When “Year” Feels Near: How Year Versus Length Framing Alters Time Perception and Consumer Decisions

Abstract

Time intervals can be framed either by a calendar year (e.g., “2015”) or by length (e.g., “ten years”), yet these ostensibly equivalent formats lead to systematically different judgments. Combining data from whiskey auctions with seven controlled experiments, the authors demonstrate that length framing elongates time perception compared with year framing, which they refer to as the year–length effect. As a result of changes in time perception, length framing increases the importance of time-related attributes in choice, leading to more favorable product evaluations in contexts where age enhances product value (e.g., whiskey evaluation) and to more negative evaluations in contexts where age reduces it (e.g., used goods). Process evidence implicates the logarithmic mental number line: Years with large nominal values occupy a compressed region of the line, relative to small length numerals. These findings offer practical guidance on how time framing can be used to shape time perception and customer value.

It is 2025, and Mr. Bond, a whiskey connoisseur, is searching for a bottle of Scotch. He comes across two options. The label on the first bottle states that the whiskey is 24 years old. This means that after being distilled, the whiskey aged in a cask for 24 years before being bottled. The second label indicates that the Scotch was distilled in 2000 and bottled in 2024. Objectively, both bottles contain whiskey of the same age, yet one presents length of the aging period while the other presents the aging period boundaries.

This variation in temporal framing is not unique to the whiskey category. In banking, for instance, mortgage terms can be described in terms of length (e.g., “fixed interest rate for five years”) or year boundaries (e.g., “fixed interest rate until 2030”). In secondhand goods markets, sellers may state how long ago they bought an item (e.g., “purchased two years ago”) or the year they purchased it (e.g., “purchased in 2023”). Across industries, firms may decide whether to communicate age using length framing, as do Guinness and Geico, or boundary framing, as do Jim Beam and Wells Fargo (see Table 1 and Web Appendix A for more examples).

Table 1.

Examples of Year and Length Framing in Seller Communications.

Category	Year Framing (Brand/Platform)	Length Framing (Brand/Platform)
Whiskey age	“Vintage 2000” (Berry Bros. & Rudd)	“24 Years Old” (Teeling)
Mortgage terms	“Fixed rate … until 31 May 2030” (Allied Irish Banks*)	“5 Year Fixed (Remortgage Only)” (Barclays)
Investment returns	“Since 1957 … index has delivered an average annual return of over 10%” (Investopedia)	“Average yield of 5 to 10 year: 3.12%” (Bank of Canada)
Used goods	“Barrel Sauna … bought in 2023”; “Color ink cartridge … bought new in 2023” (eBay)	“Crossbow, used for 2 years”; “Ryzen5 (Used For 2 years)” (eBay)
Brand age	“Staying true to the original since 1795” (Jim Beam)	“Over 260 years in the making” (Guinness)
Climate pledge	“Microsoft will be carbon negative by 2030” (Microsoft); “Apple commits to be carbon neutral … by 2030” (Apple)	“Carbon neutral in 10 years” (Loren Wood Builders); “Carbon Neutrality … over 10 Years” (Hitachi)
CDs/bonds	“4.56% Yield … Maturity: 12/23/2028” (JP Morgan Chase Bonda); “Market-Linked Capped Certificates of Deposit … due June 2, 2028” (Morgan Stanley CDa)	“Canada 5 Year Bond CA5Y-CA:Canada” (CNBC); “3.00% on a 5 year term Non-Redeemable GIC” (Royal Bank of Canada)
Property age	“Built in 2011” (Zillow); “2006 Year Built” (Redfin)	“Age of building 6 to 15 years” (Realtor.ca), “Property Age 1 to 5 Year old” (99acres.com)
Aged products	“Aceto Balsamico Tradizionale Di Modena DOP 1959” (La Secchia Balsamic Vinegar)	“Giusti 50 Reserve” (Giusti Balsamic Vinegar)
Recipe heritage	“Based on Mangieri Bros. family recipe since 1924” (Una Pizza Napoletana)	“Based on a 100-year-old Mangieri family recipe” (Una Pizza Napoletana)
Financing promo	“No Interest Until 2026” (Graves Bros.); “Pay 0% Intro APR Until Nearly 2026” (CompareCards)	“7 Years Interest Free” (Wren Kitchens); “Financing Offer: 0% APR for 3 years” (Mattress & Co)
Warranty period	“Sea-Dweller 16600 – with Rolex Service Card (Warranty Until 2027)” (Rolex); “Great watch … Warranty is covered till 2029” (Chrono24)	“5 Years of Protection” (Costco); “2 Years vs. 3 Years Plan” (Best Buy); “Limited Warranty 4 Years” (Tesla)

^a

In these examples, the time cue appears as a date rather than as a stand-alone year boundary. Notes: See Web Appendix A for actual images of the examples included in Table 1.

In this research, we examine how time presentation—in terms of length versus year boundaries—affects consumers’ time perception, product valuation, and choice. Using data from whiskey auctions and seven controlled experiments, we show that length framing elongates perceived time relative to year framing, a phenomenon we refer to as the year–length effect. As a result, length framing increases the importance of time-related attributes in choice and leads to more favorable product evaluations in contexts where age enhances product value (e.g., whiskey evaluation), but more negative evaluations in contexts where age reduces it (e.g., used goods).

This work has both practical and theoretical implications. First, in product categories where age is positively tied to value—such as whiskey or fine art—our findings suggest that managers may benefit from framing age in terms of length (e.g., “12 years old”). For instance, our whiskey auction data indicate that, holding all else constant, bottles with age framed in terms of length command roughly 9% higher prices than year-framed bottles. Yet in this same dataset, which spans 1,385 unique whiskey brands, over 30% of brands employed year framing for at least one product—potentially lowering consumer willingness to pay (WTP).

Conversely, in contexts where longer perceived duration may undermine consumer evaluations, such as financial investments and secondhand goods, managers and vendors may benefit from adopting year framing. For example, when describing investment returns, managers are more likely to receive favorable evaluations when using year framing (e.g., “investment value tripled since 2022”) than when using length framing (e.g., “value tripled in three years”).

Second, our findings carry implications for policymakers. Our investment evaluation and mortgage preference studies show that time presentation influences consumer judgment and choice. For instance, our choice-based conjoint experiment suggests that banks such as Barclays, which use length framing in mortgage descriptions (e.g., “interest rate fixed for five years”; Table 1), may lead consumers to place greater weight on fixed-rate periods and less weight on interest rates compared with lenders that adopt year framing (e.g., Allied Irish Banks). More broadly, given the malleability of consumers’ perceptions of time, policymakers may need to consider whether descriptions of financial products with significant implications for consumer welfare should be left to the discretion of firms or standardized by agencies such as the Consumer Financial Protection Bureau.

Our work also offers at least two theoretical contributions to consumer psychology and time perception literature. First, we extend research on date–delay effects (LeBoeuf 2006; Malkoc, Zauberman, and Bettman 2010; Munichor and LeBoeuf 2018; Read et al. 2005), which examines intertemporal preferences across length and boundary frames. Most commonly, this work contrasts durations expressed in length (e.g., “6 months,” “12 weeks”) with those expressed by end dates (e.g., “on December 12, 2003”), consistently showing that date-boundary framing makes people more willing to wait for future rewards. However, the mechanisms behind this effect remain debated, with precision (Read et al. 2005), attention (LeBoeuf 2006), and concreteness (Malkoc, Zauberman, and Bettman 2010) proposed as possible drivers (see Web Appendix B for details). Our findings suggest that boundary–length effects are at least partly driven by logarithmic compression of the mental number line, with length-framed durations (e.g., “25 years ago”) occupying a more expansive region than year-framed durations (e.g., “in 2000”).

Relatedly, we contribute to the literature on unit effects in consumer judgments. Time, like other continuous attributes, can be expressed with larger numbers and smaller units (e.g., 72 to 84 months) or with smaller numbers and larger units (e.g., 6 to 7 years). A large body of work suggests that the “large number–small unit” framing increases magnitude perceptions relative to the “small number–large unit” framing, because people focus on the face value of numbers without sufficiently adjusting for unit size (Aribarg, Burson, and Larrick 2017; Burson, Larrick, and Lynch 2009; Gourville 1998; Pandelaere, Briers, and Lembregts 2011; Raghubir and Srivastava 2002), though this pattern is not universal (Aribarg, Burson, and Larrick 2017; Monga and Bagchi 2012; Ülkümen and Thomas 2013). For instance, when people rely more on units rather than numbers, “large number–small unit” framing can reduce perceived duration (Monga and Bagchi 2012). Building on these findings, one might expect year framing (e.g., distilled in 2000) to produce larger magnitude perceptions than length framing (e.g., distilled 25 years ago), since both frames use the same unit (years) but year framing features larger numbers. Yet, we consistently find that perceived duration is longer under the small-number length framing than under the large-number year framing—a pattern we attribute to greater logarithmic compression of time under year framing. Our work therefore nuances prior findings on unit and numerosity effects by showing that, even when the unit of time is held constant, larger numbers can sometimes reduce rather than increase perceived duration.

In the following sections, we develop our predictions and present an archival study of whiskey auction prices, followed by seven preregistered experiments, to test our theorizing. We conclude with a discussion of theoretical and practical implications of this research.

Theoretical Background and Hypotheses

Although our work contrasts length- and boundary-framed intervals spanning multiple years (e.g., “24 years ago” vs. “since 2001”), prior research has already documented differences in responses to these frames for shorter periods (e.g., “in 8 weeks” vs. “on July 28, 2025”). First, length framing can reduce objective duration estimates for future tasks (LeBoeuf and Shafir 2009). For instance, when people estimate how long it would take them to finish reading a book, they anchor more on the present and produce shorter time estimates when time is framed as a duration (e.g., “How many days until …?”) rather than a date (e.g., “On what date …?”).

Second, length framing increases intertemporal discounting: It makes people less likely to choose larger, later rewards and leads them to demand larger premiums for waiting (LeBoeuf 2006; Malkoc, Zauberman, and Bettman 2010; Read et al. 2005). For example, when waiting times are presented in months (e.g., four months vs. six months) rather than dates (e.g., “on August 12, 2003” vs. “on October 12, 2003”), people become less willing to wait the additional two months for a larger return.

Third, length versus date-boundary framing can shape goal pursuit by altering the salience of competing obligations within a given time frame (Munichor and LeBoeuf 2018). Because consumers consider competing obligations less under length framing, they are more likely to adopt goals framed in terms of length (e.g., “losing two pounds within two weeks”) than goals framed by a specific date (e.g., “losing two pounds by July 3”).

Although most prior studies either do not directly measure perceived duration (e.g., Malkoc, Zauberman, and Bettman 2010; Read et al. 2005) or do not find that time framing affects time perception (e.g., Munichor and LeBoeuf 2018), their results are broadly consistent with the idea that length framing increases perceived duration compared with boundary framing. Longer perceived waiting times could explain why people are less likely to choose future rewards under length framing, while longer perceived duration of goal pursuit could make goals feel more feasible, increasing goal adoption.

But why might time feel longer under length framing? One possibility is that this framing makes thoughts about the passage of time more vivid (LeBoeuf 2006). This account, albeit untested, holds strong face validity: Length descriptions (e.g., “in six months”) state the duration explicitly, whereas date-boundary descriptions (e.g., “by December 12, 2025”) require consumers to compute it. As a result, thoughts about how much time will pass may become more accessible and vivid under length framing, increasing perceived duration.

Another possibility is that differences in perceived duration arise from the units of time used across frames. Prior work often confounds time framing with unit selection: Length framing typically presents time in weeks or months (e.g., “in eight months,” “within two weeks”), whereas boundary framing presents it in terms of days (e.g., “on December 12, 2003”). Because people often rely on the prototypical meaning of units, smaller units can imply shorter intervals (Monga and Bagchi 2012; Ülkümen and Thomas 2013). For example, a delivery window described as “1–3 weeks” rather than “7–21 days” may feel longer, since weeks are conventionally associated with more substantial spans of time (Monga and Bagchi 2012). Thus, length-framed intervals may be judged as longer both because they make the passage of time more vivid and because they are often expressed in larger, more substantial units.

In this research, we focus on length and year-boundary frames that hold the unit of time constant. This comparison allows us to isolate the effects of length versus boundary framing from any influence of unit changes, while also reflecting many marketplace contexts where product or company age is described either in years or in year boundaries. We propose that length framing elongates perceived duration relative to year framing because of logarithmic compression in number perception. Consequently, length framing should improve evaluations in contexts where age enhances value and lower evaluations in contexts where age reduces value. We elaborate on these predictions next and revisit alternative explanations of the year–length effect in the empirical section.

Logarithmic Compression in Year-Length Framing

Research in numerical cognition and consumer psychology suggests that humans represent numbers logarithmically rather than linearly (Coulter and Norberg 2009; Dehaene 2001; Dehaene et al. 2008; Grewal and Marmorstein 1994; Lenkovskaya and Sweldens 2025; Monroe 1971). Under logarithmic compression, differences between smaller numbers (e.g., 1 and 5) are perceived as larger than equivalent differences between larger numbers (e.g., 101 and 105). This property underlies many aspects of economic behavior, including diminishing marginal utility (Schley and Peters 2014), relative thinking (Saini and Thota 2010), and ambiguity aversion (Lenkovskaya and Sweldens 2025).

In marketplace contexts, logarithmic compression helps explain a wide range of consumer behaviors. For example, Grewal and Marmorstein (1994) show that consumers are less likely to search or compare prices when purchasing high-priced products, because a given dollar difference feels smaller relative to the high base price. Similarly, Saini and Thota (2010) find that consumers are less likely to travel to another store to save $5 on a $100 item than on a $20 item. Extending this logic beyond prices, Lenkovskaya and Sweldens (2025) demonstrate that logarithmic compression underlies evaluations of numeric ranges (e.g., vaccine effectiveness of 90%–98%, discount of 50%–70%). It leads people to view the range midpoint (e.g., 94% effectiveness in a 90%–98% range; 60% discount in a 50%–70% range) as closer to the upper than the lower bound, which in turn shifts preferences between products described by a range (e.g., “90%–98% effective”) and those described by a precise midpoint (e.g., “94% effective”).

In most cases, length framing expresses time with smaller numerals that occupy a more expansive region of the mental number line. As a result, time intervals should appear longer under length framing (e.g., two vs. five years) than under year framing, which typically involves larger numerals (e.g., 2020 vs. 2023). Importantly, because consumers tend to anchor their temporal evaluations on the present (Caruso et al. 2013; LeBoeuf and Shafir 2009), logarithmic compression should occur even when a single duration or year boundary is presented in isolation. For example, when a product is described as “purchased two years ago,” consumers anchor on “now” and mentally represent the interval between 0 and 2. By contrast, when the product is described as “purchased in 2024,” consumers anchor on the present year (2026) and represent the interval between 2024 and 2026, which would be more compressed. Thus, we hypothesize that:

• H₁: Time intervals described using the length framing are perceived as longer than equivalent time intervals described using the year framing.

Implications of Time Framing for Consumer Evaluations

So far, we have argued that length framing increases perceived duration because smaller numerals are less compressed on the mental number line. We further propose that this increase in perceived duration shapes product evaluations, WTP, and choice. In categories such as whiskey and fine art, where age is positively linked to value, length framing should make products feel older and therefore more valuable. Accordingly, when age is expressed as a length rather than a year boundary, consumers should evaluate the product more favorably, exhibit greater WTP, and be more inclined to buy the older products.

By contrast, in categories such as used goods or financial investments, where age diminishes value, length framing should have the opposite effect. Because length framing stretches perceived duration, it should reduce evaluations, lower WTP, and shift choice toward options associated with shorter durations. Formally:

• H_2a: In contexts where age enhances value, length framing (e.g., “15-year-old whiskey”) leads to higher evaluations compared with year framing (e.g., “2010 whiskey”).

• H_2b: In contexts where age diminishes value, length framing (e.g., “sofa bought two years ago”) leads to lower evaluations compared with year framing (e.g., “sofa bought in 2023”).

• H₃: The effect of year–length framing on consumer evaluations is mediated by perceived duration.

Overview of Studies

We report eight studies that test our theorizing (see Table 2 for a summary). Study 1 analyzes a large archival whiskey auction dataset (50,916 auctions) to examine how whiskey age framing (length vs. year) affects selling prices. Study 2 demonstrates the impact of time framing on perceived duration for both future- and past-oriented judgments. Studies 3–6 explore downstream consequences of the year–length effect for investment evaluations (Study 3), whiskey choice (Study 4), WTP for used goods (Study 5), and mortgage preferences (Study 6). Finally, Studies 7a and 7b provide evidence for the logarithmic compression account by showing that the year–length effect is attenuated when participants are prompted to represent time linearly (Study 7a) and when they evaluate temporally distant events (Study 7b). All experiments were preregistered. Preregistration links are provided in the study descriptions, and all data, stimuli, and materials are available on ResearchBox (https://researchbox.org/1319).

Table 2.

Overview of Studies.

Study: Context/Task	Goal	Design (Factors)	Outcome Variable	Year–Length Effect
1: Whiskey auction (field)	Archival test of framing on auction prices	Correlational	Sell price	Δ 8.9%, length > year
2: Time perception task (past and future)	Framing effect on time perception; rule out polarity–congruence account	2 (framing) × 2 (direction) × 3 (replicate)	Subjective time perception	${\eta }_{\mathrm{p}}^2$  = .06, length > year
3: Financial return evaluation	Does year framing raise investment return evaluations?	2 (framing) × 4 (replicate)	Return evaluation	${\eta }_{\mathrm{p}}^2$  = .02, year > length
4: Incentive-compatible whiskey choice	Does length framing increase preference for older whiskeys? Test mediation	2 (framing) × 3 (replicate)	Choice of older (expensive) bottle	Odds ratio = 3.74, length > year
5: Used goods evaluation	Framing effect on WTP for used goods; compare rival accounts	2 (framing) × 3 (replicate)	WTP	Δ 17%, year > length
6: Mortgage conjoint	Does length framing raise importance of time attribute? Rule out computational ease	Two-cell (framing)	Attribute importance (%)	Δ 6% points, length > year
7a: Time perception task	Does a linear scale moderate the framing effect on time perception?	2 (framing) × 2 (scale) × 2 (replicate)	Subjective time perception	${\eta }_{\mathrm{p}}^2$  = .03 (control scale), length > year
7b: Brand-age valuation (insurance)	Does length framing raise WTP for older brands? Brand-age moderation and rule out computational ease	2 (framing) × 2 (brand age)	WTP	Δ 44%, length > year (small ages)

Study 1: Analysis of Whiskey Auction Data––Evidence from the Field

Study 1 examines the economic impact of whiskey age framing—as a year boundary or as a length—on consumer valuations. Our theory predicts that whiskeys described with length framing will be perceived as older and more valuable, leading to higher selling prices. Study 1 tests this prediction using a large dataset of whiskey auction transactions.

We focus on whiskey because, unlike wine, it does not improve with age once bottled (Teeter 2020). When stored properly, whiskey quality remains stable, meaning that a bottle distilled in 2015 and bottled in 2025 is consistently ten years old. Moreover, while bottles often display distillation year, bottling year, and age, distillers typically highlight either the year or the length in bottle descriptions, making this context well suited for testing our hypothesis.

Data

Secondary data

We obtained our dataset from publicly available online whiskey auction listings hosted by a U.K.-based auction platform. This dataset encompasses extensive details on whiskey attributes such as age, brand, and final sold price. For the purposes of our study, we use the term “whiskey” to cover both the Irish/American spelling “whiskey” and the Scottish/Canadian/Japanese “whisky.”

This auction platform hosts monthly online auctions, attracting a global audience of buyers and sellers. Sellers must register and send their whiskey bottles to the company for authentication prior to listing. Each auction runs for a month and features hundreds of whiskeys available for bidding. Buyers worldwide can participate and place bids on their chosen bottles, facilitated by the auction company's international shipping services. At the conclusion of each auction, the highest bidder is awarded the bottle. While our dataset provides extensive details on whiskey characteristics and auction outcomes, it lacks specific demographic information about the buyers and sellers. Nevertheless, the depth of the dataset in terms of whiskey details and transactional data offers a valuable opportunity to test our hypothesis in a real-world context.

Data exclusions

We identified 21,995 whiskey listings whose text titles included both a year boundary (e.g., 1991) and a length cue (e.g., 12 years old). However, follow-up inspection revealed that bottle label images typically emphasized only one of these time cues, making it difficult to determine whether the listing primarily framed whiskey age by year or length. To avoid misclassification, we excluded these listings from the analysis. ¹ Additionally, to guard against undue influence of extreme values, we winsorized data using the interquartile range (IQR) rule for both logged price and whiskey age: observations more than 1.5 IQRs below the first quartile or above the third quartile were removed. Figure 1 presents the resulting age distribution.

Figure 1.

Whiskey Age Distribution by Framing.

Our main analysis was conducted on the final dataset of 50,916 whiskey auctions run from March 2016 to January 2024. Table 3 presents the key descriptive statistics for the whiskey bottles sold under the year (vs. length) framing. Web Appendix C presents bottle image examples and additional descriptive statistics for the auctioned bottles in the sample.

Table 3.

Descriptive Statistics: Whiskey Auction Data.

Framing	Price	Age	Size (in cL)	Strength (% ABV)	Mean Sold Time	Unique Brands	N
Length	£182	14.8	70	46%	Mid-2020	1,123	42,992
Year	£265	15.7	70	50%	Mid-2020	466	7,924
Total	£195	14.9	70	47%	Mid-2020	1,385	50,916

Dependent variable

Our main dependent variable was the log-transformed final sold price, representing the final auction price awarded to the highest bidder. It is important to note that while our dataset did not include details on competing bids, the iterative nature of the bidding process (allowing bidders to increase their offers) and the month-long auction duration suggest that the final sold price was shaped by multiple bids, reflecting the collective valuation of whiskey enthusiasts.

Independent variable

Our focal predictor was the time frame used in the listing title to convey a whiskey's age. Titles such as “1989 The Macallan Fine & Rare Vintage Single Malt Scotch Whisky” use year framing, whereas “Macallan Highland Single Malt Scotch Whisky 30 Years Old” use length framing. To confirm that the title text matched what bidders saw, we manually inspected a subset of listings and found that the cue emphasized on the bottle label image (year vs. length) aligned with the title text.

Control variables and fixed effects

In addition to sold price and time framing, our dataset includes a comprehensive set of whiskey characteristics used in the analysis. First, we included continuous control variables to account for whiskey age (i.e., time matured in casks prior to bottling), bottle size (in centiliters), delivery weight (in kilograms), alcohol strength (% alcohol by volume [ABV]), and bottle serial number. We also controlled for the number of awards mentioned (quantified by mentions of award-related terms on the label) and bottle information content, captured by the number of characters in the whiskey text description.

Next, we included (1) a dummy variable for the presence of a value-added tax (VAT) applicability badge (N = 4,607) on the bottle (1 = present, 0 = absent), which indicates that the bottle's final bidding price would be subject to standard VAT at checkout, ² and (2) two separate dummies capturing the presence of rarity descriptors in the whiskey description: one for “rare” (1 = rare, 0 = not mentioned) and another for “very rare” (1 = very rare, 0 = not mentioned).

Finally, we added fixed effects for (1) warning labels (e.g., notes about packaging flaws; warning text length split into tertiles), (2) brand (1,385 brands), (3) whiskey type (19 categories; e.g., single malt, bourbon), (4) region (eight geographical origins), (5) distillery status at listing (operational, closed, or demolished), (6) auction lot number (auction ID, pooled into 100-lot bands), (7) sale year, and (8) year of manufacture.

Results

To examine the impact of time framing (year vs. length) on whiskey prices, we employed a linear regression model. The dependent variable was the logged final sold price, and the key independent variable was the dummy-coded time framing indicator (length = 1, year = 0). Table 4 presents the results of the analysis across different model specifications. Model 1 includes the effect of the time framing variable and a core set of predictors (whiskey age, bottle size, delivery weight, alcohol strength, VAT applicability, brand name, whiskey type, region, and year of sale). Model 2 includes additional predictors, such as the year of manufacture. ³

Table 4.

OLS Regression Models for Whiskey Auctions.

Predictor Variables	Dependent Variable: ln (Price)
	Model 1	Model 2
Framinga: Length	.09*** (.07, .10)	.08*** (.07, .10)
Age	.53*** (.53, .54)	.37*** (.18, .55)
Size	.16*** (.15, .16)	.14*** (.14, .15)
Delivery weight	.16*** (.15, .16)	.15*** (.15, .16)
Strength	.39*** (.38, .39)	.35*** (.34, .36)
Bottle serial number		−.02 (−.04, .00)
Award count		.00 (−.00, .01)
Description length		.19*** (.18, .21)
VAT applicablea: Yes	−.09*** (−.10, −.07)	−.05*** (−.08, −.03)
Rarity descriptor: Very Rarea		.61*** (.55, .67)
Rarity descriptor: Rarea		.31*** (.29, .34)
Constant	2.04*** (.86, 3.22)	1.01 (−.64, 2.65)
Branda	Yes	Yes
Year of salea	Yes	Yes
Type of whiskeya	Yes	Yes
Regiona	Yes	Yes
Distillery statusa	No	Yes
Warning labelsa	No	Yes
Lot number bina	No	Yes
Year of manufacturea	No	Yes
Observations, adjusted R2	50,916, .70	50,916, .74
F-statistic	82.56***	46.56***
D.F.	1,421; 49,494	3,131; 47,784

***p < .001.

^a

Indicates categorical variables.

Notes: All continuous independent variables are standardized. 95% CIs are reported in parentheses. Caution is advised when interpreting the whiskey age coefficient in Model 2 due to multicollinearity with the year of manufacture.

Focusing first on Model 1, we find that whiskeys described using length framing, on average, commanded 8.9% higher prices than those described using year framing (B = .086, SE = .01, t = 9.40, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .002). For context, VAT applicability—an automatic tax on the final price—affected sold prices by a similar magnitude. That is, for bottles whose final sold prices were equal to the winning bid amount plus a VAT, buyers bid on average 8.1% less (B = −.085, SE = .01, t = −8.96, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .002), compared with bottles whose final sold prices were set at the highest bid amount only. Thus, the effect of time framing on sold prices is on par with the economic impact of VAT, underscoring its practical significance.

Beyond time framing, several other predictors show expected and significant effects. Age, which reflects the time whiskey spent maturing in casks before bottling, consistently emerges as a strong positive predictor. Older whiskeys fetch higher prices, aligning with industry norms that associate longer maturation with greater flavor complexity, rarity, and prestige. As a benchmark, the model suggests that the time framing effect corresponds to a perceived age difference of about .91 years. Bottle size also positively influences price, with larger bottles commanding higher prices. Strength (% ABV) exerts a similarly positive effect, as consumers often perceive higher-strength whiskeys as more premium or concentrated in flavor. Delivery weight has a positive effect, likely reflecting the premium appearance or packaging associated with heavier bottles.

Model 2 confirms that the time framing effect remains robust—with an average 8.4% price difference between length- versus year-framed bottles—after we include additional control variables and fixed effects. The signs and significance of the control variables are also generally consistent across models. The fully controlled model explains 74% of the variation in whiskey auction prices, which is an impressive fit given the absence of buyer-level data.

Robustness checks

To address possible endogeneity concerns, we ran several robustness checks (Web Appendix C). One particular concern pertains to the role of industry norms: Because more than 80% of listings in our dataset use length framing, the minority of year-framed bottles might be devalued for violating the category norm. To test this, we isolated brands that showed no clear framing convention—for which roughly half the listings used year framing and half used length framing. This restriction yielded 11,029 auctions across 370 brands. Within this subset the year–length effect on price remained significant (robustness check 3, Web Appendix C), indicating that the effect is likely not a mere artifact of brand-level framing norms.

Discussion

Our findings offer valuable guidance for whiskey brand managers deciding between year and length framing. In particular, the results caution managers against using the year framing in bottle labeling, which may alter how consumers perceive the whiskey's age and value.

While Study 1 provides initial support for H_2a using real-world auction data, its correlational nature presents some limitations. As noted previously, length framing may resonate more with category norms, leading consumers to prefer it. Moreover, bottling year information may be harder for consumers to locate or interpret, making year-framed bottles harder to evaluate. We revisit the ease account in Studies 5, 6, and 7b by measuring ease of computation and ease of evaluation, and in Study 4 we examine the year–length effect on whiskey choice in a controlled setting where the bottling year is provided explicitly. Furthermore, time framing could be confounded with unobserved factors such as product quality or marketing strategy. For example, length framing allows for consistent labeling across batches, whereas year framing may be used more strategically to highlight batch-specific terroir. These confounds raise the possibility that the lower valuations observed for year-framed whiskeys are not driven by year–length framing and the corresponding changes in age perception. Thus, in the remaining studies, we experimentally manipulated time framing to test the causal effect of time framing on time perception and consumer valuations.

Study 2: Past and Future Time Perception

Study 2 tests H₁ by examining whether time points framed in terms of length (e.g., “three years ago”) are perceived as more distant than those framed in terms of year boundaries (e.g., “the year 2022”). Beyond this main prediction, the study evaluates the numeric polarity congruence account of the year–length effect. For past judgments in the length frame, smaller numbers (e.g., “three years ago”) denote shorter temporal distances, whereas larger numbers (e.g., “five years ago”) denote longer ones. Thus, numeric polarity (small vs. large) aligns with temporal distance (close vs. far). In the year frame, this mapping reverses: smaller numbers (e.g., “2020”) indicate larger temporal distances, whereas larger numbers (e.g., “2022”) indicate shorter ones. Thus, for past events, numeric polarity and temporal distance are congruent in the length frame but incongruent in the year frame. Such mismatches can compress perceived magnitude because implicit memory follows a “larger number = larger distance” heuristic (Kyung, Thomas, and Krishna 2017).

Yet, this incongruence should arise only for past judgments. For future events, numeric polarity is aligned with temporal distance (smaller numbers = shorter distance, larger numbers = longer distance) in both length and year frames. If the year–length effect observed in Study 1 is driven by polarity congruence, it should emerge for past judgments but not for future ones. Study 2 tests this assertion. The preregistration is available at aspredicted.org/gbxj-mj8w.pdf.

Method

Participants

In May 2025, we recruited 846 U.S. adults from Prolific. Nine respondents failed our preregistered exclusion criteria, leaving a final sample of 837 participants (53.9% female, 45.0% male, 1.1% nonbinary; M_age = 40.3 years, SD = 12.8). Full exclusion details for all studies are reported in Web Appendix C.

Design and procedure

The study adopted a 2 (time framing: year vs. length, between) × 2 (time direction: past vs. future, between) × 3 (objective time distance replicates: 3, 5, and 20 years; within) mixed-factorial design.

Participants saw three time points, one at a time, in random order. In the year condition, time was expressed in calendar years (e.g., “the year 2022,” “the year 2028”). In the length condition, it was expressed as a duration (e.g., “3 years ago,” “3 years later”). Half of the participants evaluated past events (e.g., “the year 2022,” “3 years ago”), while the other half evaluated future events (e.g., “the year 2028,” “3 years later”).

For each time point, participants rated how far away it felt using a slider anchored at “very close” (left) and “very far” (right). The response scale ranged from 0 to 180, but the numeric values were not displayed, and no default cursor position was shown. The handle appeared only once a participant clicked on the scale. After completing the three judgments, participants reported their age and gender.

Results

Subjective time perception

Time perception judgments were analyzed with a 2 (time framing: year vs. length, between) × 2 (time direction: past vs. future, between) × 3 (objective time distance: 3, 5, 20 years, within) mixed-factorial ANOVA. As predicted, there was a significant main effect of time framing, such that length framing increased perceived duration (M_length = 102.08, SD = 56.46) compared with the year framing (M_year = 81.06, SD = 58.56; F(1, 833) = 55.98, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .06). There was also a main effect of objective time distance (F(2, 1,666) = 1,613.66, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .66), suggesting that participants were sensitive to changes in objective duration. The main effect of time direction was not significant (F(1, 833) = 1.90, p = .169, ${\eta }_{\mathrm{p}}^2$  = .00).

Next, there was a two-way interaction between time direction and time distance (F(2, 1,666) = 3.79, p = .02, ${\eta }_{\mathrm{p}}^2$  = .00), wherein the effect of time distance on subjective time perception was stronger for past than for future intervals. The two-way interaction between time framing and time distance was not significant (F(2, 1,666) = .26, p = .77, ${\eta }_{\mathrm{p}}^2$  = .00). Importantly, the two-way interaction between time framing and time direction was also not significant (F(1, 833) = .11, p = .744, ${\eta }_{\mathrm{p}}^2$  = .00; time framing effect in the past condition: F(1, 833) = 25.63, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .03, time framing effect in the future condition: F(1, 833) = 30.45, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .04), nor was the three-way interaction between time framing, time direction, and objective distance (F(2, 1,666) = .02, p = .98, ${\eta }_{\mathrm{p}}^2$  = .00). Thus, counter to the polarity congruence account, the year–length effect was equally strong for past and future time points. Figure 2 presents the means across the four between-subjects conditions and three within-subject replicates.

Figure 2.

Subjective Time Perception in Study 2.

Discussion

Study 2 demonstrates that describing time in terms of length rather than year boundaries increases perceived duration, consistent with H₁. The study also shows that the effect applies to both past and future judgments, ruling out the possibility that it arises from numeric polarity congruence. In the next studies, we examine the implications of the year–length effect for consumer evaluations and choice.

Study 3: Financial Investments

Study 3 tests the implications of the year–length effect for investment evaluations. So far, we have shown that length framing increases perceived duration and product valuation in a context where longer perceived duration increases value. Here, we test whether the effect reverses in a context where duration reduces value, thereby examining H_2b.

Holding all else constant, consumers should prefer an investment that produces a return over a shorter period. If length framing increases perceived duration, investments should be evaluated less favorably when the timeframe is expressed in terms of length rather than year boundaries. Study 3 tests this prediction. The preregistration is available at https://aspredicted.org/blind.php?x=BTY_1JF.

Method

Participants

In June 2022, we recruited 450 participants through the CloudResearch MTurk Toolkit. Fourteen participants were excluded based on preregistered criteria, leaving 436 responses for analysis (48.9% female, 50.7% male, .5% nonbinary; M_age = 41.3 years, SD = 12.6).

Design and procedure

The study employed a 2 (time framing: year vs. length, between) × 4 (investment scenarios: two real estate, two stock, within) mixed-factorial design. Each scenario included four attributes: type of investment (e.g., Apple stock), initial investment (e.g., $1,000), present value (e.g., $30,000), and investment timeframe (e.g., “in the year 2007” vs. “15 years ago”). The investment details were held constant across conditions, with return values grounded in real-world figures. Time framing was the only variable manipulated (see Table 5 for overview of scenarios).

Table 5.

Investment Scenarios and Return Evaluations in Study 3.

Investment Scenario Information (Year Framing)	Length Framing	Year Framing	F-Value, p-Value
Amazon stock, invested 7 years ago (in the year 2015),$1,000 initial investment, $8,000 current value	7.08(.12)	7.23(.10)	F(1, 1,122) = 1.38,p = .240
Apple stock, invested 15 years ago (in the year 2007),$1,000 initial investment, $30,000 current value	7.97(.09)	8.28(.07)	F(1, 1,122) = 5.70,p = .017
Condo in New York City, invested 20 years ago (in the year 2002)$10,000 initial investment, $300,000 current value	8.16(.10)	8.45(.07)	F(1, 1,122) = 5.20,p = .023
Beach house in Los Angeles, invested 32 years ago (in the year 1990)$10,000 initial investment, $400,000 current value	8.07(.11)	8.42(.07)	F(1, 1,122) = 7.53,p = .006

Notes: Table 5 shows raw means from Study 3, with standard errors shown in parentheses.

Participants evaluated four investment scenarios, presented one at a time in random order. For example, in one scenario, participants imagined investing $1,000 in Apple stock in 2007 [15 years ago], with a current value of $30,000. They then rated the return on a nine-point scale from “not a great return” to “extremely good return.” After evaluating all scenarios, participants reported demographics and indicated whether they currently invest in stocks or own a home.

Results

Subjective investment return

We analyzed the return evaluations in a mixed-factorial ANOVA with time framing condition (year vs. length) as a between-subjects factor and investment scenarios (four types) as a within-subjects factor. The analysis revealed a significant main effect of time framing (M_length = 7.82, SD = 1.58 vs. M_year = 8.09, SD = 1.24; F(1, 434) = 8.08, p = .005, ${\eta }_{\mathrm{p}}^2$  = .02) and of investment scenario (F(3, 1,302) = 122.83, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .22). The two-way interaction was not significant (F(3, 1,302) = .80, p = .49, ${\eta }_{\mathrm{p}}^2$  = .00); however, the significance level of the time framing effect varied across the four scenarios (see Table 5 for descriptive statistics and simple contrasts by investment scenario).

Discussion

Study 3 demonstrates that investments described using year framing are evaluated more favorably than those described using length framing, consistent with H_2b, likely due to shorter perceived time intervals associated with year framing leading to a perception of higher returns. These findings extend the year–length effect observed in the context of whiskey auctions to the realm of financial investments and demonstrate that length framing is not uniformly beneficial for product evaluations, consistent with our predictions.

Building on the evidence from Studies 1–3, which supported H₁, H_2a, and H_2b, Studies 4–6 put the full conceptual model to the test. Together, they examine (1) the impact of time framing on perceived duration (H₁), (2) its downstream effects on consumer evaluations (H_2a, H_2b), and (3) the indirect path from time framing to evaluation through perceived duration (H₃).

Study 4: Incentive-Compatible Whiskey Choice

Study 4 examines the effect of year–length framing on consumer choice in the context of whiskey selection. Study 1 showed that length framing increased WTP for whiskey in auctions, where consumers may evaluate whiskey bottles one bottle at a time. However, in many retail settings, consumers make choices between different products, and might choose, for example, between a 6-year-old whiskey and a 22-year-old whiskey (length framing) or between a whiskey labeled “2019” and one labeled “2003” (year framing).

Our framework predicts that age gaps will feel larger under length framing (6 vs. 22 years) than under year framing (2019 vs. 2003). Consequently, when age trades off against price, consumers should be more likely to select the older, costlier whiskey in the length condition, and this effect will be driven by perceived age difference. We test these predictions using an incentive-compatible design. The study is preregistered at https://aspredicted.org/WLJ_DNP.

Method

Participants

To enhance ecological validity, we preregistered the study to include only whiskey drinkers. In September 2024, we recruited U.S. adults from Prolific. At the start of the survey, participants reported how often they drink whiskey. Those who responded “never” or “less than once a year” (N = 381) were excluded. The remaining participants (N = 551; 59.1% of those screened) reported drinking whiskey at least a few times per year and continued to the main study. ⁴ As preregistered, we further excluded 82 participants for failing attention or completion checks, yielding a final sample of 469 respondents (45% female, 53% male, 2% nonbinary; M_age = 39.4 years, SD = 11.5).

Design and procedure

Participants were randomly assigned to either the year framing condition (e.g., 2018 vs. 2002) or the length framing condition (e.g., 6-year-old vs. 22-year-old). They were informed that they would be taking part in a whiskey shopping task, where they would be presented with pairs of miniature (50 mL) whiskey bottles. For each pair, participants were asked to select their preferred whiskey. Each pair featured a cheaper, younger whiskey and a more expensive, older whiskey of the same brand (see Figure 3 for an example). Participants made three such binary choices in random order, with each choice involving a different whiskey brand and varying combinations of age and price across trials––Macallan (6 vs. 22 years; $4.50 vs. $6.00), Talisker (5 vs. 14 years; $3.50 vs. $5.50), and Bowmore (8 vs. 20 years; $4.00 vs. $5.50). Bottling year was provided below each bottle in both framing conditions.

Figure 3.

Whiskey Choice Task in Study 4.

After making their whiskey choices, participants rated the perceived age difference between the two whiskeys in each pair. They used a nonnumeric slider scale anchored at “very similar” (0) and “very different” (180) to indicate how similar or different the whiskeys appeared in terms of age. Finally, participants completed demographic questions and an attention check.

Incentive compatibility

At the start of the survey, participants learned that one randomly selected respondent would receive $15 to buy the whiskey they chose in one of the tasks. For instance, if someone selected a $6 bottle, they would receive that bottle plus the remaining $9.

Results

Whiskey choice

We conducted a mixed-effects logistic regression to examine how time framing (i.e., whiskey age expressed as a year vs. length) influenced participants’ whiskey choices. Whiskey choice served as the dependent variable (1 = older, more expensive whiskey, 0 = younger, cheaper whiskey). The model included fixed effects for time framing (year vs. length), whiskey brand (Macallan, Talisker, or Bowmore), and their interaction, along with random intercepts for participants to account for within-subject variability.

The analysis revealed a significant main effect of time framing (M_length = 75.4%, SE = .9% vs. M_year = 64.4%, SE = 1.0%; Wald χ²(1) = 11.39, p < .001), indicating that participants were more likely to choose the more expensive, older whiskey in the length condition. The interaction between time framing and whiskey brand was not significant (Wald χ²(2) = .17, p = .918). Table 6 presents the choice shares across time frames and whiskey brands.

Table 6.

Choice Shares of the Older, More Expensive Whiskey in Study 4.

Miniature Whiskey Pairs (Year Framing)	Length Framing	Year Framing	Wald χ2, p-Value
Macallan 6-year-old (2018) for $4.50 vs. 22-year-old (2002) for $6	80.8%(2.6%)	69.6%(3.0%)	χ2(1) = 8.90,p = .003
Talisker 5-year-old (2019) for $3.50 vs. 14-year-old (2010) for $5.50	68.6%(3.1%)	57.1%(3.2%)	χ2(1) = 8.39,p = .004
Bowmore 8-year-old (2016) for $4 vs. 20-year-old (2004) for $5.50	76.9%(2.8%)	66.7%(3.0%)	χ2(1) = 7.23,p = .007

Notes: Table 6 shows raw choice means from Study 4, with standard errors in parentheses.

Perceived age difference

To examine how time framing influenced participants’ perceptions of the age difference between the two whiskey options, we ran a mixed-factorial ANOVA. Time framing (year vs. length) was included as the between-subjects factor, while whiskey brand (Macallan, Talisker, or Bowmore) was the within-subjects factor. The analysis revealed that participants perceived the age difference to be significantly larger in the length condition (M_length = 124.62, SD = 41.90) than in the year condition (M_year = 108.37, SD = 48.33; F(1, 467) = 22.15, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .05). The interaction between framing and brand was not significant (F(2, 934) = 1.45, p = .23, ${\eta }_{\mathrm{p}}^2$  = .00).

Mediation analysis

To test the effect of time framing on whiskey choice via perceived age differences, we conducted a mediation analysis using the PROCESS macro (Hayes 2017, Model 4). Time framing (length = 1, year = 0) served as the independent variable, the average whiskey choice across the three trials was the dependent variable, and the average perceived age difference between the two whiskey options was the mediator. The results revealed a significant indirect effect of time framing on choice via perceived age difference (indirect effect: B = .13, 95% CI = [.07, .21]; total effect: B = .29, 95% CI = [.11, .47]).

Discussion

Study 4 shows that the year–length effect extends to consumer choice. Participants were more likely to select the older, more expensive whiskey when age was expressed in length terms rather than year boundaries. Importantly, consistent with H₃, this effect was mediated by perceived age difference.

Study 5: Willingness to Pay for Used Goods

Study 5 has three objectives. First, it tests H_2b in the context of secondhand products. Because newer or less-used items are generally perceived as more valuable, we predict that year framing (e.g., “bought in 2023”) would yield higher WTP for secondhand products compared with length framing (e.g., “bought two years ago”), consistent with H_2b. Second, echoing the design of Study 4, Study 5 tests whether perceived duration mediates the effect of time framing on consumer evaluations (H₃).

Third, Study 5 explores alternative explanations for the year–length effect. One possibility is that the observed effects are driven not by numerical compression, but rather by differences in how easily consumers can evaluate time. As noted previously, length framing provides duration information directly (e.g., “five years”), while year framing requires mental calculations to obtain the same information. For instance, determining the age of a whiskey labeled as “distilled in 2020” requires subtracting that year from the bottling year. Similarly, evaluating the age of a furniture item that was bought in 2020 requires subtracting the purchase year from the present year (e.g., 2025 − 2020). This arithmetic burden may make time harder to evaluate under year framing, affecting perceived duration and downstream evaluations. To investigate this, we measured perceived ease of evaluation in this study. We also measured other candidate mechanisms that could be associated with ease of evaluation, including ease of visualization, vividness, concreteness, and familiarity, while also probing perceived time discretization as an alternative account. The study was preregistered at https://aspredicted.org/dnjk-mmsy.pdf.

Method

Participants

In September 2024, we recruited 550 participants from the Prolific U.S. panel. Following preregistered exclusion criteria, ⁵ 38 participants were removed, leaving 512 responses for analysis (51% female, 48% male, 1% nonbinary; M_age = 41.3 years, SD = 12.1).

Design and procedure

The study adopted a 2 (time framing: year vs. length, between) × 3 (age replicate: two-year-old sofa, four-year-old chair, six-year-old dining table; within) mixed-factorial design.

At the beginning of the study, participants were informed that the study has multiple independent sections. In the first section, participants were asked to evaluate three used furniture items—a three-seater sofa, an office chair, and a dining table set—listed on Craigslist and available in their local area, along with age information. The three items varied in age (two, four, or six years old) and were presented in random order, one at a time.

In the year framing condition, the age of each product was described using the purchase year (e.g., “Bought in 2018”), while in the length framing condition, it was described by the number of years since purchase (e.g., “Bought 6 years ago”). Participants reported their WTP for each product in a text box, with possible responses ranging from $20 to $1,000 (see Figure 4).

Figure 4.

Sofa Descriptions in Study 5.

In the next section, participants rated their subjective perception of time for the three intervals from the main task on a 180-point slider scale, similar to the one used in Studies 2 and 4. As in the main task, time was either expressed in terms of year boundaries or length.

At the end of the study, ⁶ we measured six alternative process measures—perceived discretization, ease of evaluation, ease of visualization, vividness, concreteness, and familiarity—adapted from Sokolova (2023). To capture perceived discretization, we asked participants to evaluate the extent to which they perceived the time period as a collection of discrete, individual years, using a seven-point scale (1 = “not at all,” and 7 = “very much”). To illustrate this concept, participants were shown an example: “People can think about time periods as a collection of discrete years. For instance, from the last US presidential election to this year's election, one can think of this period as four individual, discrete years (1 year; 1 year; 1 year; 1 year).” Participants then saw a specific time period—either “2018–2024” or “6 years”, depending on their framing condition—and were asked to rate the perceived discretization of the time period.

On the following page, participants were shown the same time period (“2018–2024” or “6 years”) and were presented with five statements. They were asked to rate their agreement with each statement on a seven-point scale, with the order of statements randomized for each participant. The statements were “The time period above is easy to evaluate,” “The time period above is easy to visualize,” “The time period above is vivid,” “The time period above is concrete,” and “The way the time period is described above is familiar to me.” At the end of the study, participants answered an attention check question and provided basic demographic information.

Results

WTP

To test the effect of time framing on participants’ log WTP for used goods, we ran a mixed ANOVA. Time framing served as the between-subjects factor, age replicate was the within-subjects factor, and log WTP was the dependent variable. The analysis revealed a significant main effect of time framing (F(1, 510) = 9.79, p = .002, ${\eta }_{\mathrm{p}}^2$  = .02), wherein, on average, participants were willing to pay 17% more (on the raw dollar scale) for products described using the year framing compared with the length framing (M_year = 4.73, SD = .90; vs. M_length = 4.57, SD = .88). The interaction between time framing and replicate was not significant (F(2, 1,020) = 1.34, p = .26, ${\eta }_{\mathrm{p}}^2$  = .00).

Subjective time perception

Next, we analyzed participants’ subjective time perception in a mixed ANOVA. The analysis revealed a significant main effect of time framing (F(1, 510) = 8.68, p = .003, ${\eta }_{\mathrm{p}}^2$  = .02), wherein time was perceived as longer in the length condition than in the year condition (M_length = 84.96, SD = 52.38 vs. M_year = 73.92, SD = 55.26). There was also a significant main effect of interval replicate (F(2, 1,020) = 772.08, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .60), with subjective time perception increasing for longer intervals (two vs. four vs. six years). The interaction between time framing and replicate was not significant (F(2, 1,020) = .33, p = .72, ${\eta }_{\mathrm{p}}^2$  = .00).

Alternative accounts and mediation analysis

To test whether any of the rival process measures could account for the year–length effect, we estimated a parallel-to-serial mediation model (PROCESS Model 80). Log-transformed willingness to pay (log WTP, averaged across the three age replicates) served as the outcome. Time framing (year = 1, length = 0) served as the independent variable. Five alternative mediators—ease of evaluation, ease of visualization, vividness, perceived discretization, and familiarity ⁷ —were entered in parallel; all five then fed into subjective time perception (averaged across the three age replicates), which in turn predicted log-transformed WTP (see Figure 5). All continuous variables were standardized.

Figure 5.

Parallel-to-Serial Multiple Mediator Analysis in Study 5.

As Figure 5 shows, none of the indirect paths that ran through the alternative mediators reached significance, with one qualified exception: The pathway Framing → Discretization → Subjective Time Perception → log WTP was statistically significant but negative (B = −.01, 95% CI = [−.01, −.00]), moving in the opposite direction of our theory. By contrast, the focal pathway Framing → Subjective Time Perception → log WTP was positive and significant (B = .03, 95% CI = [.01, .07]). Thus, the analysis confirms that subjective time perception drives the effect of time framing on WTP, but offers little support for ease-based or vividness-based explanations of the effect of time framing on subjective time perception.

Discussion

Study 5 shows that describing a used product's age with a purchase year increases WTP by about 17%. Yet sellers appear not to intuit this benefit. In a pretest with 200 U.S. Prolific participants, 75% chose length framing (“bought 2 years ago”) when asked to create an appealing resale listing, while only 25% chose year framing (“bought in 2022”; see Web Appendix C). This disconnect suggests a missed opportunity: Resale platforms and individual sellers could increase perceived value simply by shifting to year framing.

The study also provides initial evidence against alternative explanations for the year–length effect. Our theorizing suggests that time intervals appear shorter under the year framing because of logarithmic compression of large numerals. However, as year-framed intervals do not present the duration information explicitly, they may also be more difficult to evaluate and less vivid, which could reduce perceived duration. Study 5 provides initial evidence against this account, showing that ease of evaluation, and closely related measures, such as ease of visualization and vividness, did not account for the effect of time framing on perceived duration.

Study 6: Mortgage Choice

Up to this point, we have shown that length (vs. year) framing may improve evaluations (Studies 1 and 4) or lower them (Studies 3 and 5), depending on whether the decision context links duration to higher or lower customer value. In some categories, however, this association varies across consumers, meaning that the same attribute may be viewed as either a benefit or a drawback depending on individual beliefs or preferences.

One such category is mortgages. Many mortgage lenders allow borrowers to fix their interest rate for a specified number of years (Perry 2014). After this fixed-rate period ends, the interest rate typically adjusts based on market conditions. Borrowers who believe that interest rates will rise in the future (or those who are more risk-averse) tend to see longer fixed-rate periods as more attractive. In contrast, borrowers who expect interest rates to fall (or those who are more risk-seeking) may prefer shorter fixed-rate periods, viewing longer fixed-rate periods as less attractive.

Combining predictions from H_2a and H_2b, we propose that length framing will influence both types of borrowers by increasing the relative importance of fixed-rate period duration in their choices. For those who prefer longer fixed-rate periods, length framing should increase perceived differences between fixed-rate terms (e.g., “five years” vs. “ten years”) and enhance the utility and choice likelihood of mortgages with longer fixed-rate periods. For those who prefer shorter periods, the same increase in perceived difference should enhance the negative utility of longer fixed-rate terms and reduce their choice likelihood. As such for both types of borrowers, length framing should amplify the perceived differences between fixed-rate periods and, consequently, increase the relative importance of the fixed-rate attribute in choice. These predictions are consistent with prior work showing that attribute value discriminability increases attribute importance in multi-attribute choice (Aribarg, Burson, and Larrick 2017). In sum, Study 6 tests the effect of time framing on the importance of fixed-rate periods in mortgage choice and examines if this effect is driven by changes in perceived differences across fixed-rate terms.

In addition, Study 6 offers a further test of the ease-of-computation account. In contrast with Study 5, where we asked the participants to rate the ease of evaluation and ease of visualization of time, in Study 6 we directly asked participants how easy or difficult it was to calculate differences between fixed-rate period durations. This allowed us to assess whether ease of calculation could account for the framing-driven differences in mortgage choices. The study preregistration is available at aspredicted.org/kq9r-kmyq.pdf.

Method

Participants

In May 2025, we recruited 401 adults from Prolific's U.K. panel (42.4% female, 57.1% male, .5% nonbinary; M_age = 43.1 years, SD = 11.4). We chose a U.K. sample because U.K. mortgages typically feature a fixed-rate deal for an initial period that later reverts to a variable rate, whereas fully fixed loans dominate the U.S. market, making the U.K. context more ecologically valid for our stimuli (Belgibayeva et al. 2025, p. 2).

Design and procedure

At the beginning of the study, participants were randomly assigned to one of two framing conditions, where the fixed-rate period was described either by year (e.g., “ends in 2030”) or by length (e.g., “five years”). They were asked to imagine that they were looking to buy a house and needed a mortgage. They read that they would evaluate pairs of mortgage offers from different banks and pick the one they preferred in each pair. Each respondent completed 12 binary choice tasks created using Sawtooth Lighthouse Studio's Balanced Overlap CBC algorithm. Every mortgage varied on three attributes: (1) maximum amount to borrow (£300,000, £450,000, £600,000), (2) interest rate (5.0%, 5.5%, 6.0%), and (3) fixed-rate period, which appeared as “ends in 2030, 2035, or 2045” in the year frame and as “5, 10, or 20 years” in the length frame. The total loan term was fixed at 30 years. For each pair of mortgages participants indicated which mortgage they would choose if those were their only options (see Figure 6 for sample choice tasks across the two time framing conditions).

Figure 6.

Sample Choice Tasks Used in Study 6.

Process measures

After making 12 mortgage choices, participants rated how big the time differences seemed for the three pairs of fixed-rate periods from the previous task (e.g., 2030 vs. 2035 in the year frame; 5 vs. 10 years in the length frame) on a seven-point scale (1 = “very small difference,” and 7 = “very big difference”). These three ratings (presented in random order) were averaged to form a composite perceived-time-difference index. Second, for the same three pairs, presented in random order, participants rated “how easy it [was] to determine the difference between the two fixed-rate periods in each pair” on a seven-point scale (1 = “very difficult,” and 7 = “very easy”). The perceived ease ratings were averaged to create a composite ease-of-calculation index. Finally, participants reported their age and gender.

Results

Attribute importance

We estimated individual-specific attribute utilities for maximum amount (300, 450, 600; linear coding), interest rate (5.0, 5.5, 6.0; linear coding), and fixed-rate period (5, 10, 20; linear coding) using hierarchical Bayes procedure in Sawtooth, separately for the year and length conditions. Utility scores were converted into importance scores, as follows:

$$\mathrm{Importance}\mathrm{of}\mathrm{Attribut}{\mathrm{e}}_{\mathrm{m}}={\displaystyle \frac{|{\mathrm{Range}}_{\mathrm{m}}|}{|{\mathrm{Range}}_{\mathrm{amount}\mathrm{to}\mathrm{borrow}}|+|{\mathrm{Range}}_{\mathrm{interest}\mathrm{rate}}|+|{\mathrm{Range}}_{\mathrm{fixed}\text{-}\mathrm{rate}\mathrm{period}}|},}$$

where Range_m = β_m × (Max Level_m − Min Level_m), β_m = estimated utility of attribute m, Max Level_m = maximum level of attribute m (e.g., 20 for the fixed-rate period), and Min Level_m = minimum level of attribute m (e.g., 5 for the fixed-rate period).

To illustrate, Participant A could have a utility of .10 for amount to borrow (i.e., a higher likelihood of choosing “larger amount” mortgages), −10 utility for interest rate (i.e., a higher likelihood of choosing “lower interest rate” mortgages), and −1 utility for fixed-rate period (i.e., a higher likelihood of choosing “shorter fixed-rate period” mortgages). Participant B is similar to Participant A but has a utility of +1 for the fixed-rate period (i.e., a higher likelihood of choosing “longer fixed-rate period” mortgages). The importance of the fixed-rate period for Participants A and B would then be 27.27%:

$${\displaystyle \frac{1\times (20-5)}{[.1\times (600-300)]+[10\times (6-5)]+[1\times (20-5)]}.}$$

Web Appendix C provides additional details on utility and importance score estimation.

We entered the importance scores into a 2 (time framing: year vs. length, between) × 3 (attribute: amount, interest rate, fixed-rate period, within) mixed-factorial ANOVA. The analysis revealed a significant main effect of attribute (F(2, 798) = 50.69, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .11), and no significant main effect of time framing (F(1, 399) = .31, p = .581, ${\eta }_{\mathrm{p}}^2$  = .00).

Critically, there was a significant two-way interaction between attribute and time framing (F(2, 798) = 5.50, p = .004, ${\eta }_{\mathrm{p}}^2$  = .01). Simple contrasts revealed that the importance of the fixed-rate period was higher in the length condition than in the year condition (M_length = 42.62%, SD = 23.40% vs. M_year = 36.39%, SD = 21.59%, F(1, 798) = 9.05, p = .003, ${\eta }_{\mathrm{p}}^2$  = .01). Conversely, the importance of interest rate was lower in the length condition than in the year condition (M_length = 34.72%, SD = 19.50% vs. M_year = 40.34%, SD = 22.45%; F(1, 798) = 7.37, p = .007, ${\eta }_{\mathrm{p}}^2$  = .01). The importance of maximum amount to borrow did not vary significantly across conditions (F(1, 798) = .09, p = .769, ${\eta }_{\mathrm{p}}^2$  < .001). As such, consistent with our predictions, describing fixed-rate periods in terms of length (vs. year) increased the relative importance of the fixed-rate period attribute in mortgage choice. ⁸

Perceived time difference and ease of calculation

One-way ANOVAs on the two indices showed that framing significantly affected perceived differences between fixed-rate periods: participants judged the differences as larger in the length condition than in the year condition (M_length = 5.23, SD = 1.17 vs. M_year = 4.67, SD = 1.01; F(1, 399) = 24.62, p < .001, ${\eta }_{\mathrm{p}}^2$  = .06). Yet, framing did not influence perceived computational ease (M_length = 5.20, SD = 1.44 vs. M_year = 5.03, SD = 1.42; F(1, 399) = 1.35, p = .246, ${\eta }_{\mathrm{p}}^2$  = .003). Hence, while length framing amplifies perceived temporal differences, it does not change perceived ease of calculation.

Mediation analysis

We conducted a PROCESS Model 4 analysis with time framing as the independent variable (length = 1, year = 0), fixed-rate-period importance as the outcome, and perceived time difference as the mediator. The analysis confirmed that there was a significant indirect effect of framing on fixed rate attribute importance via perceived time difference (B = .06, 95% CI = [.01, .13]; total effect = B = .27, 95% CI = [.08, .47]).

We also tested for serial mediation using PROCESS Model 6, where time framing would affect fixed rate attribute importance via ease of calculation (M1) and perceived time difference (M2). We again found a significant indirect effect of time framing on attribute importance via perceived duration (B = .06, 95% CI = [.01, .12]; total effect = B = .27, 95% CI = [.08, .47]), but did not find a significant indirect effect via ease of calculation (B = .01, 95% CI = [−.01, .04]), nor did we find support for the serial mediation path (B = .001, 95% CI = [−.001, .005]).

Discussion

Study 6 shows that framing the fixed-rate period in terms on length (e.g., “fixed for five years”) rather than as a year boundary (e.g., “fixed until 2030”) increased the importance of that attribute in mortgage choice, and this shift was driven by greater perceived differences between fixed-rate periods, consistent with H₃. In addition, Study 6 provides further evidence against the computational ease account: Although length framing increased perceived temporal differences and attribute importance, these effects were not driven by ease of calculation.

Study 7a: Moderation by Scale––Time Perception

With H₁–H₃ were supported by Studies 1–6, Studies 7a and 7b test the proposed logarithmic compression account. Our theorizing holds that the year–length effect arises because larger numbers (e.g., in 2030 in year framing) occupy a more compressed segment of the mental number line, whereas smaller numbers (e.g., in five years in length framing) map onto a more expansive, relatively linear segment. As a result, intervals denoted by large year boundaries feel shorter than equivalent intervals expressed with smaller length numbers. If this account is correct, then temporarily imposing a linear representation of time should attenuate the year–length effect by expanding perceived duration under the year frame. Study 7a tests this prediction. The preregistration is available at https://aspredicted.org/RQ8_LHY.

Method

Participants

In March 2024, we recruited 1,003 participants using the Prolific U.S. panel. Per our preregistered exclusion criteria, we removed 48 participants, leaving 955 participants for the analysis (50.4% male, 47.6% female, 2% nonbinary; M_age = 40.9 years, SD = 13.3).

Design and procedure

The study employed a 2 (time framing: year vs. length, between) × 2 (scale: control vs. linear, between) × 2 (time distance: five and nine years later, within) mixed-factorial design. Participants answered two subjective time perception questions—presented in random order—rating how far the year 2029 (five years later) and the year 2033 (nine years later) felt from the present. Similar to Studies 2, 4, and 5, the time perception questions used a nonnumeric slider scale anchored at “feels very close” (0) and “feels very far” (180).

Importantly, in the control-scale conditions, participants provided their time perception ratings immediately. In the linear-scale superimposition conditions, however, participants first encountered a visual, linear time scale designed to impose linear representation of time, before providing their subjective time perception ratings. Specifically, participants were shown a horizontal timeline representing the relevant interval—either from 2024 to 2035 in the year framing condition or from 0 to 11 years in the length framing condition—with each unit (i.e., year) evenly spaced. They were then instructed to locate the target time point (e.g., “the year 2029” or “5 years from now”) on this scale (see Figure 7). Presenting each unit at identical visual distances encourages a linear mapping of successive time units, counteracting the natural logarithmic compression that larger year boundary numbers often provoke. To reinforce this linear, expansive time interpretation in the year frame and to prevent participants from implicitly starting the mental number line at zero, we anchored the left end of the scale at “NOW” (2024).

Figure 7.

Linear Scale Superimposition in Study 7a.

After marking the correct position on this linear timeline, the participants proceeded to answer the subjective time perception questions. Next, participants completed a fraud detection check and reported their demographic information.

Results

Subjective time perception

The subjective time distance responses were analyzed using a mixed factorial ANOVA, with time framing (year vs. length) and scale (control vs. linear) as between-subjects factors and time distance (five and nine years) as a within-subjects factor. The analysis revealed significant main effects for time framing (F(1, 951) = 19.39, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .02), scale (F(1, 951) = 9.20, p = .0025, ${\eta }_{\mathrm{p}}^2$  = .01), and time distance (F(1, 951) = 1825.13, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .66).

There was also a significant two-way interaction between time framing and scale (F(1, 951) = 7.78, p = .005, ${\eta }_{\mathrm{p}}^2$  = .01). Follow-up simple contrasts revealed that the time framing effect was significant in the control scale condition (M_length = 105.16, SD = 48.50; M_year = 84.13, SD = 48.48; F(1, 951) = 26.17, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .03). Yet, the time framing effect did not manifest when the linear scale was superimposed (M_length = 105.87, SD = 49.73; M_year = 101.16, SD = 50.15; F(1, 951) = 1.29, p = .260, ${\eta }_{\mathrm{p}}^2$  = .00; see Table 7). Moreover, the scale manipulation shifted responses by elongating perceived duration under the year framing (F(1, 951) = 17.08, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .02), but not under the length framing (F(1, 951) = .03, p = .86, ${\eta }_{\mathrm{p}}^2$  = .00).

Table 7.

Subjective Time Perception in Study 7a.

Scale, Time Distance (Year Framing)	Length Framing	Year Framing	F-Value, p-Value
Control, 5 years (2029)	90.2 (3.1)	70.3 (3.0)	F(1, 951) = 21.43, p < .0001
Control, 9 years (2033)	120.1 (2.9)	98.0 (3.0)	F(1, 951) = 28.08, p < .0001
Linear, 5 years (2029)	88.4 (3.1)	84.0 (3.1)	F(1, 951) = 1.04, p = .308
Linear, 9 years (2033)	123.3 (3.0)	118.3 (3.0)	F(1, 951) = 1.40, p = .237

Notes: Table 7 shows raw means from Study 7a, with standard errors shown in parentheses.

Finally, the two-way interaction between time framing and time distance (F(1, 951) = .87, p = .35, ${\eta }_{\mathrm{p}}^2$  = .00) and the three-way interaction between time framing, scale, and time distance were not significant (F(1, 951) = .31, p = .58, ${\eta }_{\mathrm{p}}^2$  = .00).

Discussion

Study 7a provides process evidence for the logarithmic compression account via moderation: Imposing a linear representation of time attenuated the year–length effect. Crucially, the manipulation shifted perceptions only under year framing, leaving length framing unchanged. This pattern aligns with the idea that length-framed durations already lie on a more expansive segment of the mental timeline, whereas year-framed durations fall on a more compressed segment. Thus, making people represent time points on a more linear, expansive timeline in both year and length conditions attenuates the year–length effect.

Study 7b: Moderation by Brand Age––Insurance Valuation

Study 7b again tests our process account through moderation, this time examining downstream consequences for WTP. Building on the idea that consumers often view a brand's age as a signal of consistency, reliability, and quality, leading to a preference for older brands over younger ones (Desai, Kalra, and Murthi 2008), we expect that preferences for older brands will vary depending on how age is communicated. When brand age is expressed in terms of length (e.g., “operating for one year” vs. “operating for four years”) rather than year boundaries (e.g., “operating since 2024” vs. “since 2021”), consumers should perceive larger differences between brand ages, thereby amplifying the older brand's appeal.

Critically, our logarithmic compression account predicts that this framing advantage should weaken when the same age gap is applied to larger absolute ages. As numbers increase, both year and length frames draw on the compressed part of the mental number line, reducing the distinctiveness of the older brand and attenuating the year–length effect. To illustrate, a three-year difference between a one-year-old brand and a four-year-old brand should feel larger than the same three-year difference between a brand operating since 2024 and one operating since 2021. As brand age increases, the difference in time perception between the length frame (e.g., 21 vs. 24 years) and the year frame (e.g., 2001 vs. 2004) should diminish.

Importantly, this moderation logic helps distinguish between the logarithmic compression account and the computational ease account. One could argue that age and age differences are easier to evaluate under the length framing, where age is provided explicitly (e.g., 21 vs. 24 years), than under the year framing (2001 vs. 2004), where determining age requires an extra computation step. If ease of calculation drives the year–length effect, the framing effect should remain unchanged whether brands are relatively young (e.g., 1 vs. 4 years) or old (e.g., 21 vs. 24 years), as ease of calculation is unlikely to change with firm age. By contrast, if the framing effect attenuates, it would support the logarithmic compression account: As brand ages increase, both year and length frames are subject to compression, narrowing perceived differences despite continued differences in ease of calculation. We also measured perceived ease of calculation in this study. The preregistration is available at https://aspredicted.org/93hb-4r9m.pdf.

Method

Participants

In April 2025, we recruited 1,050 U.S. adults from Prolific. Forty-seven participants failed at least one preregistered exclusion criterion, leaving 1,003 responses (52.2% male, 46.6% female, 1.2% nonbinary; M_age = 40.2 years, SD = 13.2).

Design and procedure

The study adopted a 2 (time framing: year vs. length) × 2 (brand-age magnitude: small vs. large) between-subjects design. All respondents imagined purchasing renters insurance for their home. They read that two policies offered identical coverage—up to $25,000 of personal property, $1 million in liability protection, and a $300 deductible—but differed in company age and price. In the small-age conditions, Provider A was one year old (“operating since 2024” or “operating for 1 year”) and Provider B was three years older (“operating since 2021” or “operating for 4 years”). In the large-age conditions, there was the same three-year gap for 21-year-old and 24-year-old providers (see Table 8 for an overview). Provider A's premium was fixed at $50 per year.

Table 8.

Brand Age Values for Insurance Providers in Study 7b.

Brand Age Magnitude	Time Framing
	Length		Year
	Provider A ($50)	Provider B ($?)	Provider A ($50)	Provider B ($?)
Small	1 year	4 years	2024	2021
Large	21 years	24 years	2004	2001

Participants saw two insurers listed side-by-side, their ages (framed per condition), and Provider A's fixed $50 premium. They then entered the maximum annual premium (whole dollars, $40–$1,000) they would pay for Provider B.

On the next screen, participants completed a single-item measure of ease of calculation. The brand age information from the main task was included in the question. For example, those in the year/small-age condition saw the prompt “How easy or difficult was it for you to determine the age difference between the two companies? (i.e., operating since 2004 vs. operating since 2001).” Responses were recorded on a five-point scale (1 = “very difficult,” and 5 = “very easy”). Finally, participants reported their demographic information.

Results

WTP

Log-transformed WTP values were submitted to a 2 (time framing: year vs. length) × 2 (brand-age magnitude: small vs. large) between-subjects ANOVA. The analysis revealed a significant main effect of time framing (F(1, 999) = 20.39, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .02) and of brand age magnitude (F(1, 999) = 86.40, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .08).

Crucially, these effects were qualified by a significant interaction (F(1, 999) = 42.37, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .04; see Table 9). Planned contrasts showed that in the small-age condition, expressing brand age as a length (vs. year) increased WTP by 44% on the raw-dollar scale (M_length = 4.43, SD = .57; M_year = 4.15, SD = .39; F(1, 999) = 59.66, p < .0001, ${\eta }_{\mathrm{p}}^2$  = .06). In contrast, in the large-age condition, the same framing manipulation produced a nonsignificant 8% reduction in raw WTP (M_length = 4.03, SD = .24; M_year = 4.08, SD = .34; F(1, 999) = 2.02, p = .156, ${\eta }_{\mathrm{p}}^2$  = .00).

Table 9.

WTP ($) for Older Insurance Brand in Study 7b.

	Time Framing
Brand Age Magnitude	Length	Year	F-Value, p-Value
Small	$101.00 ($4.30)	$69.90 ($2.50)	F(1, 999) = 59.87, p < .0001
Large	$58.50 ($1.40)	$63.50 ($1.90)	F(1, 999) = 2.02, p = .156

Notes: Table 9 shows raw WTP means from Study 7b, with standard errors shown in parentheses. F- and p-values were obtained from the ANOVA model on log WTP.

Ease of calculation

The same ANOVA on ease-of-calculation ratings revealed no significant effects of time framing (M_length = 4.30, SD = .97; M_year = 4.31, SD = .97; F(1, 999) = .02, p = .882, ${\eta }_{\mathrm{p}}^2$  = .00) or of brand-age magnitude (M_small = 4.31, SD = .95; M_large = 4.30, SD = .98; F(1, 999) = .03, p = .876, ${\eta }_{\mathrm{p}}^2$  = .00). The interaction was also not significant (F(1, 999) = .74, p = .389, ${\eta }_{\mathrm{p}}^2$  = .00).

Discussion

Study 7b shows that the discriminability advantage of length framing depends on the magnitude of the numbers involved. When comparing relatively young brands (one vs. four years old), length wording (“operating for … years”) increased WTP for the older brand relative to year wording. In contrast, when the same age gap was applied to older brands (21 vs. 24 years), framing had little effect on WTP. This boundary condition, together with previous findings, points to logarithmic compression as the mechanism behind the year–length effect. As company age increases, logarithmic compression makes brand age differences appear smaller for both year and length frames, thereby attenuating the effect of time framing on WTP for the older brand. Importantly, we would not expect, nor did we find, that ease of calculation differences between the two frames would diminish with company age. The attenuation of the year–length effect at higher ages thus supports the logarithmic compression account and challenges the computational ease explanation.

General Discussion

Consumers regularly encounter and evaluate time information. It can be information about product age for gourmet foods and alcohol, data on the purchase year of used goods, or information on the time horizon for their investments or loans. Although judgments of time may be common and consequential, the perception of temporal distance—and the judgments that arise from it—can be powerfully influenced by how time intervals are described. Building on prior work on date-delay effects, we demonstrate that framing a time interval in terms of its overall length (e.g., “in 25 years”), versus its year boundary (e.g., “in 2050”), can elongate perceived duration and shift consumer evaluations and choices. We refer to this as the year–length effect.

Across multiple studies, we find that the year–length effect emerges for both past and future judgments of duration (Study 2), in controlled experiments (Studies 2–7b) and in the field (Study 1), and across diverse domains, including whiskey (Studies 1 and 4), financial products (Studies 3, 6, and 7b), and used goods (Study 5). Using a process-by-moderation approach (Studies 7a and 7b), we show that the effect is driven by the logarithmic nature of numerical evaluations: larger boundary-year numbers (e.g., “2050”) are mentally compressed and perceived as closer together, whereas smaller length numbers (e.g., “25 years”) are spaced farther apart, making durations feel longer.

Theoretical Implications

Date-delay effect literature

This work adds to the temporal framing literature. Extant research shows that framing time in terms of overall duration (i.e., length) versus boundaries can increase impatience in intertemporal choice (LeBoeuf 2006; Malkoc, Zauberman, and Bettman 2010; Read et al. 2005) and boost goal adoption (Munichor and LeBoeuf 2018). While these results are consistent with the idea that time feels longer under the length framing, the mechanisms behind date-delay effects remain debated. One account emphasizes a preference for precision, where dates provide a specific point in time, increase certainty of receiving the reward, and thereby increase patience (Read et al. 2005). In the context of goal setting, date framing may make it easier to recall competing obligations (Munichor and LeBoeuf 2018). By contrast, duration framing isolates the interval, allowing individuals to concentrate on the positive outcome of accomplishing their goal, which increases goal adoption. See Web Appendix B for an overview of these process accounts.

Extending this work, we test perceived duration directly as the driver of the year–length effect in consumer decisions. We show that changes in perceived duration drive time framing effects on product preference (Study 4), WTP (Study 5), and attribute importance (Study 6). Unlike prior accounts that emphasize context-specific processes, our findings highlight perceived duration as a general mechanism across diverse decisions. We also explain why length versus boundary framing elongates perceived duration: Drawing on numerical cognition research (Dehaene 2001; Dehaene et al. 2008; Lenkovskaya and Sweldens 2025; Monroe 1971; Thomas and Morwitz 2005), we argue that larger year numbers are subject to logarithmic compression. At the same time, our data provide little support for the intuitively appealing ease-of-computation or vividness explanation put forward by LeBoeuf (2006). This account suggests that length descriptions may prompt consumers to pay more attention to time information and make it more vivid, because these descriptions readily provide objective duration information (e.g., five years, two months, eight weeks), while boundary descriptions require mental calculations to estimate objective duration. Across studies we measured ease of computation (Studies 6 and 7b), ease of evaluation (Study 5), and vividness (Study 5); yet none consistently differed by frame. Our moderation tests also do not support the ease account: Mapping time onto a linear scale (Study 7a) or using longer durations (Study 7b) shrinks the gap between year and length frames, exactly as the compression account predicts, but inconsistent with ease or vividness explanations.

While logarithmic compression offers a parsimonious explanation for the year–length effect, it may not fully account for all forms of length–boundary framing effects, particularly those involving calendar dates. In date–duration comparisons, for instance, dates often feature smaller numerical values than their duration counterparts (e.g., “October 17” vs. “120 days”) and typically blend numerals with words, making the direct application of the numerical compression account less straightforward. More importantly, dates may lead people to represent time in more granular units, such as days instead of months or weeks (e.g., “October 17” vs. “four months”), and may even activate broader temporal scales, such as the annual calendar (Sokolova 2023). In contrast, duration framing tends to evoke narrower temporal scales (e.g., of a month). When these underlying scales shift, so do the subjective spacing between time units and overall duration perception. Thus, while logarithmic compression helps explain year–length effects, it also points toward a broader, scale-based framework of time representation, setting an agenda for future research on length–boundary effects and temporal framing more generally.

Time perception

Our work also contributes to the time perception literature (Malkoc, Zauberman, and Bettman 2010; Shah and Li 2025; Tonietto, Malkoc, and Nowlis 2019; Zauberman et al. 2009). Prior studies show that as objective intervals increase, perceived time compresses: a 20-year horizon feels less than 20 times longer than a 1-year horizon, and 2045 feels less than 20 times farther away than 2026 (Zauberman et al. 2009). We extend this insight by demonstrating that compression can occur purely from framing, even when objective duration is held constant. Expressing time as a year boundary (e.g., “2028”) rather than a duration (e.g., “three years”) shortens perceived time because larger nominal values are compressed on the mental number line. Thus, differences between large numbers feel smaller. While prior work attributes compression to objective magnitude, we show that symbolic numeric format alone—independent of actual duration—can produce comparable effects.

Finally, our findings suggest that time perception is deeply tied to the broader magnitude system. Previous work documents links between numbers and space (the spatial-numerical association of response codes [SNARC] effect; Wood et al. 2008) and between time and space (Kim, Zauberman, and Bettman 2012). Building on this, we show that time, space, and numbers may be processed by a shared system encoding magnitudes symbolically. This supports Walsh's (2003) generalized magnitude theory and aligns with recent findings (e.g., Donnelly, Compiani, and Evers 2022) that symbolic representations can reshape perception across domains. Rather than treating time, space, and quantity as distinct, our results point to a unified representational system in which symbolic framing reshapes magnitude perception—including temporal distance.

Managerial Implications

Our research demonstrates the importance of time framing in marketing communication. The following examples illustrate how our findings can be applied to enhance communication strategies.

Product marketing and firm communications

In industries where age enhances value—such as whiskey, fine art, and specialty foods—careful choice of time framing can meaningfully influence firm outcomes. In the whiskey category, for instance, length framing (e.g., “12 years old”) may highlight maturity, increase WTP, and steer consumers toward older, higher-priced bottles. More broadly, firms can use length framing to reinforce perceptions of longevity and value. Many companies highlight age to signal experience and build trust. As shown in Table 1 and Web Appendix A, some do so using year-based formats (“since 2001”), while others adopt length-based formats (“operating for 24 years”). Our findings suggest the latter format may convey a stronger sense of enduring presence. Thus, stating that a company has been operating “for 24 years” rather than “since 2001” may be more impactful.

By contrast, year framing can be more effective for secondhand goods (e.g., “bought in 2022”), as it can make items feel newer and therefore more valuable. This insight is especially relevant for online marketplaces, where little guidance exists on how to frame product age (see Web Appendix A). Sellers of used goods—who, as our pretest suggests, may naturally default to length framing—could benefit from nudges that align framing with their valuation goals. Given the scale of the global secondhand market, even small framing shifts could generate meaningful economic gains.

In some categories, however, industry norms or logistical constraints limit framing flexibility. Wines, for example, continue to age after bottling, making continuous label updates impractical. As a result, producers typically use year framing (e.g., “2021 Cabernet Sauvignon”). Yet marketers can still incorporate length-based cues through supplemental materials, such as shelf tags or back labels that translate vintage years into elapsed time (e.g., “Bottled in 2021: 4 years old as of 2025”). A similar strategy applies to heritage properties: a listing that reads “Indian Banks, 1699 Colonial-style house” could also include a “326 years old” label, amplifying perceived historic value. Supporting this idea, Supplementary Study 1 shows that combining length-based cues with year framing expands perceived duration compared with year-only framing (see Web Appendix D).

Investor perceptions and financial product preference

Our findings are of relevance for investor communications. Year framing tends to make future outcomes feel closer, which can increase the appeal of investment opportunities. For instance, stating that a venture will be profitable “by 2028” (rather than “in three years”) or that an investment will triple in value “by 2035” (rather than “in ten years”) can reduce perceived delay, highlight forward-looking ambition, and improve evaluations. Conversely, length framing can be more effective for signaling stability and resilience. Saying that a company is funded “for the next five years” (rather than “until 2030”) emphasizes sustained security. Firms can therefore align their framing strategy with the impression they wish to convey—urgency and progress with year framing, or endurance and reliability with length framing.

Time framing can also shape how consumers navigate complex financial decisions, such as selecting a mortgage, car loan, or retirement plan, where time often trades off against interest rates or other attributes. For example, in a low-interest-rate environment, a borrower might compare a lower-rate mortgage fixed for 5 years with a slightly higher-rate mortgage fixed for 10 or 15 years. Similarly, loan offers often differ in both interest rates and the duration of zero-interest introductory periods. Our conjoint experiment (Study 6) reveals that when repayment horizons are presented in terms of length (e.g., “10 years”), consumers place greater weight on time and less on interest rates. This highlights a subtle but powerful lever for financial marketers: framing repayment periods in terms of length rather than year boundaries can shift preferences across loan options.

Broader societal applications

Finally, our findings extend beyond the marketplace. Organizations addressing climate change or other urgent societal challenges can choose frames that evoke urgency. For example, stating a goal as “achieve net-zero carbon by 2040” (rather than “in 15 years”) may make the target feel closer and spur faster engagement. Similarly, framing can guide future-oriented behaviors in domains like retirement savings, health planning, or lifestyle change. When encouraging people to save, plan long-term, or adopt healthier habits, framing time as a fast-approaching deadline (year framing) or as a long window of opportunity (length framing) can meaningfully influence how people perceive time, prioritize, and act.

Supplemental Material

sj-pdf-1-mrj-10.1177_00222437251399115 - Supplemental material for When “Year” Feels Near: How Year Versus Length Framing Alters Time Perception and Consumer Decisions

Supplemental material, sj-pdf-1-mrj-10.1177_00222437251399115 for When “Year” Feels Near: How Year Versus Length Framing Alters Time Perception and Consumer Decisions by Deepak Sirwani, Tatiana Sokolova, and Suzanne Shu in Journal of Marketing Research