support for H₄. These results are in line with our theory that offering people a chance to enroll “now or later” (simultaneous precommitment) decreases inferred urgency, which curbs overall adoption of a farsighted behavior.⁹

In support of H₃, participants in the Sequential Precommitment condition (M = 2.14, SD = .82) signed up for more benefits on average than did participants in the No Precommitment condition (M = 1.86, SD = .78; b = .28, p < .001, d = .35). Inferred urgency significantly and positively mediated this effect (b = .013, 95% CI = [.003, .025]), suggesting that sequential precommitment increases inferred urgency, which predicts greater overall adoption (offering further support for H₄).

Using a Wald test, we confirmed that participants in the Sequential Precommitment condition enrolled in more benefits on average than did participants in the Simultaneous Precommitment condition (p < .001, d = .37), consistent with H₃. We also confirmed that inferred urgency significantly and positively mediated this difference (b = .019, 95% CI = [.002, .037]), further supporting H₄. In other words, compared with the Simultaneous Precommitment condition, people may have enrolled in more benefits programs in the Sequential Precommitment condition in part because they inferred that adoption was more urgently recommended.

When we added decision difficulty as another potential mediator in all of the aforementioned mediation models, inferred urgency always remained a significant mediator. See detailed results of these multimediator models in Web Appendix I.

Discussion

Study 2 presents support for our complete theory and tests all four of our hypotheses. Offering simultaneous precommitment (i.e., an invitation to enroll in benefits “now or later”) decreases the inferred urgency of adopting farsighted behaviors, whereas offering sequential precommitment (i.e., an invitation to enroll in benefits “later” only if people do not enroll “now”) increases inferred urgency (H₁). Furthermore, simultaneous precommitment decreases immediate adoption of farsighted behaviors (H₂) but fails to increase overall adoption (replicating the findings from our field experiment). Meanwhile, sequential precommitment increases overall adoption of farsighted behaviors (H₃). Importantly, inferred urgency significantly mediates these effects (H₄), helping explain the divergent impact of different forms of precommitment on consumer choice.¹⁰

Study 3: An Incentive-Compatible Test of Our Full Theory

In Study 3, we aimed to conceptually replicate Study 2 with an incentive-compatible design in a distinct context. Specifically, we invited people to take a real, ten-minute financial well-being assessment and tested our full theory by examining whether and when they chose to take it.

This study also aims to reconcile our findings with previous literature by measuring an additional mechanism that should make precommitment attractive based on extant theory. Specifically, previous work suggests that the immediate costs of adopting a farsighted behavior should feel less aversive when people contemplate taking up the behavior later because they steeply discount future costs (Frederick, Loewenstein, and O’Donoghue 2002) and expect to have more resources in the future (Zauberman and Lynch 2005). In this sense, taking the financial well-being assessment in our study should feel less costly when precommitment is an option, regardless of how it is offered. We operationalized this mechanism with a measure of perceived convenience.

Methods

For this preregistered study, we recruited 2,398 MTurk participants (M_age = 39.73 years, SD_age = 12.06; 50.5% male, 49.0% female, .1% nonbinary, .3% preferred not to say) who passed an attention check. To provide a cover story for the purpose of the study, we first asked participants to take a financial knowledge test (Knoll and Houts 2012). Participants next reported their employment status, age, gender, education, and income. Then, we offered them an opportunity to take an optional, unpaid financial well-being assessment. We explained,

In collaboration with Dr. [anonymized], a [university affiliation] professor and world expert on financial decision making, we have prepared an assessment that will provide feedback on your financial well-being and offer scientific tips for improving your financial future. Completing the assessment will take about 10 minutes. It is voluntary and won't affect your pay. But we hope that taking the assessment will be worth your time in the long run.

At this point, participants were randomly assigned to one of three conditions. In the No Precommitment condition, participants were invited to take the financial assessment immediately. In the Simultaneous Precommitment condition, participants were invited to take the financial assessment and given the choice to either complete it now or in one week. In the Sequential Precommitment condition, participants were invited to take the assessment immediately. If they declined, participants were then invited to take the assessment in one week. All participants who chose to take our assessment received a real financial well-being assessment at the time they elected.

After participants made their choice(s), they responded to additional questions. First, we asked participants “To what extent do you think we urgently recommend that you take the financial assessment?” (1 = “Not at all,” and 7 = “Very much”) to measure inferred urgency.

We also asked participants to rate how inconvenient it would be to take the assessment (1 = “Not at all inconvenient,” and 7 = “Very inconvenient”). We reverse-coded this measure and included “perceived convenience” as a competing mediator in each of the reported mediation models (for the full results, see Figure 4 and Web Appendix L).

Figure 4.

Multimediator Models for Overall Adoption (Study 3).

We focus on two preregistered outcome variables in this study. Immediate adoption is a measure of whether a participant elected to take the financial well-being assessment immediately (by selecting the “now” option). Overall adoption is a measure of whether a participant ever enrolled (by selecting the “now” option or the “in 1 week” option).

Results

Inferred urgency

Consistent with H₁, participants inferred that our implicit recommendation that they take the assessment was less urgent in the Simultaneous Precommitment condition (M = 3.57, SD = 1.89; b = −.30, p = .002, d = .16) and more urgent in the Sequential Precommitment condition (M = 4.23, SD = 1.81; b = .36, p < .001, d = .20) than it was in the No Precommitment condition (M = 3.87, SD = 1.79).

Immediate adoption

Offering some support for H₂, participants in the Simultaneous Precommitment condition were marginally less likely to immediately enroll in the assessment (28.2%) than were participants in the No Precommitment condition (32.0%; b = −.038, p = .098). In terms of its relative effect, this represents an 11.9% decrease in immediate adoption. In support of H₄, this marginal negative effect was mediated by inferred urgency (b = −.011, 95% CI = [−.020, −.004]). This result is consistent with our theory that simultaneous precommitment may reduce immediate adoption of farsighted options because it decreases inferred urgency.

Overall adoption

Offering simultaneous precommitment (inviting people to take the assessment “now or later”) resulted in greater overall adoption of the assessment (48.0% enrolled) than not offering precommitment (32.0% enrolled; b = .160, p < .001; representing a 50.0% relative increase). In support of H₄, inferred urgency negatively mediated the relationship between simultaneous precommitment and overall adoption (b = −.014, 95% CI = [−.025, −.005]), as illustrated in Figure 4, Panel A. This suggests that offering consumers simultaneous precommitment decreases inferred urgency, which may curb take-up of farsighted activities like completing a financial well-being assessment.

Confirming H₃, we found that participants in the Sequential Precommitment condition were more likely to enroll in the assessment (57.4% enrolled) than were participants in the No Precommitment condition (32.0% enrolled; b = .254, p < .001). In terms of its relative effect, this represents a 79.4% increase in overall adoption. Again, supporting H₄, inferred urgency positively mediated this effect (b = .014, 95% CI = [.006, .023]), as illustrated in Figure 4, Panel B. These results are consistent with our theory that sequential precommitment may increase overall adoption of a farsighted behavior because it signals heightened urgency.

In addition, a Wald test confirmed that participants in the Sequential Precommitment condition were more likely to enroll in the assessment than were participants in the Simultaneous Precommitment condition (p < .001; representing a 19.6% relative increase), which supports H₃. In addition, we confirmed that the difference in overall adoption between the Sequential Precommitment and Simultaneous Precommitment conditions was also mediated by inferred urgency (b = .028, 95% CI = [.018, .040]), in line with H₄.

Perceived convenience

Consistent with predictions from previous research, as shown in Figure 4, offering precommitment increased the perceived convenience of taking the assessment, regardless of whether the precommitment offer was simultaneous (b = .37, p < .001) or sequential (b = .29, p = .001). When participants perceived greater convenience, they were more likely to enroll in the assessment (Simultaneous Precommitment: b = .09, p < .001; Sequential Precommitment: b = .08, p < .001). Perceived convenience helps explain why the precommitment offers increased enrollment in the assessment (Simultaneous Precommitment indirect effect: b = .034, 95% CI = [.018, .050]; Sequential Precommitment indirect effect: b = .024, 95% CI = [.010, .040]). Of particular importance, perceived convenience did not differ between the two precommitment designs (p = .36), and thus, it cannot explain why sequential precommitment resulted in greater overall adoption than simultaneous precommitment (see Web Appendix L for details).

Follow-through behavior

Finally, we measured whether participants actually completed the optional financial well-being assessment. Compared with participants in the No Precommitment condition (in which 9.6% completed the assessment), participants were more likely to complete the assessment if they were assigned to the Sequential Precommitment condition (18.8% completed it; b = .092, p < .001) or the Simultaneous Precommitment condition (15.5% completed it; b = .059, p < .001). A Wald test confirmed that participants in the Sequential Precommitment condition were marginally more likely to finish the assessment than participants in the Simultaneous Precommitment condition (p = .061).

Discussion

In an incentive-compatible experiment, we again found support for our theory that inferred urgency shapes the way different forms of precommitment influence consumer choice. Notably, unlike in our field experiment and in Study 2, where offering simultaneous precommitment did not have a significant effect on overall adoption, in this study offering simultaneous precommitment increased overall adoption of a financial well-being assessment compared with not offering precommitment. This may be because the positive mechanisms that extant theory predict to make precommitment offers attractive had a strong influence in this study. Specifically, as shown previously, participants in the Simultaneous Precommitment condition found taking a financial assessment to be more convenient than did participants in the No Precommitment condition (see Figure 4). Our evidence suggests that the lack of urgency signaled by the simultaneous precommitment offer still curbed take-up of the financial well-being assessment, but the added convenience of this offer mattered more to participants. As a result, simultaneous precommitment had a positive total effect on the take-up of the assessment in this study. This suggests that accounting for the relative magnitudes of various competing mechanisms is necessary to predict the main effect of simultaneous precommitment on overall adoption of farsighted behaviors. Sequential precommitment, however, increases adoption by harnessing both convenience and heightened urgency. Consequently, sequential precommitment has an unambiguously positive effect on the overall adoption of farsighted behaviors and produces more adoption of such behaviors than does simultaneous precommitment.

General Discussion

Offering precommitment—that is, offering people a choice to commit now to do something later—has previously been theorized to uniformly increase the adoption of farsighted behaviors. In this article, we reexamine this assumption. We develop a theory about how the design of precommitment offers can lead consumers to draw inferences about the urgency of a marketer's recommendation to act, which helps determine when precommitment will and will not promote the adoption of farsighted behavior.

In a field experiment (Study 1), we show that, contrary to experts’ predictions, simultaneously offering consumers the option to start saving now and the option to precommit to save in the future (i.e., simultaneous precommitment) causes people to save less money over our study period than only offering them the option to start saving now. This is because offering simultaneous precommitment reduces the number of people who save immediately, without causing more people to save overall. Two preregistered online experiments (Studies 2 and 3) support our theoretical account of why simultaneous precommitment reduces total saving in our field experiment and, conversely, why offering sequential precommitment has been shown to increase savings (Thaler and Benartzi 2004). Specifically, simultaneous precommitment decreases inferred urgency, which helps explain why it reduces immediate adoption and may not increase overall adoption. Moreover, offering sequential precommitment—that is, inviting the future adoption of a farsighted behavior only after people decline to adopt it immediately—increases inferred urgency, which helps explain why it increases overall adoption compared with both simultaneous precommitment and not offering precommitment. Altogether, across one field and two laboratory experiments including over 10,000 participants, involving diverse populations, and examining a variety of farsighted behaviors, we provide robust support for our theory about the role of inferred urgency in consumers’ responses to different types of precommitment offerings.

Practical Implications

The current research has important implications for marketers and policy makers hoping to increase the adoption of farsighted behaviors like saving. First, our research sheds light on how to design effective precommitment strategies. Prior research has primarily shown that offering people a single precommitment option (i.e., inviting people to commit now to adopt a behavior later) leads to more future-oriented choices than only inviting people to adopt the behavior now. But when applying this knowledge, marketers and policy makers may assume that they can simply add a precommitment option on top of an immediate adoption option (perhaps due to their assumption that offering more options can better cater to individuals’ heterogeneous preferences). They may particularly favor simultaneous precommitment because sequential precommitment can be costlier to implement (given that it requires repeated communication). These same considerations motivated us to test the efficacy of simultaneous precommitment in our field experiment. However, our work suggests that simultaneous precommitment is less effective than sequential precommitment and sometimes (as in our field experiment) less effective than only offering the option to adopt a behavior immediately. Importantly, it can lead consumers to delay action in contexts where such delays are costly (e.g., delays to enroll in savings programs can lead to less accumulated savings, delays to update software can increase likelihood of malware attacks, delays to obtain recommended health screenings can prevent early disease detection, delays to vaccinate can lead to unnecessary illness).

Furthermore, our findings highlight the value of understanding when choice sets presented by marketers and policy makers inadvertently communicate an urgent recommendation (or a lack thereof) to take action. More generally, our research suggests that seemingly innocuous aspects of the design of interventions (e.g., the simultaneous vs. sequential presentation of a precommitment option) can shape people's inferences and responses. As marketers attempt to leverage psychological principles (e.g., present bias) to motivate behavior change and adopt interventions from prior research, it is natural to modify the designs of those interventions to fit specific field settings, but these modifications can inadvertently leak information that ultimately harms the efficacy of the interventions. This points to the critical need to pilot-test interventions and probe what inferences they produce (Reiff et al. 2022). After such pilots, marketers and policy makers can then revise their designs to guard against unintentional information leakage before rolling out interventions at scale.

Limitations and Future Directions

Our research has several limitations, which also open up interesting directions for future research. First, our studies find that inferred urgency partially mediates the relationship between precommitment offers and farsighted decisions, suggesting that additional processes beyond those we theorized about in this article may influence people's responses to precommitment offers. For instance, participants may choose randomly over options in a given menu, which could have contributed to the observed effects of the precommitment offers in our hypothetical scenario in Study 2. Since simultaneous precommitment is the only offer we studied that included “now” and “later” enrollment options on the same menu, random choice could partially explain why people in the simultaneous precommitment condition were less likely to enroll immediately than those only given the option to enroll “now” in the control condition.

Another limitation of this research is that sequential precommitment is the only offer we studied that asks people twice whether they would adopt a behavior, and the mere repetition could contribute to the offer's positive effects on overall take-up of farsighted behaviors.¹¹ Future research should further explore these alternative accounts and others.

We also study farsighted behaviors involving both monetary (Studies 1 and 2) and time (Study 3) costs, and we find support for our theory about inferred urgency across both resource types. That said, the effect of simultaneous precommitment on overall adoption seemed to vary across resource types; simultaneous precommitment had a null effect in Studies 1 and 2 but a positive effect in Study 3. As we have proposed, the impact of simultaneous precommitment on overall adoption appears to depend on the relative magnitude of competing mechanisms: a lack of inferred urgency reduces adoption, while mechanisms related to present bias and resource slack increase adoption. The latter positive mechanisms may have played a stronger role in driving the impact of simultaneous precommitment when time costs (as opposed to monetary costs) were involved.¹² Future research should further investigate how the overall impact of precommitment offers varies with the resource type.

Further, all of the precommitment offers we study involve nonbinding commitments; that is, after committing to do something later (e.g., to increase their savings contribution rate in six months), people can always change their mind and nullify their decisions. An alternative design for precommitment offers could include binding commitments, which require people to stick to their initial commitment, and this design feature may be a key determinant of take-up (Karlan and Linden 2014). Future research should test whether our theory about inferred urgency also applies to binding precommitment offers.

In addition, though our research suggests that simultaneous precommitment leads consumers to infer that the adoption of a farsighted behavior is not urgently recommended, marketers and policy makers may be able to improve the effectiveness of simultaneous precommitment offers by changing the framing of options. For instance, when Google prompts consumers to update their notification settings, it offers simultaneous precommitment, presenting the options “continue” or “ask me later.” However, Google prints “continue” in bright blue letters, while “ask me later” appears in a light gray font, which may signal that the company recommends completing the update sooner rather than later. Future research testing different strategies for changing the information leaked by precommitment offers would be valuable. Alternatively, Beshears et al. (2021) suggest leveraging a conceptually distinct psychological process to improve simultaneous precommitment. They show that offering precommitment with the delayed behavior starting shortly after a moment that feels like a fresh start (e.g., after a birthday, the first day of spring) increases retirement savings (relative to offering precommitment at an equivalent, unlabeled time delay; e.g., in two months). Future work can more broadly examine how to frame precommitment options in ways that enhance overall adoption of farsighted behaviors.

Finally, it would be valuable to study moderators of the effects documented in our research. For instance, the extent to which people are influenced by the urgency of a marketer's recommendation may depend on consumers’ trust in that marketer. Future research is needed to understand how underlying attitudes toward whoever presents choices may moderate responses to implicit recommendations and thus influence the effects of different types of precommitment offerings.

Supplemental Material

sj-pdf-1-mrj-10.1177_00222437231153396 - Supplemental material for Save More Today or Tomorrow: The Role of Urgency in Precommitment Design

Supplemental material, sj-pdf-1-mrj-10.1177_00222437231153396 for Save More Today or Tomorrow: The Role of Urgency in Precommitment Design by Joseph Reiff, Hengchen Dai, John Beshears, Katherine L. Milkman and Shlomo Benartzi in Journal of Marketing Research

Footnotes

Acknowledgment

The authors thank their partners at the universities and at the retirement savings plan administrator for implementing the field experiment. They also thank Alessandro Previtero for his helpful comments and Hae Nim Lee, Predrag Pandiloski, and Christine Nguyen for excellent research assistance.

Associate Editor

Vikas Mittal

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This material is based on work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-1650604. The authors acknowledge financial support from the National Institutes of Health (grants P01AG005842 and P30AG034532), the Social Security Administration (grant RRC08098400-06-00 to the National Bureau of Economic Research as part of the SSA Retirement Research Consortium), the Pension Research Council/Boettner Center, the TIAA Institute, the Wharton School, the Wharton Behavioral Lab, the UCLA Anderson School of Management, and the UCLA Behavioral Lab. The opinions and conclusions expressed are solely those of the authors and do not represent the opinions or policy of NIH, SSA, any agency of the Federal Government, the Pension Research Council/Boettner Center, the TIAA Institute, or the NBER. See the authors’ websites for lists of outside activities.

ORCID iDs

Joseph Reiff

Hengchen Dai

Katherine Milkman

Notes

References

Ashraf

Nava

Karlan

Dean

Yin

Wesley

(2006), “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines,” Quarterly Journal of Economics, 121 (2), 635–72.

Benartzi

Shlomo

(2001), “Excessive Extrapolation and the Allocation of 401 (k) Accounts to Company Stock,” Journal of Finance, 56 (5), 1747–64.

Benartzi

Shlomo

Thaler

Richard H.

(2013), “Behavioral Economics and the Retirement Savings Crisis,” Science, 339 (6124), 1152–53.

Beshears

John

Dai

Hengchen

Milkman

Katherine L.

Benartzi

Shlomo

(2021), “Using Fresh Starts to Nudge Increased Retirement Savings,” Organizational Behavior and Human Decision Processes, 167, 72–87.

Brown

Christina L.

Krishna

Aradhna

(2004), “The Skeptical Shopper: A Metacognitive Account for the Effects of Default Options on Choice,” Journal of Consumer Research, 31 (3), 529–39.

Choi

James J.

(2015), “Contributions to Defined Contribution Pension Plans,” Annual Review of Financial Economics, 7 (1), 161–78.

D’Adda

Giovanna

Galliera

Arianna

Tavoni

Massimo

(2020), “Urgency and Engagement: Empirical Evidence from a Large-Scale Intervention on Energy Use Awareness,” Journal of Economic Psychology, 102275.

Dai

Hengchen

Milkman

Katherine L.

Riis

Jason

(2014), “The Fresh Start Effect: Temporal Landmarks Motivate Aspirational Behavior,” Management Science, 60 (10), 2563–82.

Fox

Craig R.

Ratner

Rebecca K.

Lieb

Daniel S.

(2005), “How Subjective Grouping of Options Influences Choice and Allocation: Diversification Bias and the Phenomenon of Partition Dependence,” Journal of Experimental Psychology: General, 134 (4), 538–51.

10.

Frederick

Shane

Loewenstein

George

O'Donoghue

Ted

(2002), “Time Discounting and Time Preference: A Critical Review,” Journal of Economic Literature, 40 (2), 351–401.

11.

Goodman

Joseph K.

Broniarczyk

Susan M.

Griffin

Jill G.

McAlister

Leigh

(2013), “Help or Hinder? When Recommendation Signage Expands Consideration Sets and Heightens Decision Difficulty,” Journal of Consumer Psychology, 23 (2), 165–74.

12.

Kardes

Frank R.

Posavac

Steven S.

Cronley

Maria L.

(2004), “Consumer Inference: A Review of Processes, Bases, and Judgment Contexts,” Journal of Consumer Psychology, 14 (3), 230–56.

13.

Karlan

Dean

Linden

Leigh L.

(2014), “Loose Knots: Strong Versus Weak Commitments to Save for Education in Uganda,” Working Paper 19863, National Bureau of Economic Research, https://www.nber.org/papers/w19863.

14.

Knoll

Melissa A.Z.

Houts

Carrie R.

(2012), “The Financial Knowledge Scale: An Application of Item Response Theory to the Assessment of Financial Literacy,” Journal of Consumer Affairs, 46 (3), 381–410.

15.

Krijnen

Job M.T.

Tannenbaum

David

Fox

Craig R.

(2017), “Choice Architecture 2.0: Behavioral Policy as an Implicit Social Interaction,” Behavioral Science & Policy, 3 (2), 1–18.

16.

Lieberman

Alicea

Duke

Kristen E.

Amir

(2019), “How Incentive Framing Can Harness the Power of Social Norms,” Organizational Behavior and Human Decision Processes, 151, 118–31.

17.

McKenzie

Craig R.M.

Liersch

Michael J.

Finkelstein

Stacey R.

(2006), “Recommendations Implicit in Policy Defaults,” Psychological Science, 17 (5), 414–20.

18.

Milkman

Katherine L.

Rogers

Todd

Bazerman

Max H.

(2008), “Harnessing Our Inner Angels and Demons: What We Have Learned About Want/Should Conflicts and How That Knowledge Can Help Us Reduce Short-Sighted Decision Making,” Perspectives on Psychological Science, 3 (4), 324–38.

19.

Milkman

Katherine L.

Rogers

Todd

Bazerman

Max H.

(2009), “Highbrow Films Gather Dust: Time-Inconsistent Preferences and Online DVD Rentals,” Management Science, 55 (6), 1047–59.

20.

Milkman

Katherine L.

Rogers

Todd

Bazerman

Max H.

(2010), “I’ll Have the Ice Cream Soon and the Vegetables Later: A Study of Online Grocery Purchases and Order Lead Time,” Marketing Letters, 21 (1), 17–35.

21.

Morewedge

Carey K.

Holtzman

Leif

Epley

Nicholas

(2007), “Unfixed Resources: Perceived Costs, Consumption, and the Accessible Account Effect,” Journal of Consumer Research, 34 (4), 459–67.

22.

O’Donoghue

Ted

Rabin

Matthew

(1999), “Doing It Now or Later,” American Economic Review, 89 (1), 103–24.

23.

Read

Daniel

Loewenstein

George

Kalyanaraman

Shobana

(1999), “Mixing Virtue and Vice: Combining the Immediacy Effect and the Diversification Heuristic,” Journal of Behavioral Decision Making, 12 (4), 257–73.

24.

Read

Daniel

Van Leeuwen

Barbara

(1998), “Predicting Hunger: The Effects of Appetite and Delay on Choice,” Organizational Behavior and Human Decision Processes, 76 (2), 189–205.

25.

Reiff

Joseph S.

Zhang

Justin

Gallus

Jana

Dai

Hengchen

Pedley

Nathaniel

Vangala

Sitaram

Leuchter

Richard K.

Goshgarian

Gregory

Fox

Craig R.

Han

Maria

Croymans

Daniel M.

(2022), “When Peer Comparison Information harms Physician Well-Being,” Proceedings of the National Academy of Sciences of the United States of America, 119 (29), e2121730119.

26.

Rogers

Todd

Bazerman

Max H.

(2008), “Future Lock-In: Future Implementation Increases Selection of ‘Should’ Choices,” Organizational Behavior and Human Decision Processes, 106 (1), 1–20.

27.

Rogers

Todd

Milkman

Katherine

Volpp

Kevin

(2014), “Commitment Devices: Using Initiatives to Change Behavior,” Journal of the American Medical Association, 311 (20), 2065–66.

28.

Schwartz

Janet

Mochon

Daniel

Wyper

Lauren

Maroba

Josiase

Patel

Deepak

Ariely

Dan

(2014), “Healthier by Precommitment,” Psychological Science, 25 (2), 538–46.

29.

Smith

N. Craig

Goldstein

Daniel G.

Johnson

Eric J.

(2013), “Choice Without Awareness: Ethical and Policy Implications of Defaults,” Journal of Public Policy & Marketing, 32 (2), 159–72.

30.

Tannenbaum

David

Fox

Craig R.

Goldstein

Noah J.

(2013), “Partitioning Menu Items to Nudge Single-Item Choice,” working paper.

31.

Thaler

Richard H.

Benartzi

Shlomo

(2004), “Save More Tomorrow™: Using Behavioral Economics to Increase Employee Saving,” Journal of Political Economy, 112 (S1), S164–87.

32.

Ubel

Peter A.

Comerford

David A.

Johnson

Eric

(2015), “Healthcare.gov 3.0—Behavioral Economics and Insurance Exchanges,” New England Journal of Medicine, 372 (8), 695–98.

33.

Valenzuela

Ana

Raghubir

Priya

(2009), “Position-Based Beliefs: The Center-Stage Effect,” Journal of Consumer Psychology, 19 (2), 185–96.

34.

VanEpps

Eric M.

Downs

Julie S.

Loewenstein

George

(2016), “Advance Ordering for Healthier Eating? Field Experiments on the Relationship Between the Meal Order–Consumption Time Delay and Meal Content,” Journal of Marketing Research, 53 (3), 369–80.

35.

Zauberman

Gal

Lynch

John G.

Jr. (2005), “Resource Slack and Propensity to Discount Delayed Investments of Time Versus Money,” Journal of Experimental Psychology: General, 134 (1), 23–37.

36.

Zhu

Meng

Yang

Hsee

Christopher K.

(2018), “The Mere Urgency Effect,” Journal of Consumer Research, 45 (3), 673–90.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.63 MB